Global Health Action Supplement 2, 2010

CONTENTS

Forewords INDEPTH WHO-SAGE study Osman Sankoh 2 The INDEPTH WHO-SAGE collaboration Á coming of age Ties Boerma 3 Guest Editorial The INDEPTH WHO-SAGE multicentre study on ageing, health, and well-being among people aged 50 years and over in eight countries in Africa and Asia Richard Suzman 5 Participating Sites - List of Staff 8 Ageing and adult health status in eight lower-income countries: the INDEPTH WHO-SAGE collaboration Paul Kowal, Kathleen Kahn, Nawi Ng, Nirmala Naidoo, Salim Abdullah, Ayaga Bawah, Fred Binka, Nguyen T.K. Chuc, Cornelius Debpuur, Alex Ezeh, F. Xavier Go´mez-Olive´, Mohammad Hakimi, Siddhivinayak Hirve, Abraham Hodgson, Sanjay Juvekar, Catherine Kyobutungi, Jane Menken, Hoang Van Minh, Mathew A. Mwanyangala, Abdur Razzaque, Osman Sankoh, P. Kim Streatfield, Stig Wall, Siswanto Wilopo, Peter Byass, Somnath Chatterji and Stephen M. Tollman 11 Assessing health and well-being among older people in rural South Africa F. Xavier Go´mez-Olive´, Margaret Thorogood, Benjamin D. Clark, Kathleen Kahn and Stephen M. Tollman 23 Health status and quality of life among older adults in rural Tanzania Mathew A. Mwanyangala, Charles Mayombana, Honorathy Urassa, Jensen Charles, Chrizostom Mahutanga, Salim Abdullah and Rose Nathan 36 The health and well-being of older people in Nairobi’s slums Catherine Kyobutungi, Thaddaeus Egondi and Alex Ezeh 45 Self-reported health and functional limitations among older people in the Kassena-Nankana District, Ghana Cornelius Debpuur, Paul Welaga, George Wak and Abraham Hodgson 54 Patterns of health status and quality of life among older people in rural Viet Nam Hoang Van Minh, Peter Byass, Nguyen Thi Kim Chuc and Stig Wall 64 Socio-demographic differentials of adult health indicators in Matlab, Bangladesh: self-rated health, health state, quality of life and disability level Abdur Razzaque, Lutfun Nahar, Masuma Akter Khanam and Peter Kim Streatfield 70 Health and quality of life among older rural people in Purworejo District, Indonesia Nawi Ng, Mohammad Hakimi, Peter Byass, Siswanto Wilopo and Stig Wall 78 Social gradients in self-reported health and well-being among adults aged 50 and over in Pune District, India Siddhivinayak Hirve, Sanjay Juvekar, Pallavi Lele and Dhiraj Agarwal 88 Health inequalities among older men and women in Africa and Asia: evidence from eight Health and Demographic Surveillance System sites in the INDEPTH WHO-SAGE study Nawi Ng, Paul Kowal, Kathleen Kahn, Nirmala Naidoo, Salim Abdullah, Ayaga Bawah, Fred Binka, Nguyen T.K. Chuc, Cornelius Debpuur, Thaddeus Egondi, F. Xavier Go´mez-Olive´, Mohammad Hakimi, Siddhivinayak Hirve, Abraham Hodgson, Sanjay Juvekar, Catherine Kyobutungi, Hoang Van Minh, Mathew A. Mwanyangala, Rose Nathan, Abdur Razzaque, Osman Sankoh, P. Kim Streatfield, Margaret Thorogood, Stig Wall, Siswanto Wilopo, Peter Byass, Stephen M. Tollman and Somnath Chatterji 96

In addition to the mentorship and editing provided by the Supplement Editors, each paper has been subjected to regular peer review. FOREWORD æ INDEPTH WHO-SAGE Supplement

INDEPTH WHO-SAGE study

y foreword to the first INDEPTH supplement establish a training centre for health and demographic published in GHA, which comprised a series of surveillance systems so that many more scientists from Mpapers by the INDEPTH NCD Surveillance in low- and middle-income countries could be trained in Asia Working Group, stated that the work demonstrated complex longitudinal data analysis techniques. ‘the increasing ability of the INDEPTH Network to On behalf of the INDEPTH Board and myself, I wish to harness the collective potential in Health and Demo- thank the World Health Organization who have been graphic Surveillance Systems in low- and middle-income exemplary partners in this collaboration, and also the key countries to provide a better, empirical understanding of funder, National Institute on Aging and National Institutes health issues of populations under continuous evaluation.’ of Health (NIA, NIH). I wish to highlight the pivotal role In that foreword I also noted, ‘with that collaborative played by Dr. Richard Suzman (NIA, NIH) who ‘was research, we have seen some of the objectives of always there’ as a funder and competent scientist during INDEPTH being achieved: we have strengthened the this collaboration and is the Senior Editor of this Supple- capability of several of our young scientists to conduct ment. I also want to acknowledge the Health and Popula- and analyse longitudinal health and demographic studies; tion Division, School of Public Health, University of the and some of them have become first authors of scientific Witwatersrand, South Africa, for its ongoing role as papers for the first time’ (1). satellite secretariat of the INDEPTH Adult Health and The current supplement, by the INDEPTH Adult Ageing Working Group. This multi-site and multi-country Health and Ageing Working Group, is a compilation of a INDEPTH project has succeeded because of the commit- series of excellent site-specific and cross-site papers, ment and scientific leadership of Professor Stephen Toll- which has reinforced the opinions I expressed previously. man, the leader of the INDEPTH Adult Health and Ageing I feel privileged to be writing these forewords at a time Working Group. Furthermore, I wish to appreciate the when these studies are being completed and their results advice provided by the INDEPTH Advisory Committee through its member Professor Stig Wallat Umea˚University. are being disseminated in scientific publications. The Through resources provided for core institutional INDEPTH WHO-SAGE collaboration started several support to INDEPTH by the Wellcome Trust, Sida/ years ago during the tenure of office of my predecessor, GLOBFORSK, Rockefeller Foundation, Gates Founda- Professor Fred Binka. It was he who provided the initial tion and Hewlett Foundation, we were able to contribute support to the Adult Health and Ageing group, enabling financially to the Adult Health and Ageing Working it to engage with WHO in this partnership. I therefore Group for the successful completion of this work. I was wish to share with him the credit for this success. happy to learn of WHO’s success in securing further I am delighted to have taken part in two key analysis resources from NIA, NIH for a Phase II of these workshops graciously hosted by the Umea˚ Centre for INDEPTH WHO-SAGE studies and, in this regard, Global Health Research, Umea˚ University, Sweden in look forward to our continuing collaboration. 2008, and by the Harvard Centre for Population and The dataset generated by these studies is being made Development Studies, Cambridge, MA, USA in 2010. I am freely available and INDEPTH will encourage wider use also well aware of those previously hosted by the University of the data. of Witwatersrand’s School of Public Health as well as the WHO, more recently in June 2010. I saw INDEPTH Congratulations! scientists presenting their work and taking part in rigorous data analysis, and witnessed exemplary collaboration Osman Sankoh demonstrated by our partners in Umea˚ and Boston. Executive Director, INDEPTH Network They contributed expertise and resources to strengthen the capacities of our scientists to take leading roles in this work. Reference While in Umea˚ and Boston, I saw our colleagues there demonstrating expertise in data analysis and in how to 1. Sankoh O. Foreword. Global Health Action Supplement 1, 2009. interrogate and make sense out of data that had been DOI: 10.3402/gha.v2i0.2085 collected thousands of miles away. That experience made me feel that there was a great need for INDEPTH to

Global Health Action 2010. # 2010 Osman Sankoh. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 2 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5441 FOREWORD æ INDEPTH WHO-SAGE Supplement

The INDEPTH WHO-SAGE collaboration Á coming of age

t is no surprise that there is a lack of evidence on the data. INDEPTH sites have relatively large populations health of older populations in low- and middle- under surveillance with regular monitoring of vital events, Iincome countries. Much current attention is focused which allows the inclusion of a standard short module to on the Millennium Development Goals, prioritising examine health and health-related outcomes in regular maternal and child health and leading infectious diseases. surveillance rounds. In addition, innovative strategies can The epidemiological transition is relatively recent and be developed to link survey and surveillance data to inform health researchers and policy-makers are still grappling larger national estimates as well as developing and testing with the new data demands. And even in high-income strategies for robust small area estimates. In three of the countries, which face increasingly large older populations eight countries with sites Á Ghana, India and South Africa and predominance of chronic diseases, there are major Á reported in this volume of Global Health Action (GHA), evidence gaps. national SAGE studies are ongoing. The set of papers in this Supplement represent a The collaboration will also draw upon the expertise significant step towards better evidence on the health within INDEPTH sites to improve methods in data of older populations. The papers are based on studies collection in older populations in low- and middle- in four African and four Asian countries as part of a income countries. This includes improved recording of collaboration between two multi-country networks. The age, development of verbal autopsy tools to assess the first network is the well-established International Network cause of death in the ageing population, the measurement for the Demographic Evaluation of Populations and Their of health and health-related outcomes for ageing care Health (INDEPTH) in developing countries. It is an providers caring for HIV/AIDS orphans, and the care- international platform of sentinel demographic sites that giving burden and its association with health. Other provides health and demographic data and research to routinely collected demographic data such as migration enable developing countries to set health priorities and and its relationship to health outcomes will also be policies based on longitudinal evidence and includes more essential. Furthermore, some sites have data from other than 30 sites, mostly in Africa and Asia. It has an studies on changing patterns in risk factors and can relate outstanding record of collecting vital statistics and has that to the health status of older adults. been a vehicle for the generation of information on a wide This Supplement to GHA brings together the first set of range of health topics. papers from this collaboration. This set of papers focuses The second network is the World Health Organization on describing the current situation among older people (WHO) Study on Global AGEing and Adult Health and identifies a number of consistent patterns. For (SAGE). SAGE is a multi-country study that addresses health and health-related outcomes and their determinants instance, the health of women among older adults is worse in populations around the world with a focus on low- and than that of men; living alone jeopardises health and well- middle-income countries. The emphasis is on common being; and being poor is bad for health. There are, methodological approaches to ensure cross-population however, important differences within and between sites comparability. SAGE country studies aim for a long- as well. For example, older adults in Vadu, India, who are itudinal cohort design with the inclusion of populations not in a partnership are not as badly off as in other study 50 years and over along with a comparative cohort of sites, probably because of support from extended and persons aged 18Á49 years. The first round has recently been adjoined families; older adults with the poorest health in completed in China, Ghana, India, Mexico, Russia and Purworejo, Indonesia, are clustered in the semi-urban belt South Africa. of the district; and patterns of the older adult population The SAGE and INDEPTH networks have initiated a structure are changing as exemplified by the predominance collaboration to study adult health and ageing in low- and of older men in Agincourt, South Africa and of older middle-income settings. This collaboration offers several women in the slums of Nairobi, Kenya. The results also unique features which will allow both the generation of reveal close relationships between declining health, in- unique evidence and detailed methodological work to creasing disability and worsening of quality of life in the validate self-reported morbidity and survey mortality ageing population.

Global Health Action 2010. # 2010 Ties Boerma. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 3 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5442 Ties Boerma

These first results herald the coming of more substantive This kind of evidence will be increasingly essential to analyses of the complex relationship between non-fatal shape policies and programmes for the health of older health status and subsequent mortality and the factors that populations in low- and middle-income countries. influence that relationship within and across SAGEÁ INDEPTH sites. Ties Boerma, Director This unique collaboration between WHOÁSAGE and Health Statistics and Informatics the INDEPTH Network will lead to ongoing efforts to World Health Organization follow these populations over time, to look at longitudinal Geneva, Switzerland changes in the key outcomes of interest and their predictors.

4 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5442 GUEST EDITORIAL æ INDEPTH WHO-SAGE Supplement

The INDEPTH WHO-SAGE multicentre study on ageing, health and well-being among people aged 50 years and over in eight countries in Africa and Asia

his supplement to Global Health Action presents standardised data for the four African and four Asian the first results from the INDEPTH WHO-SAGE country sites presented in this Supplement represent a Tmulticentre study, comprising background infor- significant advance on previously available information mation (1), site-specific results (2Á9) and an overall for charting the evolution of the demographic and multicentre analysis (10). Reporting on one of the first epidemiological transitions in low-income countries. cross-national studies of ageing in Africa and Asia, this Two decades ago, there was a distressing paucity of supplement might be termed historic, especially when demographic, economic, and health data on adult health coupled with the demographic circumstances of popula- and ageing for low-resource countries (12). Most of the tion ageing, and the simultaneous public release of the available data were cross-sectional. However, longitudinal microdata from the eight sites. According to a UN studies, most especially ones that combine health and projection, the world is only a few years away from a economic status data within the same study, are needed to historic watershed Á when for the first time in human understand many of the dynamics of ageing. To remedy history those aged 65 and over will outnumber those the abysmal lack of information on older populations in under age 5 (11). Awareness of population ageing and its low-income countries, the U.S. National Institute on consequences is by now quite widespread in European Ageing (NIA), a component of the National Institutes of policy circles; but the issue is only just reaching the radar Health (NIH), commissioned a series of reports screens of most low-income nations. What steps should on ageing in developing countries from the U.S. Bureau low-resource countries take (and when), in advance of the of the Census (13, 14), and the U.S. National Academy of demographic, epidemiologic, and economic transitions Sciences (15). Although as recently as 1990 almost all associated with population ageing? Industrialised nations industrialised societies also suffered from a lack of experienced population ageing after they became wealthy; adequate data (especially longitudinal), significant pro- most low-resource countries will have to cope with this gress has since been made in establishing nationally transition prior to becoming wealthy. Minimal attention representative longitudinal studies, such as the Health has been given to the dynamics of health and their and Retirement Study USA (HRS), the English Long- economic consequences in developing countries, which itudinal Study on Ageing (ELSA), and the Survey of are now among the fastest ageing nations. To date, the Health, Ageing and Retirement in Europe (SHARE). attention of global institutions has been riveted almost These surveys, with their data on health and economic solely on children rather than the needed dual focus on status, cognitive functioning, and biological assessment both groups of societies’ dependents: children and older are transforming several areas of social and behavioural people. Unfortunately, no manual exists to guide the science (16). Over the past several years, NIA has preparations of nations at different levels of development encouraged efforts to develop nationally comparable or stages of the demographic ageing transition, and representative studies in low-resource countries. We are governments have to navigate without adequate maps or now seeing successes in developing comparable and GPS systems. While the demographic changes occur over coordinated national surveys in countries such as Mexico a timeline measured in decades, the development of new (MHAS), China (CHARLS), and in earlier stages, India institutions and systems, including sound pension and (LASI). Additionally, the NIA, in concert with WHO, insurance systems, need to be set up decades in advance of seized the opportunity to develop a network of low-cost any transition. The long-term costs of public sector adult health and ageing-related surveys that piggy-backed pensions in Africa are already giving rise to expressions on the World Health Survey. The network, known as the of anxiety in some financial circles. The results from the Study on Global AGEing and Adult Health (SAGE), has

Global Health Action 2010. # 2010 Richard Suzman. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 5 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5480 Guest Editorial

fielded studies in Ghana, South Africa, India, China, ageing circles was that while modern medicine could Mexico, and Russia. INDEPTH WHO-SAGE resulted delay death, it could not prevent or delay the onset of from an opportunity to field a standardised set of surveys degenerative diseases, which could not be treated effec- of adult health and ageing in eight INDEPTH health and tively. Most believed that increases in old age longevity demographic surveillance system (HDSS) sites, with the would lead to a pandemic of disability, with disabled life survey content drawn heavily from the SAGE, SHARE, expectancy increasing substantially. However, an impor- and HRS surveys. While there have been a number of tant finding was that in the United States, between 1982 cross-national surveys focusing on ageing in Asia, this is and 2001, disability among those aged 65 and over the first involving sub-Saharan Africa. declined by 25%, demonstrating the substantial plasticity From the beginning, the INDEPTH network of socio- of individual ageing (18). More recently, concern has demographic surveillance sites offered significant poten- been rising that the epidemic of obesity will lead to tial for understanding health and demographic processes substantially increased disability, offsetting the gains. As within low-income countries, most especially within rural life expectancy increases in these middle and low-income areas. The addition of new survey data on adult health countries, no one knows yet whether disabled life and ageing to the data portfolio of the INDEPTH sites expectancy will outpace healthy life expectancy, or significantly enhances the value of the surveillance sites whether there will be any compression of morbidity and themselves, and adds value to the survey data through disability, especially if onset starts later in life. linkage to the rich local epi-demographic history and The collection of data on the same individuals in later context created by the INDEPTH sites. The new survey waves of INDEPTH WHO-SAGE will allow researchers data also substantially enhance the capacity of the to investigate a whole set of questions not amenable to Network to evaluate or assess the impact of policy analysis within the current cross-sectional data. Long- interventions, such as the establishment or major mod- itudinal data are needed to tackle a variety of questions ification of pension or health systems. Further, the ability posed by the authors of this supplement. Answering to compare the results of three of the INDEPTH WHO- questions such as how chronic disease-related disability SAGE sites [South Africa (2), India (9), and Ghana (5)] evolves, how long individuals with specific diseases with the nationally representative SAGE surveys for survive, whether self-reported health predicts survival those countries will provide the opportunity to assess better than the health score, or how living arrangements the generalisability of INDEPTH WHO-SAGE small- and widowhood affect health and well-being, require area results for these three countries. panel data. Longitudinal data are also needed, for In 1996, the Global Burden of Disease project made example, to identify the mechanisms by which old age the remarkable projection that within a few decades, non- pensions can improve the health and general welfare of communicable disease would outpace infectious diseases grandchildren if part of the pension is distributed to as a cause of morbidity and mortality in all regions of the those grandchildren. Similarly, in the absence of a globe (17). Although the projected epidemiological randomised trial, longitudinal data would be essential transition was largely a function of population ageing, to assess the impact of pensions on the health of the implications of these projections were largely ignored. pensioners Á do old age pensions that end when the INDEPTH WHO-SAGE will become an important pensioner dies improve the health and well-being of the observatory of the epidemiological transition in low- pensioner? If so, is it by means of increasing pensioners’ income countries. The introductory article in this supple- ability to purchase food and health care, or is it because ment (1) clearly shows that at baseline, the four they feel more needed by their family, or do their families INDEPTH WHO-SAGE Asian countries (Viet Nam, take better care of them to keep the pension income Bangladesh, Indonesia, and India) have moved further flowing? It is therefore important that the current toward the relative predominance of non-communicable samples are followed up regularly and that every effort disease than the African countries (South Africa, Tanza- is made to track individuals during the interim periods Á nia, Kenya, and Ghana). Based on the experience of a strength of health and demographic surveillance Á in industrialised nations, the projected increase in degen- order to ensure a high response rate for these follow-ups. erative non-communicable diseases that tracks increases The decision to release the microdata simultaneously in adult life expectancy will be accompanied by an with this supplement, via the Global Health Action Web increasing loss of physical and cognitive functioning site (http://www.globalhealthaction.net), is a noteworthy and growing levels of disability. The increase in disability milestone for INDEPTH and will be a great boon for will result in reduced capacity for work among older research on adult health and ageing in the respective workers, loss of autonomy, and the need for substantial countries. Cross-national research in both developing and care in old age, which is enormously costly in terms of developed countries has been seriously hampered by slow both economics and well-being. During the 1980s in the release of microdata, sometimes more than a decade after United States, the prevalent view in epidemiological and collection, and sometimes not ever as in the case of the

6 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5480 Guest Editorial

first WHO cross-national survey on ageing conducted countries: the INDEPTH WHO-SAGE collaboration. Global around 1979Á1980. The tension between speedy data Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0. release and the desire of the data collectors to hold onto 5302. 2. Go´mez-Olive´ FX, Thorogood M, Clark BD, Kahn K, Tollman the data until they have had a chance to fully mine those SM. Assessing health and well-being among older people in data laboriously collected from the study they had rural South Africa. Global Health Action Supplement 2, 2010. designed, is perhaps greatest today in low-income coun- DOI: 10.3402/gha.v3i0.2126. tries of Africa and Asia. However, in order to justify the 3. Mwanyangala MA, Mayombana C, Urassa H, Charles J, very considerable expense of cross-national longitudinal Mahutanga C, Abdullah S, et al. Health status and quality of life among older adults in rural Tanzania. Global Health Action studies, costs of the data need to be amortised over as Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142. many secondary data projects as possible, and the research 4. Kyobutungi C, Egondi T, Ezeh A. The health and well-being of products must also become useful to policy makers as older people in Nairobi’s slums. Global Health Action Supple- soon as possible. Science requires replication, and the lack ment 2, 2010. DOI: 10.3402/gha.v3i0.2138. of data sharing can slow down research and the produc- 5. Debpuur C, Welaga P, Wak G, Hodgson A. Self-reported health tion of policy-relevant results. It has been the experience and functional limitations among older people in the Kassena- Nankana District, Ghana. Global Health Action Supplement 2, of studies such as HRS, ELSA, and SHARE that such 2010. DOI: 10.3402/gha.v3i0.2151. longitudinal studies catalyse new fields of social and 6. Minh HV, Byass P, Chuc NTK, Wall S. Patterns of health status behavioural science and coalesce whole groups of re- and quality of life among older people in rural Viet Nam. searchers around the studies’ data, forming new scientific Global Health Action Supplement 2, 2010. DOI: 10.3402/ communities. In this case, every effort should be made to gha.v3i0.2124. get these data as rapidly as possible to pre- and post- 7. Razzaque A, Nahar L, Khanam MA, Streatfield PK. Socio- demographic differentials of adult health indicators in Matlab, doctoral students and junior faculty of at least the eight Bangladesh: self-rated health, health state, quality of life and countries involved in the study. At the same time appro- disability level. Global Health Action Supplement 2, 2010. DOI: priate efforts must be made to maintain the ethics of data 10.3402/gha.v3i0.4618. confidentiality, ensuring that respondent anonymity is not 8. Ng N, Hakimi M, Byass P, Wilopo S, Wall S. Health and quality breached, especially since these studies were all conducted of life among older rural people in Purworejo District, in specific and known geographic areas, which makes the Indonesia. Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125. protection of anonymity more challenging. 9. Hirve S, Juvekar S, Lele P, Agarwal D. Social gradients in self- The agreement by the INDEPTH WHO-SAGE prin- reported health and well-being among adults aged 50 and over cipal investigators to conduct the study with the under- in Pune District, India. Global Health Action Supplement 2, standing that the data would be speedily released is highly 2010. DOI: 10.3402/gha.v3i0.2128. commendable, and one can predict that the dividends to 10. Ng N, Kowal P, Kahn K, Naidoo N, Abdullah S, Bawah A, et al. Health inequalities among older men and women in Africa the study will perhaps be greater than the INDEPTH and Asia: evidence from eight Health and Demographic team imagines. Surveillance System sites in the INDEPTH WHO-SAGE study. Commendations and acknowledgements are due to Global Health Action Supplement 2, 2010. DOI: 10.3402/ several institutions and groups, including the INDEPTH gha.v3i0.5420. leadership, WHO staff, faculty at Umea˚ and Harvard 11. Kinsella K, He W US. Census Bureau, International Population who facilitated important data analysis workshops for Reports, P95/09-1, An aging world. Washington, DC: U.S. Government Printing Office; 2009 pp. 7Á8. INDEPTH WHO-SAGE, and the many peer reviewers 12. Kinsella K, Suzman R. Developing dimensions of population involved in this supplement. aging in developing countries. Am J Human Bio 1992; 4: 3Á8. 13. Kinsella, K. Aging in the third world. CIR staff paper, No. 35, Richard Suzman Feb 1988. ix, 80 pp. U.S. Bureau of the Census, Center for Division of Behavioral and Social Research International Research, Africa and Latin America Branch: National Institute on Aging Washington, DC. National Institutes of Health 14. Kinsella K, He W US. Census Bureau, International Population Bethesda, MD, USA Reports, P95/09-1, An aging world. Washington, DC: U.S. Government Printing Office; 2009. 15. Cohen B, Menken J. Aging in sub-Saharan Africa: recommen- All views expressed in this editorial are entirely those of dations for furthering research. Washington, DC: National the author, and do not necessarily reflect those of NIA or Academies Press; 2006. NIH. 16. Suzman R. Epilogue: Research on population aging at NIA: retrospect and prospect. Popul Dev Rev 2004; 40: 239Á64. 17. Lopez AD, Murray CJL. The global burden of disease, 1990Á2020. Nat Med 1998; 4: 1241Á3. 18. Cutler DM, Wise DA, editors. National Bureau of Economic References Research. Health at older ages: the causes and consequences of declining disability among the elderly. Chicago, IL: University of 1. Kowal P, Kahn K, Ng N, Naidoo N, Abdullah S, Bawah A, Chicago Press; 2009. et al. Ageing and adult health status in eight low-income

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5480 7 Participating Sites æ INDEPTH WHO-SAGE Supplement

The INDEPTH WHO-SAGE multicentre study was only possible because of the hard work of many staff at each participating site, as well as the authors of papers in this Supplement:

Agincourt, South Africa John Killian Jane Wahake Gitonga Hector Dhlamini Athuman Kipembe James Hotendo Victoria Dlamini Celsius Kipinga David Ireri Regan Gumede Charles Kuwonga David Otieno Juma Simon Khosa Amoses Kyovecho Gedion Kennedy Juma Glory Khoza Mary Lazaro Maureen Kadogo Thoko Khensani Machavi Nassoro Likumi Deborah Kagai Muziwakhe Solly Maluka Silivanus Lisoadinge Adan Kalicha Olga Mambane Zuhura Lungombe Phanuel M. Kasuni Nash Manzini Emanuel Luvanda Joel Kasyoka Sinah Manzini Jacob Lyanga George Kidiga Merriam Perseverence Maritze Sauda Magubikira Catherine Kimatu Lawrence Pedney Mashale Stephen Magwaja Joshua Musila Kivonge Ishmael Mashigo Ishamel Athuman Makanganya Esther Nyambura Macharia Ntanga Moses Mathabela Albert Masalu Mary Marubu Phanuel Mathebula Isaya Mashinga Catherine Mbalu Council Mbetse Shabani Matengana Kennedy Mose Momanyi Warren Mdluli Madunda Mkalimoto Geoffrey Ndungu Mondia Gordon Mkhabe Ally Mpangile David Karuga Muhika Obed Mokoena Raphael Msabana Wanjiru Murigi Linneth Mthetho Edimund Msalabule Stanley Murithi Violet Ndlovu Mshamu Mshamu Samuel Mutuma Simon Delly Ndzimande Bernadi Mwambale Hawa Hassan Mwangangi Sizzy Ngobeni Elisha Mwandikile Damaris R Mwangi Vusi Ngwenyama David Bonaventura Mwarabu Grace Mumbua Mwania Morris Sibuyi Simbani Mwikola Booker Ndayhaya Busisiwe Sibuyi Abdala Mwinshehe Henry Ndungu Morris Mdawu Sibuyi Joseph Mwonja Deborah Nganga Promise Sibuyi Honesta Mzyangizyangi Moses Mwithiga Ngugi Ellah Sihlangu Ellah Mwanaid Ngagonja Esther Wanjiru Njeri Bernard Silaule Calstus Ngalanga Jedidah Njeri Phamela Nombulelo Tibane Msafiri Ngalisoni Melchizedek Nyakundi Nomsa Ubisi Jonson Ngenga Thomas Ondieki Nyandika Mwadawa Ngumbi Peter Nyongesa Ifakara, Tanzania Joseph Njavike Audrey Achieng Ocholla Novatus Chagodola Hadija Nyanga George Ochieng Oduor Deogratias Chamanga Amina Salumu Clement Oduor Yassin Chikoko Joyce Shayo David Ouma Ojuka Timoth Chogo Athumani Utwakumwambu Mildred Adhiambo Onyango Panga Husein Peter Onyango Godwin John Nairobi, Kenya Evaline Achieng Otteng Lukresia Kadungula Mohammed Ali Meshack Odede Owino Luitfrid Kaduvaga Callen Bwari Benson Mbithi Peter Tukae Kapati Wekesah Murunga Frederick Jacqueline Ratemo Gonzaga Kasanga Abduba Salesa Galgalo Sarah Nabalayo Simiyu Sophia Kayera Anthony Chomba Gathuita Ruth Waithera Wairimu Godfrey Kidege Antony Kagiri Gichohi Peter Agutu Waka

Global Health Action 2010. # 2010 List of participating staff This is an Open Access article distributed under the terms of the Creative Commons Attribution- 8 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5493 List of participating staff

Moses Wanyama Michael Banseh Phung Thi Dinh Philip Kibet Wendot Emma Chiratogo Tran Thanh Do Abdikadir Adan Yarrow Afia Damwura Phuong Thuy Duyen Abdikadir Adan Yarrow Everest Dery Nguyen Thi Ngoc Ha Atinganne Dominic Dinh Cong Ha Navrongo, Ghana Theresa Fumjegeba Phung Thi Hai Diana Abagale Yeji Godwin Do Thi Thanh Hien Catherine Abakis Francis Gweliwo Nguyen Phuong Hoa Irene Abase Mohammed T Ibn-Salia Nguyen Thi Mai Huan Aboyinga Abokiya Memuna Issaka Hoang Thi Hue Gana Abongbe Fatima Issaka Chu Phi Hung Anthony Achana Dauda Ahmed Jadeed Nguyen Quoc Hung Bawotua Kwoyire Adda Martin Kambonga Do Manh Hung Clare Addah Joana Kampoe Bui Thi Huong Sixtus Addah Edmond Kanyomse Nguyen Thi Huong Martin Adiga Joseph Katasuma Nguyen Thi Thanh Huong Emefa Adiku Mac Kolley Ngo Thi Huyen Ophelia Adjei Felix Kondayire Nguyen Thanh Huyen Charles Adongo Fati Kumangchira Nghiem Thi Hy Immaculate Adongo Ferreol B. Lagejua Nguyen Van Lam Beatrice Afobiku Richard Latinga Ngo Thi Lien Akwia Agangba Jerry Atua Lucas Phan Thi Thanh Lieu Charles Agangzua Christina Luguchura Giang Thi Tuyet Loan Evelyn Agomah Rose Mary Luguyimang Truong Hoang Long Timothy Ajubala Rita Luguzuri Nguyen Thi Luyen Felicia Akanlassi Ziblim Mahama Nguyen Thi Ly Emmanuel Akantosige Luuse Matholomew Nguyen Thi Nguyet Minh James Akayasi John Memang Phung Thi Minh Roger Akobga Ayangba A. Mensah Nguyen Binh Minh Isaac Akumah William Minyila Phung Thi My Jacob Anabia Ismail M Mohammed Nguyen Thi Duy Na Robert Kwotera Ane Abangba Moses Phan Thi Nang Christopher Aniwe Anastasia Musah Phung Thi Nga Rufina Anoah Maxwell Naab Nguyen Thi Minh Nham Scholastica A-Oho Vitus Nabengye Tran Thi Nhan Raphael Apana Andrews Opoku Nguyen Thi Nhung Mathew Apatinga Rose Parese Dinh Thuy Nhung Vida Apayire Lucy Pelabia Phuong Thi Nhung Freda Apee Boniface Pwadurah Tran Thi Kim Oanh Thompson Apempale Habibatu Salifu Doan Thi Hoang Oanh Peter Asobayire Andriana Sumboh Phung Thi Thu Phuong Gilbert Asuliwono Felicity Titigeyire Tran Thi Mai Phuong Rita Asumboya Patience Tito Dao Dinh Sang Justina Asumboya Francis Yeji Nguyen Thi Sinh Martina Atenka Yahaya Zulhaq Nguyen Thi Thanh Tam Ajentio Atulugu Nguyen Thi Tam Joana Awineboya Filabavi, Viet Nam Tran Thi Tha Francis Awineboya Dang Thi Minh Anh Bui Thi Thanh Thao Tamgomse Ayaam Nguyen Thi Be Phung Thi Thanh Thao Peter Ayangba Nguyen Thi Ngoc Bich Nguyen Thi Thu Denisia Ayibello Le Thi Thanh Binh Dang Thi Hong Thuy Akua Ayirewora Quach Thi Thanh Binh Nguyen Thi Thuy Raymond Azagisiya Phung Thi Chien Nguyen Thi Thuyet Jesse Jackson Azambugi Nguyen Thi Dau Phung Thi Tinh

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5493 9 List of participating staff

Tran Khanh Toan Joko S. Titik Rahayu Phung Thi Toan Juana Linda Tri Atmi Nguyen Thanh Tu Kartini Tri Wahyu Nguyen Lan Viet Khotib Subhan Tri Wantoro Khuat Thi Xuan Kusen Utari Marlinawati Lasmi Wahyu Fatmawati Matlab, Bangladesh Ledjar Warsiyah Ali Ahmed Lidya Hastuti Winarti Halena Akhter Lilik Dewanti Wisnu Abul Kalam Azad Mintorowati Yekti Utami Jiabonessa Begum Muhtadi Yudha Prastika Nazma Khanam Murtiyah Yunardi Bahadur Mia Nur Wicaksono Yusmiyati Shirajum Munira Nurtiyah Yusuf Samira Akhter Sultana Pitoyo Purnawati Vadu, India Purworejo, Indonesia Puspita Handayani Kalpana Agale Abdul Wahab Ratih Widayanti Jyoti Bhosure Agung Nugroho Ratna Bharat Choudhari Ami Rumhartinah Retno Handayani Nilam Fadtare Ardiyanti Robert Arian Datusanantyo Shilpa Fulaware Arif Rosyid Budiman Prashant Gaikwad Bambang Sukma Widadi Rustiningsih Vijay Gaikwad Budi Hartiningsih Ruwayda Tejashri Ghawte Devie Caroline Sendy Trupti Joshi Didi Yudha Prastika Siti Aminah Deepak Mandekar Didik Fery Kristianto Siwi Rahmawati Anita Masalkar Djaswadi Dasuki Sri Purnaningsih Sayaji Pingale Dwi Lestari Priastuti Sri Suryani Ratan Potdar Dwi Rosmalawati Sugeng Somnath Sambhudas Eka Yuli Astuti Sugeng Dinesh Shinde Eko Setianto Sugun Eni E. Suharyani Secretariat Erry Ariyanti Sujarwo Raymond Akparibo Fahruddin Sukarman Sixtus Apaliyah Fatma Yunita Sukirman Zubeida Bagus Feri Budiarto Sumarta Sadiya Ooni Hafsah Tahir Supriyo Pratomo Dereshni Ramnarain Haryanto Sutaryo Jackie Roseleur Hendras Bintar Teguh Imam Titus Tei Hendro Budi Teguh Rohaji Birgitta A˚ stro¨m Irfan Cahyadi Tetra Ita Saraswati Tetra Lintang

10 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5493 æINDEPTH WHO-SAGE Supplement Ageing and adult health status in eight lower-income countries: the INDEPTH WHO-SAGE collaboration Paul Kowal1,2*, Kathleen Kahn3,4,5#, Nawi Ng4,5,6#, Nirmala Naidoo1, Salim Abdullah5,7, Ayaga Bawah5, Fred Binka5, Nguyen T.K. Chuc5,8, Cornelius Debpuur5,9, Alex Ezeh5,10, F. Xavier Go´ mez-Olive´ 3,5, Mohammad Hakimi5,6, Siddhivinayak Hirve5,11, Abraham Hodgson5,9, Sanjay Juvekar5,11, Catherine Kyobutungi5,10, Jane Menken12,13, Hoang Van Minh5,8, Mathew A. Mwanyangala5,7, Abdur Razzaque5,13, Osman Sankoh5, P. Kim Streatfield5,13, Stig Wall4#, Siswanto Wilopo5,6, Peter Byass4#, Somnath Chatterji1 and Stephen M. Tollman3,4,5#

1Multi-Country Studies Unit, World Health Organization, Geneva, Switzerland; 2University of Newcastle Research Centre on Gender, Health and Ageing, Newcastle, NSW, Australia; 3MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa; 4Centre for Global Health Research, Epidemiology and Global Health, University of Umea˚ , Umea˚ , Sweden; 5INDEPTH Network, Accra, Ghana; 6Purworejo HDSS, Faculty of Medicine, Gadjah Mada University, Yogyakarta, Indonesia; 7Ifakara Health Institute, Ifakara, Morogoro, Tanzania; 8FilaBavi HDSS, Faculty of Public Health, Hanoi Medical University, Hanoi, Viet Nam; 9Navrongo HDSS, Navrongo, Ghana; 10African Population & Health Research Center, Nairobi, Kenya; 11Vadu Rural Health Programme, KEM Hospital Research Centre, Pune, India; 12University of Colorado, Boulder, CO, USA; 13Matlab HDSS, ICDDR,B, Dhaka, Bangladesh

Background: Globally, ageing impacts all countries, with a majority of older persons residing in lower- and middle-income countries now and into the future. An understanding of the health and well-being of these ageing populations is important for policy and planning; however, research on ageing and adult health that informs policy predominantly comes from higher-income countries. A collaboration between the WHO Study on global AGEing and adult health (SAGE) and International Network for the Demographic Evaluation of Populations and Their Health in developing countries (INDEPTH), with support from the US National Institute on Aging (NIA) and the Swedish Council for Working Life and Social Research (FAS), has resulted in valuable health, disability and well-being information through a first wave of data collection in 2006Á2007 from field sites in South Africa, Tanzania, Kenya, Ghana, Viet Nam, Bangladesh, Indonesia and India. Objective: To provide an overview of the demographic and health characteristics of participating countries, describe the research collaboration and introduce the first dataset and outputs. Methods: Data from two SAGE survey modules implemented in eight Health and Demographic Surveillance Systems (HDSS) were merged with core HDSS data to produce a summary dataset for the site-specific and cross-site analyses described in this supplement. Each participating HDSS site used standardised training materials and survey instruments. Face-to-face interviews were conducted. Ethical clearance was obtained from WHO and the local ethical authority for each participating HDSS site. Results: People aged 50 years and over in the eight participating countries represent over 15% of the current global older population, and is projected to reach 23% by 2030. The Asian HDSS sites have a larger

#Supplement Editor, Kathleen Kahn, Editor, Nawi Ng, Chief Editor, Stig Wall, Deputy Editor, Peter Byass, Supplement Editor, Stephen M. Tollman, have not participated in the review and decision process for this paper.

Global Health Action 2010. # 2010 Paul Kowal et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 11 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 Paul Kowal et al.

proportion of burden of disease from non-communicable diseases and injuries relative to their African counterparts. A pooled sample of over 46,000 persons aged 50 and over from these eight HDSS sites was produced. The SAGE modules resulted in self-reported health, health status, functioning (from the WHO Disability Assessment Scale (WHODAS-II)) and well-being (from the WHO Quality of Life instrument (WHOQoL) variables). The HDSS databases contributed age, sex, marital status, education, socio-economic status and household size variables. Conclusion: The INDEPTH WHOÁSAGE collaboration demonstrates the value and future possibilities for this type of research in informing policy and planning for a number of countries. This INDEPTH WHOÁ SAGE dataset will be placed in the public domain together with this open-access supplement and will be available through the GHA website (www.globalhealthaction.net) and other repositories. An improved dataset is being developed containing supplementary HDSS variables and vignette-adjusted health variables. This living collaboration is now preparing for a next wave of data collection. Keywords: ageing; survey methods; public health; burden of disease; demographic transition; disability; well-being; health status; INDEPTH WHO-SAGE

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘SAGE data’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 18 May 2010; Revised: 6 July 2010; Accepted: 8 July 2010; Published: 27 September 2010

he ageing of populations is often considered as a older ages across countries has been increasing in global public health success, but results in many recent years. The potential benefits of cross-national Tensuing challenges, particularly in lower- and studies of ageing that enable us to understand the nature middle-income countries where societies did not grow of demographic and epidemiological transitions have wealth before growing old, as in higher-income countries. been widely recognised (12, 13). The US Health and Societal ageing will affect economic and health systems in Retirement Study (HRS) and other notable surveys, all nations, including the ability of states and societies such as the English Longitudinal Study on Ageing to both maintain contributions from and also provide (ELSA), have provided the necessary evidence base to resources for older population groups. begin to address the needs and contributions of older But will population ageing affect lower- and higher– persons in higher-income countries. However, the ma- income countries in similar ways? The projected jority of older persons now and into the future will macroeconomic and health impacts from longer life reside in lower-income countries where the evidence base expectancies have only recently become clearer for is very limited. higher-income nations (1Á5); but few non-Organization The HRS and ELSA studies, and more recently the for Economic Cooperation and Development (OECD) World Health Organization’s (WHO) multi-country countries have the data to determine if extended longevity Study on global AGEing and adult health (SAGE), coincides with healthier lives until older ages (that is, a have also been used as the basis for harmonisation with compression of morbidity). Unlike wealthier countries, other national studies and many cross-national compar- the existing formal social protection systems in most isons. Longitudinal ageing studies are critical to develop lower-income countries cover only a small proportion of the evidence base to better understand ageing processes the older population (6); however, if we believe in and adult health dynamics, especially in countries with demographic dividends, lower-income countries will limited mortality data due to poorly functioning or low have a long lead period to collect data which can be coverage of vital registration systems. They have parti- used to inform economic and health systems (7). Burden cular advantages in their ability to examine multiple of disease shifts from maternal/child health and acute exposures, determinants and outcomes, and to measure communicable diseases to chronic infectious and non- relationships over time: all essential aspects for under- communicable diseases in lower-income countries will standing ageing across different contexts. However, while challenge health systems without the data necessary to critical to research, policy and planning, longitudinal inform policy and planning (8Á11). studies are resource and time intensive. Interest in the measurement and comparability of The extent to which lower-income countries have adult health, the ageing process and well-being at begun to generate and use critical evidence for an

12 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 Ageing and adult health status in eight lower-income countries

effective health response has been slow and suboptimal in Navrongo, Ghana; Asia: Filabavi, Viet Nam; Matlab, many countries (14). This lack of evidence is particularly Bangladesh; Purworejo, Indonesia and Vadu, India) and prominent in low- and middle-income countries, partly provides another valuable data collection platform for because the demographic transitions have been relatively cross-national comparisons of ageing. The NIA BSR was recent and also because political will and financial instrumental in bringing the two groups together from support have not been sufficient. Combining standar- the outset and has provided technical guidance through- dised survey modules with existing surveillance infra- out in combining survey and surveillance data collection structures, especially systems collecting vital registration efforts to fill needed data gaps on ageing and adult details, offers a unique opportunity to reduce research health. WHO SAGE, the INDEPTH Adult Health and costs and efficiently collect needed data in low- and Ageing Working Group, the NIA and the eight partici- middle-income countries. pating INDEPTH HDSS sites have developed a colla- If populations in any country are to age well, an boration built on these survey and surveillance data improved understanding of ageing processes, of resilience collection platforms. This included health and well-being factors for well-being, and of the determinants of health survey data collected within or parallel to HDSS house- status (HS) across countries are needed. This knowledge hold (HH) census update rounds and linked socio- will in turn inform health care and social protection demographic household data. While this initial dataset policies and planning. Results from a collaboration is cross-sectional, there are plans to include longitudinal between the WHO-SAGE survey platform and the HDSS data and further waves of data collection using an International Network for the Demographic Evaluation adapted summary version of the SAGE instrument in the of Populations and Their Health in developing countries HDSS sites. This will significantly enhance the value of (INDEPTH), involving Health and Demographic the collaboration and resulting datasets by tracking Surveillance Sites (HDSS) in eight countries (four changes over time in the same population samples and African and four Asian) will provide HS, disability and relating them to health determinants, predictors and well-being results for ageing and adult health in South outcomes, such as mortality in older adults. An introduc- Africa, Tanzania, Kenya, Ghana, Viet Nam, Bangladesh, tion to the countries, HDSS sites and research methods Indonesia and India. Data collection included methods follows. to improve cross-country comparability, thereby provid- ing a basis for comparisons with data from higher-income Setting the stage countries, such as the US Health and Retirement Study and the ELSA. This article describes the back- Country characteristics ground to the INDEPTH WHO-SAGE collaboration The ongoing demographic shift provides concrete evi- and introduces the methods used to generate the first dence that most countries will be faced with an increas- wave of results which includes site-specific analyses and Á ingly old or ageing population Á the challenge is for cross-site comparisons. national and international health communities to use available data to best prepare for these changes. At Background present, 62% of older persons reside in less developed The WHO’s Multi-Country Studies unit, with the sup- countries and this is projected to increase to almost 80% port of the US National Institute on Aging’s Behavioral by 2050 (15). and Social Research Program (NIA BSR), has imple- Table 1 includes the estimated and projected total mented multi-country ageing and adult health studies to populations and proportions of older adults for the world fill data gaps in lower-income countries and has worked and participating INDEPTH countries in 2009 and 2030. to improve cross-national comparability with available The World Bank income category is also included for data. WHO’s SAGE conducts nationally representative each country, with a mix of five low- and three middle- household health surveys in six countries, with direct income countries (16). In 2009, over 281 million people links to an additional 14 countries through various aged 50 years and over resided in the eight nations collaborations. SAGE is guided by an international included in this collaboration, which constitutes 20% of expert Advisory Committee and coordinated from the global population in that age group (15). Similarly, WHO’s Multi-Country Studies unit. In addition, com- 18% of the global population aged 60 and over lives in parisons with ageing research in higher-income countries, these eight countries. These proportions will increase to such as the US HRS, English ELSA and the pan- 23% and 21%, respectively, by 2030. Over the same time European Survey of Health, Ageing and Retirement in period, the percentage of the population aged 0Á14 years Europe (SHARE) are ongoing. in these countries will drop from 29.9 to 28.5% and five of WHO’s collaboration with INDEPTH has generated the eight countries will have a larger proportion of data from HDSS sites in eight countries (Africa: Agin- persons aged 60 and over than under 15 years by 2050 court, South Africa; Ifakara, Tanzania; Nairobi, Kenya; (the four Asian countries and South Africa). Overall, the

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 13 Paul Kowal et al.

Table 1. Population totals and proportions of older adults for the world and by INDEPTH country, in 2009 and projected to 2030

2009 2030

Country income Country categorya Total, Nb 50, N (%) 60, N (%) Total, N 50, N (%) 60, N (%)

World 6,829 1,379 (20.2) 737 (10.8) 8,309 2,283 (27.5) 1,370 (16.5) Sub-Saharan Africa 843 110 (10.9) 54 (5.3) 1,308 157 (12.0) 78 (5.9) South Africa UMI 50 8 (15.0) 4 (7.1) 55 10 (19.1) 6 (11.1) Tanzania Low 44 4 (9.5) 2 (4.8) 75 8 (10.6) 4 (5.3) Kenya Low 40 3 (8.8) 2 (4.1) 63 7 (11.5) 3 (5.5) Ghana Low 24 2 (11.2) 1 (5.7) 35 5 (15.3) 3 (7.7) Asia 4,121 785 (19.1) 400 (9.7) 4,917 1,398 (28.4) 821 (16.7) Viet Nam Low 88 15 (17.2) 6 (8.6) 105 32 (30.6) 19 (18.2) Bangladesh Low 162 20 (12.9) 10 (6.0) 203 46 (22.9) 23 (11.3) Indonesia LMI 230 40 (17.4) 20 (8.8) 271 79 (28.9) 43 (16.0) India LMI 1,198 187 (15.6) 89 (7.4) 1,485 343 (23.1) 185 (12.4) Pooled INDEPTH 1,836 281 (15.3) 135 (7.3) 2,293 531 (23.2) 286 (12.5) country (8) totals aWorld Bank country income category: Low, low income; LMI, lower-middle income; UMI, upper-middle income. bN in millions (,000,000). Sources: UN Population Division (15) and World Bank (16). percentage increase in population aged 60 will grow collaboration (Fig. 1). South Africa’s burden profile is more in the African than Asian countries. exceptional here because as an upper-middle income With ageing populations and increasing life expectan- country, a lower communicable disease burden is expected; cies, countries will inevitably see changing population however, the massive HIV/AIDS burden clearly shifts the disease burdens. Burdens of disease, risk factors and burden distribution. Similarly, despite being a lower- patterns of injury are changing through a complex income country, Viet Nam has a comparatively lower combination of evolving social, demographic, health, communicable disease burden. political and economic processes. Diseases thought to Shifting to morbidity, the top three contributors to be the domain of higher-income countries are now morbidity burdens in middle-income countries in 2004 significant causes of morbidity and mortality in a number were unipolar depressive disorders, ischaemic heart dis- of lower- and middle-income countries (17Á19). ease and cerebrovascular disease (20). The top three for The most recent Global Burden of Disease (GBD) lower-income countries were lower respiratory infections, 2004 update includes distributions of mortality and diarrhoeal diseases and HIV/AIDS. Fig. 2 illustrates the morbidity by three major groupings: (Group I) commu- burden of non-fatal health outcomes by major grouping nicable diseases, maternal health and nutrition; (Group II) non-communicable diseases; and (Group III) violence 100% and injuries. The 2004 update incorporates revisions and 90% new data working from the initial 1990 GBD (20). The 80% 1990 GBD results estimated 44% of total burden was 70% Group I, 41% for Group II and 15% for Group III 60% worldwide (21). These data show that even in 1990, NCDs 50% were a significant contributor to mortality rates. Fig. 1 Violence/injury 40% Non-communicable shows the distributions of fatal disease burden by geo- 30% Communicable graphic grouping and country for 2004. Preliminary results 20% indicate a substantial increase in the proportion of deaths 10% due to non-communicable diseases from 59% in 2002 to 0% 69% in 2030 (19). All the participating Asian HDSS sites had higher NCD rates than the 1990 estimates Á and Indonesia had a much higher Group III burden. Countries Africa Asia that are at an earlier phase of the demographic transition typically have a higher mortality burden from Group I Fig. 1. Mortality profiles (age-standardised death rates) by conditions, and this is more clearly the case for the African major Burden of Disease grouping and country, 2004 (WHO countries participating in the INDEPTH WHOÁSAGE 2008).

14 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 Ageing and adult health status in eight lower-income countries

and INDEPTH HDSS site country for 2004, indicating SAGE characteristics those conditions which lead to longer years of life lived in The SAGE project (http://www.who.int/healthinfo/ a state of less than full health (non-fatal health outcomes systems/sage) has become a leading multi-country study or disability). The figure illustrates the mixture of disease on ageing and adult health in lower- and middle-income burden in the participating low- and middle-income countries. Launched in 2003 as part of the WHO’s World countries, with Group I burden featuring more promi- Health Survey (WHS), SAGE has implemented nation- nently in African countries and Group II in Asian ally representative population surveys in six core coun- countries. Still, a majority of the main chronic conditions tries: China, Ghana, India, Mexico, the Russian predominate in older age groups in both regions (19). Federation and South Africa. The specific aims of From currently available data, the overall contribution of SAGE are to: disability from non-communicable diseases is projected to . Obtain reliable, valid and comparable data on levels of grow substantially and ageing will be one of the major health on a range of key domains for older adult drivers of the burden (22). populations. . Examine patterns and dynamics of age-related changes in health using longitudinal follow-up of HDSS characteristics survey respondents as they age, and to investigate INDEPTH (http://www.indepth-network.org) is a net- socio-economic consequences of these health changes. work of 37 sites in 19 countries in Africa, Asia, . Supplement and cross-validate self-reported measures Central America and Oceania based on health and of health and the anchoring vignette approach to socio-demographic surveillance within defined areas. improve comparability of self-reported measures, The network brings together virtually all of the world’s through measured performance tests for selected HDSSs located in low- and middle-income settings, and health domains. currently covers over 2 million individuals. Regular . Collect data on health examinations and biomarkers household census updates at each HDSS site allow for to improve reliability of data on morbidity, risk factors continuous, household-level monitoring of all vital events and monitor effect of interventions. (births, deaths and migrations) in the defined population. INDEPTH provides an exceptional resource with which The baseline data collection for SAGE (Wave 0) was to characterise the health, demographic and social conducted as part of the 2002/2003 WHS with SAGE dynamics of some of the world’s most vulnerable Wave 1 data collected between 2007 and 2010. Biennial populations. The INDEPTH Adult Health and Ageing longitudinal follow-up is planned with Wave 2 in 2011 Working Group has established INDEPTH’s capability and Wave 3 in 2013. to contribute critical insights into the adult health, ageing SAGE provides data on the levels and differences in and disease transitions evolving in Africa and Asia, and health and well-being across low- and middle-income to use this understanding to inform policy and evaluate countries, and methodologies that improve health mea- interventions of potentially high impact. surement and cross-national comparability. SAGE covers a broad range of topics, with a focus on health, disability, 100% risk factors, stress, happiness, social networks, economic 90% well-being, care-giving, health care utilisation and health 80% systems responsiveness. Furthermore, a host of biomar- 70% ker data was collected, including anthropometrics, phy- 60% sical performance tests and dried blood spots. 50% Another objective for SAGE is to develop working Violence/injury relationships and linkages to other data collection plat- 40% Non-communicable 30% Communicable forms, including surveys and surveillance sites, to better understand changing health over the life course, 20% compression of morbidity and perceptions of health, 10% quality of life and economic well-being within and across 0% countries. SAGE has a history of collaborating with other ageing research, like the US HRS; ELSA; SHARE; China Health, Ageing, Retirement Longitudinal Study; Africa Asia Longitudinal Ageing Study in India; and, now with Fig. 2. Morbidity profiles (age-standardised DALYs) by INDEPTH HDSS sites. The collaboration with IN- major Burden of Disease grouping and country, 2004 DEPTH extends the possibilities of longitudinal house- (WHO 2008). hold-based research through the combination of survey

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 15 Paul Kowal et al. and surveillance methods and provides opportunities to Methods apply new methodological techniques to cross-country The initial step was to develop a health and well-being ageing research. module that could be nested within a typical census update round in an INDEPTH HDSS site. This meant that the interview needed to be approximately 15Á20 min The collaboration in duration with minimal impact on interviewers and The collaboration between WHO-SAGE and INDEPTH respondents. A meeting between WHO and INDEPTH has pursued four main goals: (1) to develop and at the University of the Witwatersrand, South Africa in implement a survey instrument that can be incorporated 2006 was used to examine psychometric properties of the into a regular census update round placing minimal health and quality of life sections of the SAGE survey additional demands on existing research infrastructure; instrument based on results from the 2005 SAGE pilot (2) to implement the full SAGE survey in parallel to a study (n1,500) conducted in Ghana, India and Tanza- summary short survey round, but with separate infra- nia, to determine priorities, to outline the scope of the structure and resources; (3) to determine key areas where working relationship and to invite interested HDSS sites INDEPTH HDSS sites could be used as methodological to participate. During the meeting, the survey instru- laboratories to pilot new methods and test hypotheses Á ments and results from the SAGE pilot were reviewed so as to exploit the complementary strengths of both with commentary from each INDEPTH HDSS site. The survey and surveillance data; and (4) to derive more group then worked together to create a shortened integrated analytical plans to assess ageing and adult summary version of the full SAGE questionnaire (the health at national and sub-national levels. INDEPTH WHOÁSAGE instrument, available as a For this article, we address goals 1 and 4 above using a supplementary file to this article, including variants of summary version of the full SAGE instrument which was vignettes) which consisted of questions on HS and implemented in eight INDEPTH HDSS sites. This part vignettes, functioning and subjective well-being. This of the collaboration had two primary aims. The first was summary questionnaire was subsequently piloted in to use survey and surveillance data to describe the each HDSS site in 2006/2007 before implementing the situation of ageing and adult health within and across full data collection. Pilot results and interview debriefings participating HDSS sites. This included the adaptation were used to refine and finalise the standardised ques- and implementation of standardised SAGE survey mod- tionnaire to be used across all HDSS sites. This version ules on health and wellbeing in INDEPTH HDSS sites. was then translated and back-translated in local lan- The HDSS sites identified overlapping content in their guages using translation protocols from both the WHS respective surveillance data and the SAGE survey instru- and INDEPTH HDSS sites. ments. HDSS sites then worked to enhance the compar- Standard interview protocols, training curricula ability of the socio-demographic data collected at each (including a DVD with video clips of example interviews) site to be included in a cross-site dataset (for example, and quality assurance procedures were used across all comparing socio-economic status indicators and map- HDSS sites. Training sessions with experienced inter- ping education levels to an international standard). The viewers were conducted for survey teams at each HDSS second aim was to determine the feasibility of collecting site. These training sessions lasted an average of 4.5 days. longitudinal data through combining the two types of The interview teams had the added advantage of long- data collection efforts as a means to establish ageing and standing relationships within the surveillance sites. adult health trends in a range of countries. A first step Face-to-face interviews with participants aged 50 and was to develop a survey instrument adapted from the full over were conducted in the course of the regularly SAGE questionnaire that could be inserted into a regular scheduled census in three HDSS sites. Separate survey census round without significant disruption to the infra- activities were used in five HDSS sites, where in one site it structure and process. The belief was that the potential was part of a broader ageing survey (Nairobi). Feedback increase in efficiency from adding modules to the regular from the survey teams indicated that it took about data collection rounds, coupled with new analytical three weeks to become maximally efficient at interviews techniques, could provide data on changing health and and data collection. Across all the sites, the mean well-being at a reduced cost whilst retaining the strengths interview time, excluding vignettes, was 20 minutes of both surveillance and survey data. These data would towards the end of the survey process. This was about then be used to inform the design of interventions 14 minutes less than the average time at the beginning of addressing vital aspects of older adult health and the interview process. The vignettes took an average of functioning and, importantly, have the potential to be 13 minutes of interview time, again, the time decreasing monitored more frequently within the HDSS sites than from an average of 19 minutes at the beginning of the with the national-level surveys. process.

16 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 Ageing and adult health status in eight lower-income countries

Vignette methodology Ethical clearance was obtained from research review Cross-national comparative data analysis enhances un- boards local to each participating HDSS site (several of derstanding of HS differences, ageing dynamics and which are linked to universities), plus from the WHO cultural differences, but also facilitates the evaluation of Ethical Review Committee as part of SAGE. Informed the performance of health, social and economic systems, consent was obtained from each respondent prior to and policies to address ageing and health. Typically, the interview. measurement of HS relies on self-reported responses in Sample: Six HDSS sites collected data from the entire surveys and the self-response data take the form of population aged 50 in their HDSS. Sampling in the two ordered categorical (ordinal) responses. Eight domains remaining HDSS sites (Navrongo, Ghana and Matlab, of health were used, which account for up to 80% of the Bangladesh) was based on random selection of persons variance in HS (23). As part of the WHO cross-country aged 50 and over within the HDSS site. For comparison health survey approach, anchoring vignettes have been purposes, a smaller sample of younger adults (aged used to position self-reported responses onto a common 18Á49, n5,794) was interviewed in five HDSS sites scale comparable across individuals. An anchoring vign- using similar methods. ette is a description of a concrete level on a given health Questionnaire: The abbreviated survey instrument domain that respondents are asked to evaluate with consisted of two modules adapted from the full SAGE the same questions and response scales applied to questionnaire: the HS and associated vignette questions self-assessments on that domain. plus Activities of Daily Living (ADL)-type questions A concrete example of the HS questions and vignettes (following the WHO Disability Assessment Scale version for one health domain, mobility, follows: II (WHODAS-II) model), and questions on subjective Female respondent X is asked two questions about her well-being as measured by the 8-item version of the WHO own level of mobility, Quality of Life (WHOQoL) instrument (28). Some HDSS sites chose to add additional modules and/or questions, Q1 Overall in the last 30 days, ‘Was it none, mild, moderate, but the primary goal was a standardised questionnaire how much difficulty did you severe, extreme or cannot do that could be applied in all HDSS sites embedded within have with moving around? this?’ existing HDSS census rounds. Q2 In the last 30 days, how ‘Was it none, mild, moderate, Additional data targeted for inclusion into the final much difficulty did you have severe, extreme or cannot do dataset, and deriving directly from the HDSS, included in vigorous activities? this?’ socio-demographic characteristics, such as age, sex, education, marital status, socio-economic status and household information, such as the number of household Next the respondent is asked to respond to questions members. about the vignettes. Vignettes are brief stories that describe a certain fixed level of health, with five vignettes covering a range of mobility levels. The respondent is Dataset instructed to put herself in the shoes of the person Following site-level data entry and cleaning, and after a described in the vignettes and answer the same question data-sharing agreement was reached between the partici- as if she were that person: pating INDEPTH HDSS sites and with WHO, data were [Someshni] has a lot of swelling in her legs due to her forwarded to a central location (Umea˚, Sweden) for health condition. She has to make an effort to walk cleaning and imputation of missing data. Regular corre- around her home as her legs feel heavy. spondence between HDSS sites improved the efficiency of the data checking and cleaning process. A working Q3 How much difficulty did ‘Was it none, mild, moderate, meeting held in 2008 at Umea˚ University, Sweden, was [Someshni]have with moving severe or extreme or cannot used to harmonise data across the sites, finalise the around? do this?’ dataset and agree on initial outputs. A first dataset was generated and included: Q4 How much difficulty did ‘Was it none, mild, moderate, [Someshni] haveinvigorous severe or extreme or cannot activities? do this?’ . Comprehensive HH information including roster of all members (by age, sex, marital status, education, location (urban or rural), HH head) and socio- By mapping responses to various questions on the same economic status. health domain to a common comparable scale, anchoring . For each respondent: age and date of birth, sex, vignettes may provide a bridge between data collected marital status and education. across cultures or population sub-groups [further detailed . From the adapted SAGE modules: overall general self- information about anchoring vignettes and statistical rated health, HS from eight domains plus related models is available elsewhere (24Á27)]. vignette information, functioning assessment from the

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 17 Paul Kowal et al.

12-item WHODAS and subjective quality of life Functional status (WHODAS) results from the 8-item WHOQoL. Self-reported functioning was assessed through the stan- . Plans to archive the data at WHO, INDEPTH and the dardised 12-item WHO Disability Assessment Scale, University of Michigan’s National Archive of Com- Version 2 (WHODAS) (35). It is a well-tested instrument, puterized Data on Aging (NACDA) to maximise with published psychometric properties and a good opportunities to share data and provide multiple predictor of global disability (36Á38). The WHODAS is access portals. compatible with the International Classification of Func- tioning, Disability and Health (ICF) and contains many of The four main outcome variables derived from this data the most commonly asked ADL and Instrumental Activ- and reported in the site-specific and cross-site articles in ities of Daily Living (IADL) questions. The WHODAS this issue are self-rated general health (SRH), overall HS, instrument also provides an assessment of severity of disability levels (WHODAS) and subjective quality of life disability (39). Results from the 12-items were summed to (WHOQoL). get an overall WHODAS score, which was then trans- formed to a 0Á100 scale, with 0 as best functioning (no Overall general self-reported health (SRH) disability) and 100 maximum disability. Two overall general health questions were asked, each with 5-point Likert-type response scales. The first is a Subjective well-being and quality of life (WHOQoL) question asked very often in surveys: ‘In general, how An 8-item version of the World Health Organization would you rate your health today? Would you say, very Quality of Life instrument (WHOQoL) was used to good (1), good (2), moderate (3), bad (4) or very bad assess perceived well-being (28). This is a cross-culturally (5)?’; and the other was a question related to general valid instrument for comprehensively assessing overall difficulties in day-to-day tasks: ‘Overall in the last 30 subjective well-being, yet is also very brief. Knowing that days, how much difficulty did you have with work or health and quality of life are strongly associated yet household activities? Was it, none (1), mild (2), moderate distinct concepts, WHOQoL will help describe the (3), severe (4) or extreme/cannot do (5)?’ These types of relationship in older persons across countries and over global measures of self-rated health are commonly used time. Results from the 8-items were summed to get an in health surveys and as measures of population health. overall WHOQoL score which was then transformed to a At the individual level, the global self-rated health 0Á100 scale, similar to the health score. question is a good predictor of many health and health- related outcomes (29, 30). However, the true meaning of Implementation results responses to a single question for a multi-dimensional Eight INDEPTH HDSS sites collected data using the construct and the reliability of this measure over time has summary questionnaire (see Table 2). Sample sizes been questioned (31, 32). ranged from almost 2,100 to over 12,000, with a total combined sample of over 46,000 persons aged 50 and Health status (HS) over. Additionally, a random sample of persons aged 18Á Health scores were calculated based on self-reported 49 was included in five HDSS sites Á as a comparison health in eight health domains covering affect, cognition, population Á but these were not included in the initial interpersonal activities and relationships, mobility, pain, dataset or analyses. self-care, sleep/energy, and vision. Each domain included The survey took an average of 4.7 months to complete at least two questions. Asking more than one question with a range of 3Á8 months. Five sites implemented the about difficulties in a given domain provides more robust survey as a stand-alone effort, with the three remaining assessments of individual health levels and reduces HDSS sites (Navrongo, Ifakara and Agincourt) imple- measurement error for any single self-reported item. menting the survey as part of a scheduled census update. Item response theory (IRT) was used to score the Two of these three HDSS sites finished on schedule, with responses to the self-reported health questions using a the one site requiring additional time and staff to partial credit model which served to generate a composite complete the census and survey. HS score (33, 34). An item calibration was obtained for each item. In order to determine how well each item Discussion contributed to common global health measurement, chi- square fit statistics were calculated. The calibration for Platform for research on adult health and ageing each of the health items was taken into account and the In light of the projected demographic and epidemiologic raw scores were transformed through Rasch modelling transitions associated with an ageing world, a WHO and into a continuous cardinal scale where a score of INDEPTH collaboration has demonstrated the capacity 0 represents worst health and a maximum score of to generate data across African and Asian settings to 100 represents best health. better understand health outcomes and their determi-

18 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 Ageing and adult health status in eight lower-income countries

Table 2. Selected features of participating HDSS sites: INDEPTH WHO-SAGE study, 2006Á2007

Approximate HDSS site populations Study population

Total 50 Anticipated study Final study Year Periodicity of Total years population, all population HDSS site Country started census updates population and over ages 50 years and over

Africa Agincourta South Africa 1992 Annually 70,000 8,400 6,500 4,085 Ifakaraa Tanzania 1996 Every 4 months 84,000 9,400 5,000 5,131 Nairobia Kenya 2000 Every 4 months 69,000 2,700 2,700 2,072 Navrongoa Ghana 1993 Every 4 months 144,000 22,900 5,000 4,584 Asia Filabavib Viet Nam 1999 Every 3 months 50,000 8,500 8,500 8,535 Matlaba Bangladesh 1966 Every 2 months 212,000 33,800 5,000 4,037 Purworejob Indonesia 1990 Annually 53,000 14,200 14,200 12,395 Vadua India 2003 Every 6 months 68,000 8,000 8,000 5,430 Totals 750,000 107,900 54,900 46,269 aSupport from the US National Institute on Aging. bSupport from Swedish Council for Working Life and Social Research. nants in older adult populations. The initial results from persistent and emerging health challenges for older the collaboration between WHO-SAGE and INDEPTH populations, and to monitor and evaluate health and HDSS sites are a milestone for longitudinal research on social programmes to determine what works, assess how ageing and adult health and provide an exceptional specific programmes are performing and inform decisions platform for multi-site and multi-country, longitudinal regarding programme design and implementation. research on ageing and adult health in lower-income countries in Africa and Asia. The data collection platform has the potential to Limitations and difficulties substantially enhance the applications of findings from As with any longitudinal study, problems were experi- both survey (SAGE) and surveillance-based (INDEPTH) enced with locating respondents to be included Á data collection. The very nature of the HDSS sites, with especially men, many of whom may be migrant labourers. geographic boundaries defining their populations, along Interviewers found difficulty in questioning the oldest with established infrastructure and human resources, old, even after training and increased awareness about the present a number of opportunities for methodological potential issues with interviewing this population seg- development and hypothesis testing prior to scaling to a ment. In addition, difficulties were experienced with national-level survey. A number of topics could be explaining the vignettes, some of which included scenar- explored, such as the relationships between morbidity, ios possibly foreign to rural settings. As part of the well-being, social networks and mortality, because of the analysis of results, response patterns to the vignette documentation levels and frequency of contact. Similarly, questions would clearly indicate if, in the end, a respondent did not understand the vignettes. surveillance sites benefit from enhanced generalisability of results, expansion of objectives and comparability to other survey data, to name a few. Additionally, the Feasibility of longitudinal monitoring of adult health methodological and practical strengths of each are and ageing accentuated, resulting in improved financial efficiencies Although we aimed to assess the feasibility of incorpor- for conducting longitudinal ageing research. ating the INDEPTH WHO-SAGE short questionnaire The collaboration will also support data harmonisa- into routine HDSS activities, only three of the eight sites tion, data management and analytic capacity develop- attempted this, with the other five sites conducting the ment, cross-validation and calibration of measures, survey as a separate field activity. Of the three HDSS sites contextualisation of the detailed information from integrating the survey, one found need for additional time HDSS within broader national patterns and trends, joint and staff. Interviewers needed time to gain experience efforts to disseminate results and consideration of their interviewing older respondents and to develop strategies policy implications. for high-quality interviews: the average duration of The analysis of levels, trends and differentials in interviews, excluding vignettes, decreased on average by leading health problems globally is needed to identify 14 minutes from about 34 minutes at the beginning of

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 19 Paul Kowal et al.

the interviews to 20 minutes towards the end. The average Further exploration of results using small area analyses length ended up at about 28 minutes. and optimising the combination of survey and surveil- In general, sites found value in targeting the age group lance data are needed. of 50 and over, focusing on health rather than routine Finally, another wave of data collection is planned, for HDSS questions, linking INDEPTH WHO-SAGE data which funding was recently secured. Further hypothesis with existing HDSS variables and subsequent health testing can be undertaken to take advantage of the outcomes. Any further data collection efforts will seek unique panel data that the ongoing surveillance systems to shorten the questionnaire further; incorporate the provide. For example, differences in functioning at older survey modules into the routine census round; provide ages given different socio-economic and health transition more training to implement the vignettes; and interview environments may be explored in cross-site comparisons. the entire population under surveillance rather than using The contrast, for instance, between the leading health a sample, where possible. problems in Navrongo, Ghana, which remain dominated by many persistent ‘pre-transition’ challenges (infectious Future plans and possibilities diseases, nutritional disorders, maternal and perinatal The next steps in the INDEPTH WHO-SAGE collabora- conditions) and the emerging epidemics of non-commu- tion include further work on improving the existing nicable diseases in Agincourt, South Africa, provide a dataset, incorporating additional existing HDSS vari- detailed epidemiologic backdrop for analysis of variation ables and future rounds of data collection. Work will be in levels on core health domains (40). Other hypotheses undertaken to further harmonise HDSS variables, across that could be examined relate to functioning of older INDEPTH HDSS sites, for example, re-examining the adults in the context of evolving childcare contributions education data and wealth quintiles from each site. This (for example, due to AIDS mortality of household will help to improve comparability across HDSS sites and members), levels of family and household support, and countries, and with the nationally representative full associated economic activity. Health issues of mortality, SAGE studies implemented in three of the countries the compression of morbidity and social networks will (South Africa, Ghana and India). also be pursued. The ability to connect comparable data Additional HDSS variables have already been identi- on different dimensions of functioning to rich databases fied and will be added to the current summary dataset to on individual and household variables has the potential produce an enhanced dataset. Planned additions include to support important analyses for a wide range of longitudinal HDSS data such as in- and out-migration, questions concerning shifting determinants of health in births, deaths, additional respondent characteristics older adults in settings undergoing dramatic socio- (mother tongue, ethnicity, religious denomination) or demographic changes. changes in respondent and household characteristics over time (education, marital status, walls, floors, water, Archiving and sharing sanitation, fuel use for cooking, food security), and Appropriate metadata and the summary SAGE dataset relevant data about health (non-communicable disease with selected HDSS variables included will be made risk factors for example) and household composition publicly available to researchers in concert with the (members). We will also include historical HDSS data to publication of this supplement (see Supplementary files cover at least SAGE baseline years (back through 2002). under Reading Tools online). The dataset will also Three HDSS sites (Agincourt, Navrongo and Vadu) be archived in the University of Michigan’s National collected data using both the summary and full versions Archive of Computerized Data on Aging (NACDA). of the SAGE questionnaire. Examination of data from respondents who completed both the short and full Conclusion survey will be undertaken and then compared with the This collaboration provides both the practical tools and nationally representative SAGE survey in their respective infrastructure for collecting critical evidence needed by countries. These steps will allow examination of sub- researchers and policy-makers. Health, disability, living national variation in health levels, as well as variation in conditions and social support are concerns for ageing the relationships between physical and mental function- populations throughout the world. Considering the ing and other socio-demographic factors. The perfor- dearth of health and well-being data for older people in mance of the SAGE health module and vignettes among most lower- and middle-income countries (13, 41, 42), older adults in the surveillance sites can also be compared this collaboration directly addresses this data gap now to the performance in the community SAGE samples and into the future. WHO and INDEPTH will work to from these countries. It will provide opportunities to improve availability and use of reliable, valid and compare and correlate findings from African and Asian comparable health information at the country and global countries participating in SAGE with INDEPTH sites in levels, developing and improving tools and methods for the same Á as well as contrasting Á national settings. collecting this information, and providing norms, stan-

20 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 Ageing and adult health status in eight lower-income countries

dards and technical guidance for data collection, re- lance study. Lancet 2008; 372: 893Á901. Doi: 10.1016/S0140- search, analysis and synthesis of knowledge. The articles 6736(08)61399-9. 11. Lafortune G, Balestat G. Trends in severe disability among that follow in this supplement illustrate the value and elderly people: assessing the evidence in 12 OECD countries and quality of the data collected as part of this collaboration. the future implications. OECD Health Working Papers No. 26. 2007. DELSA/HEA/WD/HWP (2007)2. 12. Banks J, Marmot M, Oldfield Z, Smith JP. Disease and Acknowledgements disadvantage in the United States and in England. JAMA 2006; 295: 2037Á45. WHO Multi-Country Studies unit contributed the SAGE survey 13. National Research Council. Preparing for an aging world: the instruments, supporting materials and technical support. The Umea˚ case for cross-national research. Panel on a research agenda and Centre for Global Health Research provided technical support and new data for an aging world. Washington, DC: National advice to the INDEPTH HDSS sites and hosted an analytic and Academy Press; 2001. writing workshop in 2008. The Health and Population Division, 14. Ferreira M. Research on ageing in Africa: what do we have, not School of Public Health, University of the Witwatersrand, provides have and should we have? Generations Rev 2005; 15: 32Á5. co-leadership for this initiative and serves as a satellite secretariat for 15. United Nations Department of Economic and Social Affairs, the INDEPTH Adult Health and Ageing Working Group. Population Division. World Population Prospects: the 2008 Revision. New York: United Nations; 2009. 16. World Bank. Country classification; 2009. Available from: Conflict of interest and funding http://data.worldbank.org/about/country-classifications [cited Financial support for six HDSS sites (four African sites 10 May 2010]. plus Matlab and Vadu) was provided by the US National 17. Ebrahim S, Smeeth L. Non-communicable diseases in low and Institute on Aging through an interagency agreement middle-income countries: a priority or a distraction? Int J with the World Health Organization, and for two HDSS Epidemiol 2005; 34: 961Á6. 18. Salomon JA, Murray CJL. The epidemiologic transition revis- sites (FilaBavi and Purworejo) from the Swedish Council ited: compositional models for causes of death by age and sex. for Working Life and Social Research (FAS) through Popul Dev Rev 2002; 28: 205Á28. Umea˚ University. 19. Mathers C, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med 2006; 3: 2011Á 30, e442. Doi: 10.1371/journal.pmed.0030442. References 20. Mathers C, Fat DM, Boerma JT. The global burden of disease: 2004 update. Geneva: World Health Organization; 2008. Avail- 1. Cairncross F. Economics: age, health and wealth. Nature 2007; able from: http://www.who.int/healthinfo/global_burden_disease 448: 875Á6. /GBD_report_2004update_full.pdf [cited 10 May 2010]. 2. Lee R, Mason A. Population aging, wealth, and economic 21. Murray CJL, Lopez AD. Global mortality, disability, and the growth: demographic dividends and public policy. Background contribution of risk factors: global Burden of Disease Study. Paper for the World Economic and Social Survey. New York: Lancet 1997; 349: 1436Á42. United Nations, Department of Economic and Social Affairs; 22. Chatterji S, Kowal P, Mathers C, Naidoo N, Smith J, Suzman R. 2007. The health of aging populations in China and India. Health Aff 3. Deaton A. Policy implications of the gradient of health and 2008; 27: 1052Á63. wealth. Health Aff 2002; 21: 13Á30. 23. Ustun TB, Chatterji S, Mechbal A, Murray CJL, WHS 4. England RS. The macroeconomic impact of global aging. A new Collaborating Groups. The World Health Surveys. In: Murray era of economic frailty? Washington, DC: CSIS Press; 2002. CJL, Evans D, eds. Health systems performance assessment: 5. World Health Organization. Macroeconomics and health: debates, methods and empiricism. Geneva: World Health investing in health for economic development. Report of the Organization; 2003. Commission on Macroeconomics and Health. Geneva: World 24. King G, Murray CJL, Salomon JA, Tandon A. Enhancing the Health Organization; 2001. validity and cross-population comparability of measurement in 6. International Social Security Association. Social Security Pro- survey research. Am Political Sci Rev 2004; 98: 567Á83. grams throughout the World. Geneva: ISSA; 2009. Available 25. Murray CJL, Ozaltin E, Tandon A, Salomon JA, Sadana R, from: http://www.issa.int/aiss/Resources/ISSA-Publications/ Chatterji S. Empirical evaluation of the anchoring vignette Social-Security-Programs-Throughout-the-World6 [cited 10 approach in health surveys. In: Murray CJL, Evans D, eds. May 2010]. Health systems performance assessment: debates, methods and 7. Bloom DE, Canning D, Sevilla J. The demographic dividend. A empiricism. Geneva: World Health Organization; 2003. new perspective on the economic consequences of population 26. Salomon JA, Tandon A, Murray CJL. Comparability of self change. Santa Monica, CA: RAND; 2002. rated health: cross-sectional multi-country survey study using 8. Frenk J. The global health system: strengthening national health anchoring vignettes. BMJ 2004; 328: 258Á61. systems as the next step for global progress. PLoS Med 2010; 7: 27. Tandon A, Murray CJL, Salomon JA, King G. Statistical e1000089. Doi: 10.1371/journal.pmed.10000898. models for enhancing cross-population comparability In: Mur- 9. Coovadia H, Jewkes R, Barron P, Sanders D, McIntyre D. The health and health system of South Africa: historical roots of ray CJL, Evans D, eds. Health systems performance assessment: current public health challenges. Lancet 2009. Doi: 10.1016/ debates, methods and empiricism. Geneva: World Health S0140-6736(09)60951-X. Organization; 2003. 10. Tollman SM, Kahn K, Sartorius B, Collinson MA, Clark SJ, 28. Schmidt S, Mu¨hlan H, Power M. The EUROHIS-QOL 8-item Garenne ML. Implications of mortality transition for primary index: psychometric results of a cross-cultural field study. Eur J health care in rural South Africa: a population-based surveil- Public Health 2006; 16: 420Á8.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 21 Paul Kowal et al.

29. Bardage C, Saskia M, Pluijm F, Pedersen NL, Deeg DJH, Jylha¨ 37. Luciano JV, Ayuso-Mateos JE, Ferna´ndez A, Serrano-Blanco M, et al. Self-rated health among older adults: a cross-national A, Roca M, Haro JM. Psychometric properties of the twelve comparison. Eur J Ageing 2005; 2: 149Á58. item World Health Organization Disability Assessment Sche- 30. Frankenberg E, Jones NR. Self-rated health and mortality: does dule II (WHO-DAS II) in Spanish primary care patients with a the relationship extend to a low income setting? J Health Soc first major depressive episode. J Aff Disord 2010; 121: 52Á8. Behav 2004; 45: 441Á52. 38. O’Donovan M-A, Doyle A. Measure of Activity and Participa- 31. Salomon JA, Nordhagen S, Oza S, Murray CJL. Are Americans tion (MAP), World Health Organization’s Disability Assessment feeling less healthy? The puzzle of trends in self-rated health. Schedule (WHODAS II); 2007. Available from: http:// Am J Epidemiol 2009; 170: 343Á51. www.hrb.ie/uploads/tx_hrbpublications/mapbulletin.pdf [cited 32. Sadana R, Mathers CD, Lopez AD, Murray CJL, Iburg KM. 3 August 2010]. Comparative analysis of more than 50 household surveys of 39. World Health Organization. International classification of health status. In: Murray CJL, Salomon JA, Mathers CD, functioning, disability and health (ICF); 2001. Available from: Lopez AD, eds. Summary measures of population health: http://www.who.int/classifications/icf/en/ [cited 10 May 2010]. concepts, ethics, measurement and applications. Geneva: World 40. WHO SAGE survey programme; 2010. Available from: Health Organization; 2002. http://www.who.int/healthinfo/systems/sage [cited 10 May 2010]. 33. Wilson M, Allen DD, Li JC. Improving measurement in health 41. INDEPTH Network, iSHARE; 2010. Available from: http:// education and health behavior research using item response www.indepth-network.org [cited 10 May 2010]. modelling: introducing item response modelling. Health Educ 42. National Research Council. Aging in sub-Saharan Africa: Res 2006; 21: i4Á18. recommendations for furthering research. Panel on policy 34. Andrich D. Controversy and the Rasch model: a characteristic research and data needs to meet the challenge of aging in of incompatible paradigms? Med Care 2004; 42: 1Á16. Africa. In: Cohen B, Menken J, eds. Committee on population, 35. World Health Organization. Measuring health and disability: a division of behavioral and social sciences and education. manual for the World Health Organization Disability Assess- Washington, DC: The National Academies Press; 2006. ment Schedule (WHODAS 2.0). Geneva: World Health Orga- nization; 2009. 36. Sousa RM, Dewey ME, Acosta D, Jotheeswaran AT, Castro- *Paul Kowal Costa E, Ferri CP, et al. Measuring disability across cultures Á Multi-Country Studies Unit the psychometric properties of the WHODAS II in older people World Health Organization from seven low- and middle-income countries. The 10/66 20 Avenue Appia Dementia Research Group population-based Surrey. Int J CH-1211 Geneva, Switzerland Methods Psych Res 2010; 19: 1Á17. Email: [email protected]

22 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 æINDEPTH WHO-SAGE Supplement Assessing health and well-being among older people in rural South Africa F. Xavier Go´ mez-Olive´ 1,2*, Margaret Thorogood1,3, Benjamin D. Clark1,4, Kathleen Kahn1,2,5# and Stephen M. Tollman1,2,5#

1MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa; 2INDEPTH Network Accra, Ghana; 3Warwick Medical School, University of Warwick, Coventry, UK; 4Centre for Population Studies, London School of Hygiene and Tropical Medicine, London, UK; 5Umea˚ Centre for Global Health Research, Epidemiology and Global Health, Umea˚ University, Umea˚ , Sweden

Background: The population in developing countries is ageing, which is likely to increase the burden of non- communicable diseases and disability. Objective: To describe factors associated with self-reported health, disability and quality of life (QoL) of older people in the rural northeast of South Africa. Design: Cross-sectional survey of 6,206 individuals aged 50 and over. We used multivariate analysis to examine relationships between demographic variables and measures of self-reported health (Health Status), functional ability (WHODASi) and quality of life (WHOQoL). Results: About 4,085 of 6,206 people eligible (65.8%) completed the interview. Women (Odds Ratio (OR) 1.30, 95% CI 1.09, 1.55), older age (OR2.59, 95% CI 1.97, 3.40), lower education (OR1.62, 95% CI 1.31, 2.00), single status (OR1.18, 95% CI 1.01, 1.37) and not working at present (OR1.29, 95% CI 1.06, 1.59) were associated with a low health status. Women were also more likely to report a higher level of disability (OR1.38, 95% CI 1.14, 1.66), as were older people (OR2.92, 95% CI 2.25, 3.78), those with no education (OR1.57, 95% CI 1.26, 1.97), with single status (OR1.25, 95% CI 1.06, 1.46) and not working at present (OR1.33, 95% CI 1.06, 1.66). Older age (OR1.35, 95% CI 1.06, 1.74), no education (OR1.39, 95% CI 1.11, 1.73), single status (OR1.28, 95% CI 1.10, 1.49), a low household asset score (OR1.52, 95% CI 1.19, 1.94) and not working at present (OR1.32; 95% CI 1.07, 1.64) were all associated with lower quality of life. Conclusions: This study presents the first population-based data from South Africa on health status, functional ability and quality of life among older people. Health and social services will need to be restructured to provide effective care for older people living in rural South Africa with impaired functionality and other health problems. Keywords: adult health; ageing; self-reported health; disability; quality of life; South Africa; rural; INDEPTH WHO-SAGE

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘SAGE data’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 3 November 2009; Revised: 10 June 2010; Accepted: 8 July 2010; Published: 27 September 2010

he world’s population is ageing and projections show that this increase will continue (1, 2). The #Supplement Editor, Kathleen Kahn, Supplement Editor, Stephen M. Tollman, have not participated in the review and decision process Tpercentage of the world’s population aged 65 and for this paper. over is projected to increase steeply in coming years

Global Health Action 2010. # 2010 F. Xavier Go´ mez-Olive´ et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 23 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 F. Xavier Go´ mez-Olive´ et al.

(1Á3). The growth in the world population aged 50 and was 9.9% in 1992, 10.7% in 2000 and 11.7% in 2007 over is expected to increase from 21% in 2011 to 34% in (Fig. 1). In this area there are high labour migration rates 2050. This increase will affect not only developed of around 60% in adult males 35Á50 years old (13) and countries but also developing countries (1). In particular, high HIV-related mortality in young adults (14, 15). in developing countries demographers have predicted an Despite a falling life expectancy at birth (14), we have increase of 140% between 2006 and 2030 (4), from 35 to seen an increase in the older population. Information more than 69 million (3). The health effects of this global from annually updated health and socio-demographic demographic change are, as yet, not fully known but surveillance has shown an increase of 15% in non- estimations predict that the change in age structure in communicable diseases during the past 10 years, while coming years will bring an increase in mortality due to the number of chronic conditions overall requiring long- non-communicable diseases, changing the pattern of the term care has increased 2.6-fold (16). This may increase most common causes of death in the different regions of the existing high burden on health services depending on the world and the world as a whole (2). In 2005 it was the proportion of older people seeking health care. estimated that a total of 37 million chronic disease deaths In addition, this may increase the demand for social occurred worldwide, and more than three-quarters (77%) support for these individuals in their communities. were in people aged above 60 (5, 6). Many of these deaths Changes in the social structure and roles and respon- were preventable and a call has already been made sibilities of older people, particularly women, have for active interventions to decrease this death rate by already occurred (10). In this new reality, older women 2015 (5). For most of the developing world, and face additional responsibilities such as nursing their sick particularly for sub-Saharan Africa, this epidemic of children and taking care of their grandchildren (17). non-communicable diseases is appearing at a time when Older people have also become the main bread winners countries are also experiencing a crippling HIV epidemic. through their social pension, which is sometimes The recent availability of highly active anti-retroviral the family’s only source of income (18). In 2006, any therapy (HAART) means that, for those people with South African citizen (women 60 years or older and men access to treatment, AIDS is becoming a chronic disease 65 years or older) living in South Africa could apply for requiring long-term clinical management (7, 8). the government monthly pension (the Old Age Grant). The high HIV prevalence and recent access to This grant also depends on the person’s income, taking HAART, together with an ageing population and the into account the total amount in the family if the person emerging epidemic of non-communicable diseases, will is married (19, 20). put immense pressure on already weak health services as For all the above reasons, the health and well-being of well as on society as a whole, with important changes in older adults in rural South Africa has become a crucial household structure (9) and in the roles and responsi- issue which may impact the well-being of the entire bilities of older people (10). population. However, the impact of the changing age In South Africa, the proportion of the population aged structure and the growth in chronic disease and disability 50 and over has slightly increased from 14.8% in 2006 is poorly understood. We have therefore set out to (11) to 15% in 2009 (12) and is predicted to be 19% in address this gap. In this article, we describe the findings of a population survey of people aged 50 and over which 2030 (1). This research is based in the Agincourt sub- included information on their self-reported health, levels district of rural northeast South Africa, where the of disability and overall quality of life (QoL), which is the proportion 50 years and over in the study population first time that such findings have been reported.

Methods

Study setting The study site covers an area of 402 km2 of semi-arid scrub land. It is situated in the rural northeast of South Africa in the Bushbuckridge sub-district of Ehlanseni District, Mpumalanga Province. In the 2006 census, there was a population of 71,587 people living in 21 villages and 11,734 households. Individuals aged 50 and over constituted 12% of the population. The MRC/WITS Rural Public Health and Health Transitions Research Unit (Agincourt Unit) has been Fig. 1. Trend in proportion of population 50 years and older monitoring causes of death, births and migration in a in Agincourt sub-district, South Africa, 1992Á2007. population of around 70,000 people since 1992 (21). Each

24 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 Cross-sectional survey of older people in rural South Africa

individual and household has a unique identifying HDSS database: data routinely collected every year were number. The information is updated annually by trained extracted from the 2006 census, while Household Asset fieldworkers through a household census. Each year, Score and Employment Status data were extracted from additional modules focusing on specific research and the most recent available data (2005 and 2004, respec- policy issues (for example, food security, household assets, tively). health care utilisation, labour participation and temporary Local staff translated the questionnaires forward and migration) are included. A verbal autopsy, to determine backward into Shangaan, the local language. The final probable cause of death, is conducted on every death. version of the questionnaire included amendments fol- Although there has been substantial development in lowing a pilot conducted in several households before the the area since democratic elections in 1994, and a start of data collection. standpipe providing clean water and an electricity supply During the 4 months of field work, three stages of to households is available in all villages, the infrastructure quality control were implemented: (1) field workers cross- remains poor. There is a high unemployment rate with checked each others’ forms on a weekly basis; (2) field 36% of the total adult population unemployed and supervisors carried out daily supervision and weekly looking for work (29% of men and 46% of women Á quality control checks; and (3) two full-time workers unpublished data, 2004). As is common in rural South checked the completeness and quality of all census Africa and the region, reflecting the structure of the questionnaires including the SAGE questionnaires prior regional economy, labour migration is high, especially in to data entry. Any identified errors were referred back to men aged 35Á50 years old of whom 60% live outside the the field worker who revisited the respondent to correct study area for more than 6 months per year (13). the data. There are six clinics and one health centre within the study area; these are served by three hospitals situated 25 Variables and 45 km away (22). The public health service staff are We considered factors that could be associated with levels heavily over-committed, staff training is limited, and of QoL and disability in our population including: age, chronic disease management programmes are not yet education, marital status, household assets, nationality, fully developed. Improvement of primary health care employment status and household conditions. We calcu- services is a priority for the Province (16). lated age at interview from the recorded date of birth and reported age in four age groups: 50Á59 years, 60Á69, Sample 70Á79 and 80. Using the 2005 Agincourt census update, all 6,206 Education was categorised according to the WHO- individuals aged 50 and over and living permanently in recommended levels of education: no formal education; the study area were highlighted on the 2006 household less than six years of formal education; and six years or roster used by field workers to update census informa- more of formal education. This information was obtained tion. In this manner, field workers knew which indivi- from the census database, which is updated every 5 years duals should be invited to complete the additional using a full questionnaire on education status (last questionnaire described in the next section. If an updated in 2006). individual was not available for interview at the first Since many unions are traditional rather than civic and visit, the field worker made up to two further visits to polygamy is practised by some people, we categorised attempt to complete the interview. Before the 2006 census marital status into two groups: (1) currently married or update, a similar but more extensive questionnaire was living as married; and (2) single, including anyone with- conducted in a sample of 575 individuals 50 years old or out a current partner (i.e. those who had never married or more. Those individuals were excluded from this study. were separated, divorced or widowed). To evaluate the potential role of socio-economic status Data collection in our analyses, we used a household asset score. This Field workers employed in the annual census update were score was developed using principal component factor trained to administer the questionnaire. We used a analysis and 34 variables derived from the 2005 census questionnaire adapted from the World Health Organiza- questionnaire Á including information collected about the tion (WHO) Study on Global AGEing and Adult Health type and size of dwelling, access to water and electricity, (23) (the SAGE study). It included questions on self- appliances and livestock owned and transport available. reported health, functionality (mobility, self-care, pain During and following the civil war in Mozambique, the and discomfort, cognition, interpersonal activities, sleep/ Agincourt area received many refugees; hence we re- energy, affect, vision and general health conditions) and corded a variable ‘nationality of origin’ (South African/ well-being, as well as the eight questions which form the Mozambican). The Mozambican group are separately WHO Quality of Life (WHOQoL) measure. Additional identified in the census data and it has been previously demographic data were extracted from the Agincourt observed that this group differs from the host South

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 25 F. Xavier Go´ mez-Olive´ et al.

Table 1. Domains and scales

Health status WHODASi WHOQoL

Domains Mobility Interpersonal activities Enough energy for daily life Self-care Difficulties in daily living: Enough money to meet needs Pain and discomfort “ Standing Satisfaction with: Cognition “ Walking “ Your health Interpersonal activities “ Household duties “ Yourself Sleep/energy “ Learning “ Ability to perform daily activities Affect “ Concentrating “ Personal relationships Vision “ Self-care “ Condition of your living place Rate your overall quality of life Scale 0 (poor health) to 100 (good health) 0 (low ability) to 100 (high ability) 0 (low quality of life) to 100 (high quality of life)

African population in measures such as education, household assets and child mortality (24). Many Mozam- Table 2. Background characteristics by response for 6,206 bicans have now taken South African nationality which adults 50 years and older living permanently in the Agin- allows them to work legally and receive state pensions. court sub-district, 2006 Employment status (currently working or not) is based on Agincourt 2004 census data, when it was most recently p-Value for collected. The majority of those not working were not difference looking for work, but had retired in the sense they had Non- respondents concluded their working career. Respondents respondents vs. non- Variables (N4,085) (N2,121) respondents In order to examine whether health and well-being were affected by the age structure of the household, we Sex (%) created a dichotomous variable for those living in Men 1,012 (24.8) 926 (43.7) B0.001 households with younger members and those living in Women 3,073 (75.2) 1,195 (56.3) Mean age (SD) 66.6 (10.6) 64.8 (11.3) B0.001 households with no one under the age of 50, using data from the 2006 census. Age group (years) 50Á59 1,297 (31.7) 923 (43.5) B0.001 60Á69 1,221 (29.9) 546 (25.7) 70Á79 1,077 (26.4) 413 (19.5) Health Status, Disability and Quality of Life 80 490 (12.0) 238 (11.2) (QoL) scores Education level (%) These three measures progress from what may be seen as No formal education 2,601 (65.8) 1,038 (67.5) B0.001 a more basic health status assessment (Health Status) Less than or equal 757 (19.2) 218 (14.1) through to more complex functioning of the person to 6 years (WHODAS) and then the person’s satisfaction with their More than 6 years 594 (15.0) 292 (18.9) life (WHOQoL). WHODAS is a scale designed to Marital status (%) measure disability (with a high score indicating a severe Single 2,223 (54.4) 1,125 (53.0) 0.302 Current partnership 1,862 (45.6) 996 (47.0) lack of physical functioning). Thus, for consistency between the scores used in this study, an inverted score Household asset score (%) First quintile 629 (15.9) 313 (18.5) 0.125 designated WHODASi has been used, with the conse- Second quintile 753 (18.9) 312 (18.5) quence that all three scores are based on a 0Á100 scale, Third quintile 766 (19.3) 330 (19.5) and in all cases a high score indicates a good outcome. Fourth quintile 841 (21.2) 329 (19.5) Table 1 shows the domains used to calculate the variables Fifth quintile 978 (24.6) 405 (24.0) and their scales. Mean number of 7.0 (4.1) 7.4 (4.6) 0.002 household Health Status is a composite score which includes members (SD) functionality and QoL domains. Health Status generally Household members 32.1 (25.9) 28.9 (25.9) B0.001 refers to physical and occupational functions, psycholo- aged 50 years and gical states, social interaction and somatic sensations (25). over (SD) This general health score was derived using item response Nationality of origin theory (IRT) parameter estimates in Winsteps, a Rasch South African 2,972 (72.8) 1,399 (66.0) B0.001 measurement software package (http://www.winsteps. Mozambican 1,111 (27.2) 720 (34.0) com). IRT uses Maximum Likelihood Estimation, which Occupational status in 2004 combines the pattern of responses as well as the char- Working 503 (14.6) 481 (28.8) B0.001 Not working 2,930 (85.3) 1,189 (71.2) acteristics of each specific item for the multiple health

26 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 Cross-sectional survey of older people in rural South Africa

Table 3a. Demographic variables by sex [n, (%)] for 4,085 adults aged 50 and over in Agincourt sub-district, 2006

p-Value for difference between Variable Males Females Total male and female

Sex (%) 1,012 (24.8) 3,073 (75.2) 4085 (100) pB0.001 Mean Age in years (95% CI) 67.8 (67.1, 68.5) 66.1 (65.7, 66.4)

Age group (years) 50Á59 275 (27.2) 1,022 (33.3) 1,297 (31.7) df3 60Á69 321 (31.7) 900 (29.3) 1,221 (29.9) p0.001 70Á79 269 (26.6) 808 (26.3) 1,077 (26.4) 80 147 (14.5) 343 (11.2) 490 (12.0)

Partnership status In a partnership 771 (76.2) 1,091 (35.5) 1,862 (45.6) df1 Currently single 241 (23.8) 1,982 (64.5) 2,223 (54.4) pB0.001

Education level No education 549 (54.2) 2,052 (66.8) 2,601 (63.7) df3 Less than 6 years 214 (21.1) 543 (17.1) 757 (18.5) pB0.001 Six years or more 209 (20.6) 385 (12.5) 594 (14.5) Missing data 40 (4.0) 93 (3.0) 133 (3.3)

Household asset score (quintiles) First (lowest) 159 (15.7) 470 (15.3) 629 (15.4) df5 Second 167 (16.5) 586 (19.1) 753 (18.4) p0.016 Third 171 (16.9) 595 (19.4) 766 (18.7) Fourth 212 (20.9) 629 (20.5) 841 (20.6) Fifth (highest) 279 (27.6) 699 (22.7) 978 (23.9) Missing data 24 (2.4) 94 (3.1) 118 (2.9)

Household with and without people aged less than 50 years With under 50 853 (84.3) 2841 (92.5) 3694 (90.4) df1 Without under 50 159 (15.7) 232 (7.5) 391 (9.6) pB0.001

Nationality of origin South African 767 (75.9) 2,205 (71.8) 2,972 (72.8) df1 Mozambican 244 (24.1) 867 (28.2) 1,111 (27.2) p0.011

Occupational status in 2004 Working 169 (19.7) 334 (13.0) 503 (14.7) df1 Not working 690 (80.3) 2,240 (87.0) 2,930 (85.4) pB0.001 questions (each with multiple response categories) to response options. Measurement of functionality was produce the final health score. The health score is then calculated by asking participants about difficulty experi- transformed to a scale of 0Á100. IRT models the relation- enced performing certain activities during the past 30 ship between a person’s reported Health Status and their days, and transformed into the WHODASi score for probability of responding to each question in a multi-item functional ability as described above. scale. A key feature of IRT modelling is that item QoL was measured using the Word Health Organisa- parameter estimates should be invariant to group mem- tion Quality of Life (WHOQoL) scale. WHO defines QoL bership (i.e. each item functions similarly across groups of as ‘the individual’s perception of their position in life in people from different cultures) (26). the context of the culture and value systems in which they To measure disability levels we used the WHODAS II live and in relation to their goals, expectations, standards (World Health Organization Disability Assessment Sche- and concerns’ (27, 28). QoL domains include questions on dule II) scale that assesses day-to-day functioning in six self-rated general health and questions on satisfaction. activity domains. There are 10 questions with multiple The WHOQoL score is presented on a scale of 8Á40

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 27 F. Xavier Go´ mez-Olive´ et al.

Table 3b. Demographic variables by age group for 4,085 adults aged 50 and over in Agincourt sub-district, 2006

Age groups 50Á59, N (%) 60Á69, N (%) 70Á79, N (%) 80, N (%) Total N (%) p-Value

Sample distribution 1,297 (31.8) 1,221 (29.9) 1,077 (26.4) 490 (12) 4,085 (100) Mean (95% CI) 54.5 (54.4Á54.7) 64.8 (64.6Á64.9) 74.5 (74.3Á74.7) 84.9 (84.6Á85.3)

Sex Male 275 (21.2) 321 (26.3) 269 (25.0) 147 (30.0) 1,012 (24.8) df3 Female 1,022 (78.8) 900 (73.7) 808 (75.0) 343 (70.0) 3,073 (75.2) p0.001

Marital status In a partnership 732 (56.4) 615 (50.4) 374 (34.7) 141 (28.8) 1,862 (45.6) df3 Currently single 565 (43.6) 606 (49.6) 703 (65.3) 349 (71.2) 2,223 (54.4) pB0.001

Education level No formal education 630 (48.6) 736 (60.3) 844 (78.4) 391 (79.8) 2,601 (63.7) df9 Primary or less than six years 304 (23.4) 253 (20.7) 144 (13.4) 56 (11.4) 757 (18.5) pB0.001 Six years or more 316 (24.4) 193 (15.8) 61 (5.7) 24 (4.9) 594 (14.5) Missing 47 (3.6) 39 (3.2) 28 (2.6) 19 (3.9) 133 (3.3)

Socio-economic quintiles First (lowest) 198 (15.3) 153 (12.5) 186 (17.3) 92 (18.8) 629 (15.4) df15 Second 233 (18.0) 198 (16.2) 220 (20.4) 102 (20.8) 753 (18.4) p B0.001 Third 238 (18.4) 246 (20.2) 199 (18.5) 83 (16.9) 766 (18.8) Fourth 258 (19.9) 258 (21.1) 231 (21.5) 94 (19.2) 841 (20.6) Fifth (highest) 337 (26.0) 326 (26.7) 217 (20.2) 98 (20.0) 978 (23.9) Missing 33 (2.5) 40 (3.3) 24 (2.2) 21 (4.3) 118 (2.9)

Adult in the household Youth plus older 1,206 (93.0) 1,123 (92.0) 964 (89.5) 401 (81.8) 3,694 (90.4) df3 Only older 91 (7.0) 98 (8.0) 113 (10.5) 89 (18.2) 391 (9.6) pB0.001

Nationality South African 957 (73.8) 919 (75.3) 740 (68.7) 356 (72.7) 2,972 (72.8) df3 Mozambican 339 (26.2) 301 (24.7) 337 (31.3) 134 (27.4) 1,111 (27.2) p0.003

Occupational status Working 284 (26.4) 160 (15.3) 44 (4.9) 15 (3.6) 503 (14.7) df3 Not working 791 (73.6) 883 (84.7) 859 (95.1) 397 (96.4) 2,930 (85.4) pB0.001

(where 8 is the best QoL) and transformed to a model sequentially and then discarded if the effect was not 0Á100 scale corresponding to the other scores. significant at the level of p0.1.

Data entry and analysis We entered data using CSPro 3.1 data entry programme Ethical clearance (http://www.census.gov/ipc/www/cspro/index.html) which Ethical clearance for the MRC/WITS Rural Public includes validation checks, and data was then extracted to Health and Health Transitions Research Unit Á Health Stata 10.1 (College Station, TX, USA) for analysis. and Socio-Demographic Surveillance System (Agincourt) Logistic regression was performed to assess the relation Á census and modules has been granted by the Committee between potentially associated factors and confounders, for Research on Human Subjects (Medical) of the and the three outcomes, i.e. health score, functionality University of the Witwatersrand, Johannesburg, South (WHODASi) and quality of life (WHOQoL). We first Africa (Ref No. M960720). Ethical clearance for the carried out a univariate analysis with each of the census Agincourt-INDEPTH Study on Global Ageing and variables and then constructed a multivariate model based Adult Health was given by the Committee for Research on the results of the univariate analyses (Tables 5, 7 and 9). on Human Subjects (Medical) of the University of the Variables which were significantly related to the outcome Witwatersrand, Johannesburg, South Africa (Ref No. measures in a univariate analysis were introduced into the R14/49).

28 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 Cross-sectional survey of older people in rural South Africa

Table 4. Range of Health Status (quintiles) by demographic variables [n, (%)] for 4,085 adults aged 50 and over in Agincourt sub-district, 2006

Health status quintile

Variable 1 (poorest) 2 3 4 5 (best) p-Value

Sex Male 160 (15.8) 170 (16.8) 175 (17.3) 215 (21.2) 292 (28.8) df4 Female 641 (20.9) 597 (19.4) 562 (18.3) 639 (20.8) 634 (20.6) pB0.001

Age group (years) 50Á59 170 (13.1) 240 (18.5) 220 (17) 315 (24.3) 352 (27.1) df12 60Á69 183 (15) 209 (17.1) 239 (19.6) 283 (23.2) 307 (25.1) pB0.001 70Á79 270 (25.1) 207 (19.2) 202 (18.8) 193 (17.9) 205 (19) 80 and over 178 (36.3) 111 (22.7) 76 (15.5) 63 (12.9) 62 (12.7)

Partnership In a partnership 277 (14.9) 341 (18.3) 328 (17.6) 411 (22.1) 505 (27.1) df4 Currently single 524 (23.6) 426 (19.2) 409 (18.4) 443 (19.9) 421 (18.9) pB0.001

Education level No education 590 (22.7) 500 (19.2) 475 (18.3) 510 (19.6) 526 (20.2) df8 Less than 6 years 120 (15.9) 140 (18.5) 147 (19.4) 166 (21.9) 184 (24.3) pB0.001 Six years or more 65 (10.9) 97 (16.3) 96 (16.2) 159 (26.8) 177 (29.8)

Household asset score (quintiles) First (lowest) 126 (20.0) 120 (19.1) 111 (17.7) 131 (20.8) 141 (22.4) df16 Second 159 (21.1) 148 (19.7) 138 (18.3) 155 (20.6) 153 (20.3) p0.321 Third 145 (18.9) 135 (17.6) 147 (19.2) 163 (21.3) 176 (23.0) Fourth 164 (19.5) 177 (21.1) 152 (18.1) 163 (19.4) 185 (22.0) Fifth (highest) 179 (18.3) 165 (16.9) 160 (16.4) 219 (22.4) 255 (26.1)

Household with and without people aged less than 50 With under 50 696 (18.8) 702 (19) 671 (18.2) 787 (21.3) 838 (22.7) df4 Without under 50 105 (26.9) 65 (16.6) 66 (16.9) 67 (17.1) 88 (22.5) p0.003

Nationality of origin South African 623 (21.0) 558 (18.8) 506 (17.0) 619 (20.8) 666 (22.4) df4 Mozambican 178 (16.0) 209 (18.8) 229 (20.6) 235 (21.1) 260 (23.4) p0.003

Occupational status in 2004 Working 59 (11.7) 74 (14.7) 93 (18.5) 119 (23.7) 158 (31.4) df4 Not working 612 (20.9) 569 (19.4) 518 (17.7) 612 (20.9) 619 (21.1) pB0.001

Results marital status or socio-economic status, but respondents From the 6,206 people aged 50 years and over selected included significantly more South Africans than Mozam- from the 2005 census, 4,085 (65.8%) responded to a bicans and proportionally more people who were cur- questionnaire. Of those that did not complete a ques- rently not working (85.3% vs. 71.2%; pB0.001). tionnaire, 1,616 (26.0%) were absent at the time of the About 85% of respondents were ‘currently not work- interview, 218 (3.5%) had died, 47 (0.75%) declined to ing’, but the majority of these were not formally take part and 240 (3.9%) were unable to answer the ‘unemployed’ (i.e. actively searching for work but not questions (mainly due to different health conditions). finding it). The 5.7% of people who were formally A comparison of respondents and non-respondents unemployed included 15% of those aged 50Á59 and (Table 2) shows that non-respondents were significantly 4.3% of those aged 60Á69 (data not shown). younger (mean age 64.8 vs. 66.6, pB0.001), included a Among the respondents, there were significant differen- higher proportion of men (43.7% vs. 24.8%, pB0.001) ces between men and women in all the variables (Table 3a). and were better educated. There were no differences in Only a quarter of the respondents were men (24.8%), and

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 29 F. Xavier Go´ mez-Olive´ et al.

Table 5. Factors associated with poor Health Statusa score with age (21.2% in 50Á59 age group vs. 30% in the 80 for 4,085 adults aged 50 and over in Agincourt sub-district, age group; p0.001); the younger age group was better 2006 educated (24.4% in the 50Á59 age group vs. 4.9% in 80 have 6 years or more of formal education; pB0.001); the Univariate model Multivariate model two younger age groups have higher socio-economic Variables OR (95% CI) OR (95% CI) status (26.0 and 26.7% in the younger groups vs. 20.2 and 20.0% in the older age groups; pB0.001). Sex Table 4 shows the range of Health Status responses by Male 1 1 each of the demographic variables, while Table 5 shows Female 1.42 (1.23, 1.64) 1.30 (1.09, 1.55) the results of univariate and multivariate logistic regres- Age group (years) sion analysis examining the odds of reporting a Health 50Á59 1 1 Status in one of the bottom two quintiles. Household 60Á69 1.13 (0.97, 1.32) 1.05 (0.88, 1.26) asset score, household age structure and nationality of 70Á79 1.81 (1.53, 2.13) 1.46 (1.19, 1.78) origin did not show a significant association in univariate 80 3.09 (2.45, 3.89) 2.59 (1.97, 3.40) analysis. In the final multivariate model, women had a 30% higher risk than men (odds ratio (OR)1.30, 95% Education level confidence interval (CI) 1.09, 1.55) of reporting a low No formal education 1.97 (1.64, 2.35) 1.62 (1.31, 2.00) Health Status. Older age (OR2.59, 95% CI 1.97, 3.40), Less than 6 years 1.51 (1.22, 1.88) 1.42 (1.12, 1.79) lower education level (OR1.62, 95% CI 1.31, 2.00), Six years or more 1 1 single marital status (OR1.18, 95% CI 1.01, 1.37) and Marital status not working at present (OR1.29, 95% CI 1.06, 1.59) Single 1.52 (1.34, 1.72) 1.18 (1.01, 1.37) were also all related to a poorer Health Status. People of In current partnership 1 1 Mozambican origin were 24% less likely to report a Health Status in the bottom two quintiles (OR0.76, Household with and without people aged less than 50 95% CI 0.64, 0.91). With under 50 1 Not included in the The quintiles for self-reported ability (WHODASi final model score) are shown in Table 6, while Table 7 shows the Without under 50 1.19 (0.97, 1.48) results of univariate and multivariate logistic regression Household asset score analysis examining the odds of reporting a WHODASi First quintile (lowest) 1.23 (1.01, 1.51) Not included in the score in one of the bottom two quintiles (poorer self- final model reported functioning). In multivariate analysis, women Second quintile 1.36 (1.12, 1.65) were more likely to be in the bottom two quintiles of self- Third quintile 1.18 (0.98, 1.43) reported functioning (OR1.38, 95% CI 1.14, 1.66), as Fourth quintile 1.33 (1.11, 1.60) were older people (OR2.92, 95% CI 2.25, 3.78), those Fifth quintile (highest) 1 with less education (OR1.57, 95% CI 1.26,1.97), those not in a current partnership (OR1.25, 95% CI 1.06, Nationality of origin 1.46) and those who were not working (OR1.33, 95% South African 1 1 CI 1.06, 1.66). Mozambican 0.95 (0.82, 1.09) 0.76 (0.64, 0.91) Although women were significantly more likely than Occupational status in 2004 men to be in the lowest two quintiles of self-reported QoL Working 1 1 Á WHOQoL (Table 8), this effect disappeared after Not working 1.69 (1.40, 2.05) 1.29 (1.06, 1.59) adjusting for other variables, as did the effect of house- hold age structure and nationality of origin (Table 9). In aIRT (Item Response Theory) used when measuring health status. the final multivariate model, older age (OR1.35, 95% The Health Status scale was divided in quintiles. The best Health CI 1.06, 1.74), lack of education (OR1.39, 95% CI Status was defined as those in the two highest quintiles, while the 1.11, 1.73), not being in a current partnership (OR1.28, worst Health Status was defined as those in the three lower 95% CI 1.10, 1.49), having a low household asset score quintiles. (OR1.52, 95% CI 1.19, 1.94) and not working at present (OR1.32; 95% CI 1.07, 1.64) were all asso- ciated with a higher odds of being in one of the lower two the men were older (67.8 years vs. 66.1 years; pB0.001), quintiles for WHOQoL (Table 9). more likely to be in a current partnership (76.2% vs. 35.5%; pB0.001) and more likely to be in paid employment. Discussion Demographic variables presented by age group In this study we describe the well-being and functionality (Table 3b) show that the proportion of males increased of the population aged 50 and over in the Agincourt

30 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 Cross-sectional survey of older people in rural South Africa

Table 6. WHODASia by demographic variables [n, (%)] for 4,085 adults aged 50 and over in Agincourt sub-district, 2006

WHODASi quintile

Variable 1 (high ability) 2 3 4 5 (low ability) p-Value

Sex Male 328 (32.4) 184 (18.2) 165 (16.3) 160 (15.8) 175 (17.3) df4 Female 701 (22.8) 542 (17.6) 526 (17.1) 642 (20.9) 662 (21.5) pB0.001

Age group (years) 50Á59 398 (30.7) 264 (20.4) 220 (17) 256 (19.7) 159 (12.3) df12 60Á69 364 (29.8) 238 (19.5) 210 (17.2) 217 (17.8) 192 (15.7) pB0.001 70Á79 198 (18.4) 177 (16.4) 188 (17.5) 233 (21.6) 281 (26.1) 80 and over 69 (14.1) 47 (9.6) 73 (14.9) 96 (19.6) 205 (41.8) Partnership In a partnership 545 (29.3) 369 (19.8) 323 (17.4) 343 (18.4) 282 (15.2) df4 Currently single 484 (21.8) 357 (16.1) 368 (16.6) 459 (20.7) 555 (25.0) pB0.001

Education level No education 583 (22.4) 419 (16.1) 443 (17) 539 (20.7) 617 (23.7) df8 Less than 6 years 214 (28.3) 149 (19.7) 127 (16.8) 147 (19.4) 120 (15.9) pB0.001 Six years or more 206 (34.7) 130 (21.9) 99 (16.7) 89 (15) 70 (11.8)

Household asset score (quintiles) First (lowest) 168 (26.7) 98 (15.6) 105 (16.7) 127 (20.2) 131 (20.8) df16 Second 181 (24) 139 (18.5) 129 (17.1) 153 (20.3) 151 (20.1) p0.218 Third 184 (24) 157 (20.5) 123 (16.1) 136 (17.8) 166 (21.7) Fourth 191 (22.7) 148 (17.6) 148 (17.6) 176 (20.9) 178 (21.2) Fifth (highest) 281 (28.7) 166 (17) 170 (17.4) 179 (18.3) 182 (18.6)

Household with and without people aged less than 50 With under 50 940 (25.5) 662 (17.9) 631 (17.1) 720 (19.5) 741 (20.1) df4 Without under 50 89 (22.8) 64 (16.4) 60 (15.4) 82 (21) 96 (24.6) p0.199

Nationality of origin South African 719 (24.2) 535 (18) 522 (17.6) 560 (18.8) 636 (21.4) df4 Mozambican 309 (27.8) 191 (17.2) 169 (15.2) 241 (21.7) 201 (18.1) p0.005

Occupational status in 2004 Working 179 (35.6) 98 (19.5) 85 (16.9) 81 (16.1) 60 (11.9) df4 Not working 686 (23.4) 523 (17.9) 502 (17.1) 574 (19.6) 645 (22.0) pB0.001 aWHODASi: Using the World Health Organization Disability Assessment Schedule II (WHODAS II) the variable scale was inverted and divided into quintiles.

Health and Socio-demographic Surveillance Site by Status. As with the Health Status, women were more measuring three main variables (scores) that flow from likely to report poorer functionality (WHODASi) than a more basic health status assessment (Health Status) men. Age significantly affected functionality only from through to more complex functioning of the person 70 years of age. People aged 80 and over had a threefold (WHODASi) and then to the person’s satisfaction with increase in risk of reporting poorer functionality. Pro- their life (WHOQoL). gressively lower levels of education related to a gradual Women were 30% more likely than men to report a increase in functional problems. Being single or ‘not poor state of health (low Health Status). Other factors working at present’ were also associated with worse associated with a worse Health Status were aged above functionality. There was no gender difference in QoL. 70 years, lower levels of formal education, being single However, our analysis showed the following factors and currently not working. On the other hand, being of related to lower QoL: older age group, no formal Mozambican origin is related to a better-reported Health education, being single and currently not working.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 31 F. Xavier Go´ mez-Olive´ et al.

Table 7. Factors associated with poor self-reported functio- their own health and therefore report health problems in a ning (WHODASia) for 4,085 adults aged 50 and over in higher proportion than men. At present, the data are not Agincourt sub-district, 2006 available to explore this issue further. The oldest age group (people aged 70 and over) Univariate model Multivariate model reported worst QoL and functioning. However, the age Variables OR (95% CI) OR (95% CI) group 60Á69 years presented no significant difference in Sex Health Status and functioning measures compared with Male 1 1 the 50Á59 year age group. Moreover, they reported a Female 1.49 (1.28, 1.73) 1.38 (1.14, 1.66) significantly better QoL than the younger 50Á59 age Age group (years) group. This may be related to the fact that women who 50Á59 1 1 retire at 60 and men at 65 are still in good health. In 60Á69 1.07 (0.90, 1.27) 1.00 (0.83, 1.21) addition, they receive old-age grants (pensions) which 70Á79 1.94 (1.64, 2.29) 1.62 (1.32, 1.99) 80 3.38 (2.73, 4.20) 2.92 (2.25, 3.78) allows them a better life with higher food security and, importantly, with greater capacity to help children in Education level No formal education 2.19 (1.80, 2.67) 1.57 (1.26, 1.97) their households who then enjoy higher food security and Less than 6 years 1.49 (1.18, 1.88) 1.33 (1.03, 1.72) better schooling (29). At older ages (70 and over), Health Six years or more 1 1 Status and functioning had deteriorated and they re- Marital status ported worse levels of both variables despite still receiving Single 1.66 (1.46, 1.88) 1.25 (1.06, 1.46) pension grant. In current partnership 1 1 The household asset score was created as a proxy for HH with and without people aged less than 50 household socio-economic status. The asset data used in With under 50 1 Not included in this study were collected in 2005, a year earlier than the the final model study was conducted. Our data did not show any relation Without under 50 1.28 (1.03, 1.57) between this score and either the Health Status or the Household asset score (quintiles) WHODASi. However, the household asset score is First quintile (lowest) 1.24 (1.03, 1.50) Not included in the final model significantly related to the WHOQoL that measures Second quintile 1.11 (0.91, 1.35) satisfaction with one’s life. This could mean that people’s Third quintile 1.16 (0.95, 1.41) socio-economic status has no relation to being physically Fourth quintile 1.19 (0.97, 1.46) and socially functional, but impacts on how satisfied Fifth quintile (highest) 1 people are with their life and expectations (30). Nationality of origin Unemployment among Agincourt’s adult population South African 1 Not included in (including both permanent and temporary residents) is the final model Mozambican 0.98 (0.85, 1.13) 36%, representing 29% of men and 46% of women Occupational status in 2004 (Collinson, personal communication). In our study Working 1 1 sample, 85% of all respondents were ‘not currently Not working 1.83 (1.48, 2.25) 1.33 (1.06, 1.66) working’, but only 5.7% were formally unemployed. There is a significant relationship between currently not aWHODASi: Using the World Health Organization Disability working and Health Status, WHODASi and WHOQoL Assessment Schedule II (WHODAS II) the variable scale was even after controlling for age group. inverted and divided into quintiles. ORs reflect odds for those in Other work in the Agincourt study site has shown the two lowest quintiles of functionality. residents of Mozambican origin to be a vulnerable sub- group (24, 31). We thus expected Mozambican nation- ality to have a significant relationship with low Health Finally there was a gradient in the expected direction in Status, low WHODASi and low WHOQoL. However, no the relationship between lower QoL and lower socio- relationship with WHOQoL and WHODASi was found, economic status measured by household asset score. and being Mozambican was associated with less like- Our data show that women report significantly poorer lihood of reporting a lower Health Status, meaning that functionality for both Health Status and WHODASi, the those of Mozambican origin reported feeling in better two measures that include variables of functionality, health than their South African counterparts. This may although they do not report a lower QoL. There are be related to a healthy immigrant selectivity that may several possible explanations for this. Women may objec- decrease over coming years (32). tively have poorer functionality but do not regard this as a The Agincourt HDSS includes individuals living problem, or women may be more active in the home than permanently in the area and those that spend more their retired partners and therefore more aware of a than 6 months per year outside the study area but remain change in functionality, or women may be more aware of linked to their rural households. Some permanent

32 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 Cross-sectional survey of older people in rural South Africa

Table 8. WHOQoLa by demographic variables [n (%)] for 4,085 adults aged 50 and over in Agincourt sub-district, 2006

WHOQoL quintile

Variable 1 (high) 2 3 4 5 (low) p-Value

Sex Male 244 (24.2) 217 (21.5) 168 (16.6) 171 (16.9) 210 (20.8) df4 Female 566 (18.4) 623 (20.3) 608 (19.8) 678 (22.1) 596 (19.4) pB0.001

Age group (years) 50Á59 269 (20.8) 274 (21.1) 246 (19.0) 261 (20.1) 246 (19.0) df12 60Á69 279 (22.9) 281 (23.0) 238 (19.5) 257 (21.0) 165 (13.5) pB0.001 70Á79 185 (17.2) 214 (19.9) 209 (19.4) 225 (20.9) 242 (22.5) 80 and over 77 (15.7) 71 (14.5) 83 (16.9) 106 (21.6) 153 (31.2)

Partnership In a partnership 432 (23.2) 394 (21.2) 371 (19.9) 371 (19.9) 292 (15.7) df4 Single 378 (17.0) 446 (20.1) 405 (18.2) 478 (21.5) 514 (23.1) pB0.001

Education level No education 454 (17.5) 508 (19.5) 513 (19.7) 565 (21.7) 558 (21.5) df8 Less than 6 years 169 (22.3) 163 (21.5) 131 (17.3) 164 (21.7) 129 (17.0) pB0.001 Six years or more 157 (26.4) 151 (25.4) 102 (17.2) 91 (15.3) 93 (15.7)

Household asset score (quintiles) First (lowest) 94 (14.9) 128 (20.4) 117 (18.6) 135 (21.5) 155 (24.6) df16 Second 119 (15.8) 158 (20.1) 144 (19.1) 168 (22.3) 164 (21.8) pB0.001 Third 162 (21.1) 155 (20.2) 141 (18.4) 177 (23.1) 131 (17.1) Fourth 157 (18.7) 183 (21.8) 157 (18.7) 174 (20.7) 169 (20.1) Fifth (highest) 269 (27.6) 200 (20.5) 187 (19.1) 165 (16.9) 155 (15.9)

Household with and without people aged less than 50 With under 50 735 (19.9) 772 (20.9) 708 (19.2) 768 (20.8) 710 (19.2) df4 Without under 50 78 (20.0) 68 (17.4) 68 (17.4) 81 (20.7) 96 (24.6) p0.099

Nationality of origin South African 624 (21) 617 (20.8) 559 (18.8) 587 (19.7) 585 (19.7) df4 Mozambican 189 (17.0) 223 (20.1) 215 (19.4) 262 (23.6) 221 (19.9) p0.014

Occupational status in 2004 Working 136 (27.0) 114 (22.7) 95 (18.9) 86 (17.1) 72 (14.3) df4 Not working 568 (19.4) 603 (20.6) 566 (19.3) 614 (21.0) 579 (19.8) pB0.001 aWHOQoL: The World Health Organization Quality of Life score was calculated and then divided into quintiles.

residents work in the surrounding area making it difficult (around 3% of the total population per year) are not to find them at home. In this study, 76% of non- followed up and so it is not possible to measure their respondents were not found at home for interview impact on the health status and functionality of the despite three visits to the household. Men participate remaining population. Thus, the study may have under- in the labour force more than women, and the non- estimated the reported health of the population given respondents represented nearly 50% of all men and 30% that the results show the health status of those that live of all women expected to participate in the study. Table 2 most of the year in the study area. shows that non-respondents included twice the propor- This study presents the first population-based data tion of workers compared to respondents. Moreover, from South Africa on Health Status, functionality and 69% of workers among the non-respondent group were WHOQoL. Other studies have focused on specific aged between 50 and 59 years (data not shown). Those diseases (33, 34) or on defining the best domains with who out-migrate permanently from the study area which to evaluate QoL and Health Status (30).

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 33 F. Xavier Go´ mez-Olive´ et al.

Table 9. Factors associated with poor quality of life with significantly impaired functionality and other health (WHOQoL)a for 4,085 adults aged 50 and over in problems. We plan to continue to monitor the health and Agincourt sub-district, 2006 well-being of older people. This will provide information on how societal changes are affecting their health and well- Univariate model OR Multivariate model OR being, assist policy makers to predict demand for health Variables (95% CI) (95% CI) services, and inform the development of appropriate and Sex cost-effective health and social services. Male 1 Not included in the final model Female 1.17 (1.01, 1.35) Acknowledgements

Age group (years) We thank the study participants, field team and local authorities. 50Á59 1 1 Special thanks to Dr. Oscar Franco (Warwick University, UK) and 60Á69 0.82 (0.70, 0.97) 0.73 (0.61, 0.89) to Ms. Marguerite Schneider (Human Sciences Research Council, 70Á79 1.19 (1.01, 1.37) 0.91 (0.75, 1.11) RSA) for providing useful comments for the improvement of the 80 1.74 (1.41, 2.15) 1.35 (1.06, 1.74) manuscript. This study was funded by the National Institute on Education level Aging of the National Institutes of Health, USA and by the No formal 1.69 (1.40, 2.05) 1.39 (1.11, 1.73) Wellcome Trust, UK (Grant Nos. 058893/Z/99/A and 069683/Z/02/ education Z). It was carried out in collaboration with the World Health Less than 6 1.41 (1.12, 1.77) 1.33 (1.04, 1.70) Organization. years Six years or 11Conflict of interest and funding more The authors have not received any funding or benefits Marital status from industry to conduct this study. Now single 1.46 (1.28, 1.65) 1.28 (1.10, 1.49) In current 11 partnership References Household with and without people aged less than 50 With under 50 1 Not included in the 1. United Nations Department of Economics and Social Affairs. final model Population division. World population prospects: the 2008 Without under 1.24 (1.01, 1.53) revision. New York: United Nations; 2009. 50 2. Murray CJ, Lopez AD. Alternative projections of mortality and Household asset score (quintiles) disability by cause 1990Á2020: global burden of disease study. First (lowest) 1.76 (1.43, 2.16) 1.52 (1.19, 1.94) Lancet 1997; 349: 1498Á504. Second 1.62 (1.33, 1.97) 1.49 (1.19, 1.86) 3. Velkoff VA, Kowal PR. US Census Bureau, current population Third 1.38 (1.13, 1.68) 1.29 (1.04, 1.61) reports, P95/07-1. Population aging in Sub-Saharan Africa: Fourth 1.41 (1.17, 1.71) 1.34 (1.09, 1.66) demographic dimensions 2006. Washington, DC: U.S. Govern- Fifth (highest) 1 1 ment Printing Office; 2007. 4. Min SK, Kim KI, Lee CI, Jung YC, Suh SY, Kim DK. Nationality of origin Development of the Korean versions of WHO Quality of Life South African 1 Not included in the scale and WHOQoL-BREF. Qual Life Res 2002; 11: 593Á600. final model 5. Strong K, Mathers C, Leeder S, Beaglehole R. Preventing Mozambican 1.18 (1.03, 1.36) chronic diseases: how many lives can we save? Lancet 2005; 366: Occupational status in 2004 1578Á82. Working 1 1 6. World Health Organization. Preventing chronic diseases Á a vital Not working 1.50 (1.22, 1.83) 1.32 (1.07, 1.64) investment. Geneva: WHO; 2005. 7. Nguyen N, Holodniy M. HIV infection in the elderly. Clin aWHOQoL: The World Health Organization Quality of Life score Interv Aging 2008; 3: 453Á72. was calculated and then divided into quintiles. A better quality of 8. Quinn TC. HIV epidemiology and the effects of antiviral therapy on long-term consequences. AIDS 2008; 22: S7Á12. life was defined as those included in the three lowest quintiles; 9. Wittenberg M, Collinson MA. Household transitions in rural while a worse quality of life was defined as those included in the South Africa, 1996Á2003. Scand J Public Health Suppl 2007; 69: two highest quintiles. 130Á7. 10. Schatz EJ. ‘‘Taking care of my own blood’’: older women’s Measuring health status, functionality and QoL at the relationships to their households in rural South Africa. Scand J population level in older people is important to under- Public Health Suppl 2007; 69: 147Á54. 11. Statistics South Africa. Mid-year population estimates, stand the health, welfare and social support needs of this South Africa: 2006. Available from: http://www.statssa.gov.za growing proportion of the population. As the Agincourt [cited 27 May 2010]. population continues to age, along with millions living in 12. Statistics South Africa. Mid-year population estimates, South similar rural settings, it will become increasingly important Africa: 2009. Available from: http://www.statssa.gov.za [cited 27 May 2010]. for health and social services to adapt and improve in order 13. Collinson MA, Tollman SM, Kahn K. Migration, settlement to provide effective care for a growing older population change and health in post-apartheid South Africa: triangulating

34 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 Cross-sectional survey of older people in rural South Africa

health and demographic surveillance with national census data. their hosts in rural South Africa. Int J Epidemiol 2004; 33: Scand J Public Health Suppl 2007; 69: 77Á84. 1271Á8. 14. Kahn K, Garenne ML, Collinson MA, Tollman SM. Mortality 25. Cieza A, Bickenbach J, Chatterji S. The ICF as a conceptual trends in a new South Africa: hard to make a fresh start. Scand J platform to specify and discuss health and health-related Public Health Suppl 2007; 69: 26Á34. concepts. Gesundheitswesen 2008; 70: e47Á56. 15. Zwang J, Garenne M, Kahn K, Collinson M, Tollman SM. 26. Reeve B. Applying item response theory to enhance health Trends in mortality from pulmonary tuberculosis and HIV/ outcomes assessment. Qual Life Res 2007; 16: 1Á3. AIDS co-infection in rural South Africa (Agincourt). Trans R 27. WHOQOL Group. Study protocol for the World Health Soc Trop Med Hyg 2007; 101: 893Á8. Organization project to develop a Quality of Life assessment 16. Tollman SM, Kahn K, Sartorius B, Collinson MA, Clark SJ, instrument (WHOQoL). Qual Life Res 1993; 2: 153Á9. Garenne ML. Implications of mortality transition for primary 28. WHOQOL Group. The World Health Organization Quality of health care in rural South Africa: a population-based surveil- Life assessment (WHOQoL): position paper from the World lance study. Lancet 2008; 372: 893Á901. Health Organization. Soc Sci Med 1995; 41: 1403Á9. 17. Menken J, Cohen B, Mann AS, Ezeh A, Kaseke E, Kuate- 29. Case A, Menendez A. Does money empower the elderly? Defo B, et al. Aging in Sub-Saharan Africa: recommendations Evidence from the Agincourt demographic surveillance site, for furthering research. In: Cohen B, Menken J, eds. Aging in South Africa. Scand J Public Health Suppl 2007; 69: 157Á64. Sub-Saharan Africa: recommendations for furthering research. 30. Jelsma J, Mkoka S, Amosun SL. Health-related quality of life Panel on Policy Research and Data Needs to Meet the (HRQoL) domains most valued by urban IsiXhosa-speaking Challenge of Aging in Africa. Washington, DC: The National people. Qual Life Res 2008; 17: 347Á55. Academies Press; 2006, pp. 9Á51. 31. Kahn K, Collinson M, Hargreaves J, Clark S, Tollman S. Socio- 18. Lam D, Murray L, Vimal R. Labor force withdrawal of the economic status and child mortality in a rural sub-district of elderly in South Africa. In: Cohen B, Menken J, eds. Aging South Africa. In: de Savigny D, Debpuur C, Mwageni E, in Sub-Saharan Africa: recommendations for furthering re- Nathan R, Razzaque A, Setel P, eds. Measuring health equity in search. Panel on Policy Research and Data Needs to Meet the small areas-findings from Demographic Surveillance Sites. Challenge of Aging in Africa. Washington, DC: The National Farnham, Surrey: Ashgate Publishing; 2005, pp. 67Á86. Academies Press; 2006 pp. 214Á49. 32. Jasso G, Massey DS, Rosenzweig MR, Smith JP. Immigrant 19. Samson M MacQuene K van Niekerk I. Inter-regional inequal- Health Selectivity and Acculturation. In: Anderson NB, Bulatao ity facility. Policy brief 1. Social grants. South Africa; 2006. RA, Cohen B, eds. Critical perspectives on racial and ethnic Available from: http://www.odi.org.uk/resources/download/ differences in health in late life. Washington, DC: National 1067.pdf [cited 30 May 2010]. Academy Press; 2004, pp. 227Á66. 20. Department of Social Development (The Government of South 33. Phaladze NA, Human S, Dlamini SB, Hulela EB, Hadebe IM, Africa). Old age pensions. Available from: http://www.capegate Sukati NA, et al. Quality of life and the concept of ‘‘living well’’ way.gov.za/eng/directories/services/11586/47491 [cited 30 May with HIV/AIDS in sub-Saharan Africa. J Nurs Scholar 2005; 2010]. 37: 120Á6. 21. Collinson MA, Mokoena O, Mgiba N, Kahn K, Tollman SM, 34. Westaway MS, Rheeder P, Gumede T. The effect of type 2 Garenne M, et al. Chapter 16: Agincourt Demographic diabetes mellitus on health-related quality of life (HRQoL). Surveillance System (Agincourt DSS). In: Sankoh O, Kahn K, Curationis 2001; 24: 74Á8. Mwageni E, eds. Population, health and survival at INDEPTH sites. Ottawa: IDRC; 2002. 22. Kahn K, Tollman SM, Collinson MA, Clark SJ, Twine R, *F. Xavier Go´ mez-Olive´ Clark BD, et al. Research into health, population and social MRC/Wits Rural Public Health and Health Transitions Research Unit transitions in rural South Africa: data and methods of the (Agincourt) Agincourt Health and Demographic Surveillance System. Scand School of Public Health J Public Health 2007; 35: 8Á20. Faculty of Health Sciences 23. WHO. Available from: http://www.who.int/healthinfo/systems/ University of the Witwatersrand sage/en/index1.html [cited 7 September 2009]. 7 York Road, Parktown 2193 24. Hargreaves JR, Collinson MA, Kahn K, Clark SJ, Tollman SM. Johannesburg, South Africa Childhood mortality among former Mozambican refugees and Email: [email protected]

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2126 35 æINDEPTH WHO-SAGE Supplement Health status and quality of life among older adults in rural Tanzania Mathew A. Mwanyangala1,2*, Charles Mayombana1,2,3, Honorathy Urassa1,2, Jensen Charles1,2, Chrizostom Mahutanga1,2, Salim Abdullah1,2,3 and Rose Nathan1,2,3

1Ifakara Site Health Institute, Ifakara, Morogoro, Tanzania; 2INDEPTH Network, Accra, Ghana; 3Mikocheni Office, Ifakara Health Institute, Tanzania

Background: Increasingly, human populations throughout the world are living longer and this trend is developing in sub-Saharan Africa. In developing African countries such as Tanzania, this demographic phenomenon is taking place against a background of poverty and poor health conditions. There has been limited research on how this process of ageing impacts upon the health of older people within such low- income settings. Objective: The objective of this study is to describe the impacts of ageing on the health status, quality of life and well-being of older people in a rural population of Tanzania. Design: A short version of the WHO Survey on Adult Health and Global Ageing questionnaire was used to collect information on the health status, quality of life and well-being of older adults living in Ifakara Health and Demographic Surveillance System, Tanzania, during early 2007. Questionnaires were administered through this framework to 8,206 people aged 50 and over. Results: Among people aged 50 and over, having good quality of life and health status was significantly associated with being male, married and not being among the oldest old. Functional ability assessment was associated with age, with people reporting more difficulty in performing routine activities as age increased, particularly among women. Reports of good quality of life and well-being decreased with increasing age. Women were significantly more likely to report poor quality of life (odds ratio 1.31; pB0.001, 95% CI 1.15Á 1.50). Conclusions: Older people within this rural Tanzanian setting reported that the ageing process had significant impacts on their health status, quality of life and physical ability. Poor quality of life and well-being, and poor health status in older people were significantly associated with marital status, sex, age and level of education. The process of ageing in this setting is challenging and raises public health concerns. Keywords: health status; quality of life; older people; ageing; Health and Demographic Surveillance System; INDEPTH WHO-SAGE

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘SAGE data’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 20 November 2009; Revised: 29 March 2010; Accepted: 8 July 2010; Published: 27 September 2010

uman populations throughout the world are people will reside in developing countries (1). The living longer than ever before Á but this is a number of older people is growing rapidly in sub-Saharan Hrelatively new phenomenon in developing coun- Africa (2). Changes in the ageing process within devel- tries. It is estimated that nearly 63% of the population oping countries have been observed through shifts in aged 60 and over are living in developing countries, and population age composition. This process is associated further projected that by 2050 nearly 1.5 billion older with rapid declines in fertility and mortality (3). In the

Global Health Action 2010. # 2010 Mathew A. Mwanyangala et al. This is an Open Access article distributed under the terms of the Creative Commons 36 Attribution-Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 Health status and quality of life among adults in Tanzania

near future, larger older populations will become ubiqui- many older people still do not access these services tous in Africa (1, 4, 5). Tanzania has a total population of due to inability to prove their age, aggravated by the 34 million of whom 4% are aged 50 and over. It is also limited availability of health services, equipment and among the countries in sub-Saharan Africa with at least expertise (6). 1 million older people, and this proportion is projected to The economies of rural Tanzanian settings are pre- rise to 10% of the total population by 2050 (6, 7). dominantly supported by subsistence agriculture, which Furthermore, the absolute number of people entering the provides little or no pension coverage and limited health older cohort is increasing (7). care services. The age structure of these settings is already In developing African countries such as Tanzania, being impacted by the emigration of younger people to many older people reach retirement age after a lifetime urban areas and the return of older people to rural of poverty and deprivation, poor access to health care environments from urban areas on retirement. and poor diet. This situation can leave them with Current health challenges and existing policies act to insufficient personal savings as a consequence of a hide the situation of older people. A large body of fragile earning history (8, 9). In most developing research has described the process of ageing using countries, formal social security systems have only contrasting perspectives: demographic characteristics, limited coverage and inadequate benefit payments (10, physical health, cognitive impairment, disability and 11). As a result, the majority of older people depend self-perceived health of older people in developed coun- on family support networks, a reality that is well tries (22Á24). In the developing world, studies of popula- appreciated in most parts of sub-Saharan Africa (12Á tion ageing have been focused primarily on Asia and 14). Furthermore, it is recognised that traditional social Latin America. In Tanzania there has been limited security systems are evolving, attenuating and rapidly research on explaining process of societal ageing and disappearing due to pressures from urbanisation, in- impact on the health of older people, especially in rural dustrialisation and HIV/AIDS (15). At the same time it settings where people are most beset by poverty and poor is widely reported that older people have more sub- health conditions. This study aims to describe the impact stantial inter-individual variability in health related to age than do younger people (16, 17). The health care of ageing on the health status and well-being of older system spends a small fraction of the budget on people in a rural Tanzanian population using data treating older adult illness and access to care is limited collected by the Ifakara Health Institute’s Health and and not a policy priority in most developing countries Demographic Surveillance System (HDSS) in collabora- (6, 18Á20). tion with the INDEPTH Network and the WHO Survey Within developing countries the demographic transi- on Adult Health and Global Ageing (SAGE). Our aim tion towards older populations is likely to constrain was to provide a better understanding of the health and future health care systems. The attitude of health care well-being of older people in developing countries. The providers towards older people makes their situation resulting information will provide a baseline for examin- even more difficult. It has been reported that older ing the relationship between ageing and other health people in Tanzania are frequently mistreated by health outcomes during demographic transition in these settings. care providers when they seek care (21). Although This will help to raise awareness about the predicament of provision of free health services to older people is older people, support possible policy interventions and stipulated in the Tanzanian National Ageing policy, stimulate further research.

Fig. 1. Maps of Africa, Tanzania and the Ifakara HDSS area.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 37 Mathew A. Mwanyangala et al.

Design instrument. A copy of the INDEPTH WHOÁSAGE summary questionnaire is available as a supplementary Geography of the HDSS area file. Additional data targeted for inclusion into the final The Ifakara HDSS area is located in southern Tanzania data set, derived directly from the HDSS, included socio- in parts of the Kilombero and Ulanga districts, both in demographic characteristics, such as age, sex, education, the Morogoro region (latitude 8.08Á8.68 S and longitude marital status, socio-economic status and household 35.98Á36.68 E), as shown in Fig. 1. The Ifakara DSS information, such as the household size. covers an area of 2,400 km2 in the Kilombero Valley. The HDSS site was initiated in September 1996. A Health status information baseline census was conducted between September and Health status scores were calculated based on health December 1996 in 25 villages covering a population of responses in eight health domains covering affect, cogni- about 93,000 people living in 19,000 households. Since tion, interpersonal activities and relationships, mobility, January 1997, each household has been visited once every pain, self-care, sleep/energy and vision. Each domain 4 months to record births and pregnancies, deaths and included at least two questions. Asking more than one migrations. In order to document community-based question about difficulties in a given domain provides causes of death, the HDSS has conducted verbal more robust assessments of individual health levels and autopsies since 2002. reduces measurement error for any single response item. The area is predominately rural with scattered house- Item Response Theory (IRT) was used to score the holds. Many local houses have brick walls but only 34% responses to the health questions using a partial credit have a corrugated iron roof. The main ethnic groups are model which served to generate a composite health status Wapogoro, Wandamba, Wabena, Wahehe and Wam- score (25, 26). An item calibration was obtained for each bunga, with several other smaller groups. Most of the item. In order to determine how well each item con- inhabitants are Christian or Muslim. All residents speak tributed to common global health measurement, chi- the Kiswahili language. Subsistence farming of maize, squared fit statistics were calculated. The calibration for rice and cassava occupies the majority of the population. each of the health items was taken into account and the Fishing is also common both for local consumption and raw scores were transformed through Rasch modelling shipping to other towns within the country. into a continuous cardinal scale where a score of zero represents worst health and a maximum score of 100 Data collection represents best health. More details on the application of In January 2007, all households with people aged 50 and the IRT approach to computing patient-reported health over were identified from the Ifakara HDSS database. outcomes are described in Chang and Reeve, and These households were subsequently visited to interview Kyobungi (27Á31). The IRT has been judged as among these older people. The questionnaires and the consent the most efficient, reliable and valid methods to evaluate forms were translated to Kiswahili. All field workers were measures of health (32Á37). trained for 3 days prior to conducting the interviews, including 1 day of tool piloting. Surveys started in the Quality of life and well-being middle of January 2007 and ended in April 2007. During In this study we define quality of life as individual field work, interviewers were closely supervised by field perceptions of life in the context of local culture and supervisors who accompanied them on interviews, per- value systems, as well as in relation to goals, expectations, formed spot-checks and re-interviewed where appropri- standards and concerns. An 8-item version of the ate. Also, desk checks on the completed questionnaires WHOQoL instrument was used to assess perceived were done to identify errors before computer data entry. well-being (38). This is a cross-culturally valid instrument All questionnaires that raised queries were returned to for comprehensively assessing overall subjective well- interviewers for clarification in the field. Data entry was being, yet is also very brief. It recognises that health conducted using a double entry system in CSPro. Verbal and quality of life are strongly associated yet distinct informed consent was obtained from all older people who concepts. Results from the 8-items were summed to get an participated in this study. All individuals were inter- overall WHOQoL score which was then transformed to a viewed using the WHO-abbreviated survey instrument 0Á100 scale, similar to the health status score. The short module adapted from the full SAGE questionnaire: WHOQoL instruments have been used in other studies the health status and associated vignette questions plus of older people in Africa (39, 40). Activities of Daily Living (ADL)-type questions (follow- ing the WHO Disability Assessment Scale version II Functional status assessment [WHODAS-II] model), and questions on subjective Personal functioning was assessed through the standar- well-being as measured by the 8-item version of the dised 12-item WHODAS-II. It is a well-tested instrument, World Health Organization Quality of Life (WHOQoL) with published psychometric properties, and a good

38 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 Health status and quality of life among adults in Tanzania

predictor of global disability (41Á43). The WHODAS is Table 1. Background characteristics of study subjects compatible with the International Classification of Functioning, Disability and Health (ICF) and contains Variables Respondents Non-respondentsa many of the most commonly asked ADL and Instru- (n5,131) (n3,075) mental Activities of Daily Living (IADL) questions. The WHODAS instrument also provides an assessment of Sex (%) severity of disability. Results from the 12-items were Men 47.8 52 summed to get an overall WHODAS score, which was Women 52.2 48 then transformed to a 0Á100 scale, with zero represent- Mean age (years) (SD) 62.6 (9.2) 61.3 (7.8) ing no disability. Since this scale runs counterintuitively Age group (years) to the WHOQoL and health status scores, it was 50Á59 43.7 48.5 inverted to a scale designated here as WHODASi, in 60Á69 32.8 33.2 which 100 represents the best situation, i.e. no disability, 70Á79 18.2 17.9 and which thus represents a measure of functional 80 and over 5.3 0.3 ability. Education level (%) Socio-economic status of households No formal education 39.3 41.4 The socio-economic status of households was assessed by Less than or equal to 56.6 45.2 constructing a household wealth index based on house- six years hold asset ownership, level of education of the head of More than six years 4.1 13.3 household and household characteristics, as proposed Marital status (%) and validated by Filmer (44). Data on asset ownership Currently single 33.3 29.0 were collected within the HDSS framework. In current partnership 66.7 71.0

Socio-economic quartile (%) Data analysis Lowest quartile 19.2 19.6 Data were analysed using Stata version 10. Simple cross- Second quartile 19.4 23.7 tabulations and multivariate analysis were done to Third quartile 21.1 19.9 describe the situation of ageing, health status, physical Highest quartile 40.3 36.7 disability, quality of life and well-being of older people. Mean no. of household 10.4 (6.0) The median values for health status, WHOQoL and members (sd) WHODASi were computed, and used to define cut-off Percentage of household 22.9 points for assessing good or poor status. Mean scores members aged 50 years were calculated for different sex and age groups. In order and over to investigate the factors associated with health and quality of life, univariate and multivariate models were aIncludes those listed in the HDSS database who had out- run. In both models, social and demographic variables migrated or died prior to interview visit, and those who did not were fitted as possible explanatory variables. Principal respond for other reasons. component analysis (PCA) was conducted on household characteristics and asset ownership data to investigate associations between these variables at the household with a standard deviation of 9.2. The majority of people level. Wealth index quartiles were constructed to investi- in this study were within the 50Á59 age group, and 67% of gate associations between health status and household the respondents were married, while 39% of respondents wealth. had no formal education. In the majority of households (54%), less than 25% of household members were Results 50 years old or above. The mean size of households A total of 8,206 older people from 3,914 households were where older people lived was 10.4 (standard deviation identified from the Ifakara DHSS. In visits, 63% were 6.0). Only 2% of households were composed solely of successfully interviewed (n5,131). The majority of non- older people living on their own. responders were men (52%) in the 50Á59 age group. The reasons for non-response included hearing impairment, out-migration, refusal, death and absence during the day Functional status assessment and quality of life of the interview. Characteristics of responders and non- The mean and median quality of life scores (WHOQoL) responders are shown in Table 1. were 68.2 and 68.8, respectively, with the proportion below Among those interviewed, the majority were women the median decreasing with increasing age (Table 2). The (n2,668). The mean age of respondents was 62.5 years mean and median functional ability scores (WHODASi)

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 39 Mathew A. Mwanyangala et al.

Table 2. Distribution of quality of life (WHOQoL) and func- Factors associated with poor quality of life and tional ability (WHODASi) outcomes by age and sex health status Odds ratios for below-median quality of life and health Variables Men (n2,463) Women (n2,668) status showed significant associations with being female, older and unmarried (Tables 4 and 5). Women were more Mean WHOQoL score (SD) likely to report poor health as well as being scored for 50Á59 years 69.3 (5.6) 68.8 (6.6) lower quality of life than men. Lower quality of life was 60Á69 years 68.4 (5.9) 67.6 (6.9) also significantly associated with the two lower socio- 70Á79 years 67.0 (7.3) 67.2 (9.4) economic quartiles. However, no association between 80 years and over 64.3 (7.1) 66.1 (11.7) socio-economic status and self-reported health was Percentage of respondents with WHOQoL less than median evident in multivariate analysis controlling for other 50Á59 years 28.8 37.0 factors (Table 5). Age composition within households 60Á69 years 39.1 50.3 and education were not appreciably associated with either 70Á79 years 52.8 59.7 quality of life or health status in multivariate analyses. 80 years and over 67.9 71.2

Mean WHODASi score (SD) Discussion 50Á59 years 90.4 (13.4) 87.5 (14.4) This study observed that among older adults men 60Á69 years 87.1 (14.9) 82.2 (16.2) reported better health status than women, and that 70Á79 years 80.5 (18.1) 74.0 (21.3) health status, quality of life and physical ability 80 years and over 68.4 (22.1) 59.0 (24.9) deteriorated markedly with increasing age. This is in line with empirical knowledge of the physiological Percentage of respondents with WHODASi less than median processes of ageing and linked to disease and ill health. 50Á59 years 35.0 43.9 These results underscore the reality of existing gender 60Á69 years 45.2 61.2 biases in relation to economic power, which may be the 70Á79 years 62.0 73.5 product of lower levels of education and savings, and 80 years and over 82.1 86.5 the poorer life-time earning histories many women have (45). The results are consistent with those reported recently by the Tanzanian Ministry of Health and were 84 and 90, respectively. Functional ability was lower Social Welfare, which found that older people make among women than men in all age groups. up around one-third of all disabled people in Tanzania (46). Higher quality of life and good health status was associated with being married, a high level of education Distributions of health status and higher socio-economic status of the household. The median health status score of the surveyed popula- This reinforces the hypothesis that individual health is tion was 68.4. Health status was associated with age and improved by education, possibly due to having greater gender (Table 3). Poor health status was associated with access to information on health, better eating habits increasing age and among women. and self-care (47, 48). These results reveal sex differences in longevity, with larger numbers of women than men aged 50 and over, Table 3. Distribution of self-reported health status outcomes despite their poorer health outcomes. The mean house- by age and sex hold size of 10 observed for households containing older people in this study area is broadly reflective of Variables Men (n2,463) Women (n2,668) socio-cultural practices in rural areas of most countries in sub-Saharan Africa, where older people tend to Mean health status score (SD) live in extended family households rather than inde- 50Á59 years 74.5 (13.0) 72.1(12.1) pendently (49). This is reflective of the current Tanza- 60Á69 years 71.5 (12.2) 68.4 (10.3) nia Ageing policy which prioritises family as the basic 70Á79 years 67.1 (11.2) 64.5 (11.0) institution of care and support for older people (50). 80 years and over 61.3 (10.2) 58.5 (9.2) Few studies have been conducted on adult health and ageing in Tanzania. The approach of assessing individual Percentage of respondents with health status less than median health status based on self-reported health status has 50Á59 years 34.8 41.3 been criticised by various scholars, and it has been 60Á69 years 43.8 54.2 suggested that self-reported health status should not be 70Á79 years 60.0 66.8 used to estimate disease prevalence and identify indivi- 80 years and over 82.7 84.7 duals with disease (47, 51). Thus, although the current

40 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 Health status and quality of life among adults in Tanzania

Table 4. Factors associated with below-median quality of life (WHOQoL)

Variables Univariate model (OR and 95% CI) p-value Multivariate model (OR and 95% CI) p-value

Sex Men 1 Women 1.37 (1.22Á1.53) pB0.001 1.27 (1.11Á1.45) pB0.001

Age group (years) 50Á59 1 1 60Á69 1.63 (1.43Á1.86) pB0.001 1.57 (1.38Á1.80) pB0.001 70Á79 2.60 (2.22Á3.04) pB0.001 2.37 (2.01Á2.80) pB0.001 80 4.52 (3.44Á5.92) pB0.001 4.33 (3.26Á5.75) pB0.001

Education level No formal education 1.63 (1.22Á2.19) p0.001 1.17 (0.86Á1.60) p0.315 Less than or equal to six years 1.46 (1.30Á1.64) pB0.001 1.03 (0.76Á1.39) p0.845 More than six years 1

Marital status Now single 1.62 (1.44Á1.82) pB0.001 1.19 (1.04Á1.37) p0.010 In current partnership 1 10

Proportion aged 50 years and over in the same household (%) B25 0.79 (0.63Á0.98) p0.035 0.92 (0.69Á1.23) p0.575 25Á49 0.80 (0.63Á1.00) p0.049 0.96 (0.75Á1.23) p0.749 50Á74 0.86 (0.65Á1.13) p0.272 1.05 (0.83Á1.33) p0.697 575 1 1

Socio-economic quartile Lowest quartile 0.71 (0.61Á0.82) pB0.001 0.71 (0.69Á0.99) p0.042 Second quartile 0.61 (0.52Á0.71) pB0.001 0.62 (0.63Á0.87) pB0.001 Third quartile 0.81 (0.70Á0.94) p0.006 0.75 (0.75Á1.03) p0.118 Highest quartile 1 1

study indicates a clear association between older people’s Reduction (57), older people are not fully considered. perception of age and health, further medically based Older people are widely recognised as being a valuable studies are required to confirm the health burden of older source of information, knowledge and experience. people in rural Tanzania. Following up this sample over Thus, attempts should be made to consider and time would be useful to see how these data relate to improve their health status and quality of life within subsequent health outcomes. this and other rural settings in Tanzania and other Several studies have shown socio-economic status to be developing countries. associated with older people’s health status, quality of life and well-being (52Á54). However, the current study also detected an association between household socio- Conclusion economic status and quality of life, but not between The health status and quality of life of older people in wealth and self-reported health description. Similar rural Tanzania is reduced significantly during the observations have been documented elsewhere (55), and ageing process. Perceptions of physical disability also may be due to the fact that household asset-based wealth increase with age in this population. Poor quality of life indices can be unrelated to individual health status, and well-being, and health status in older people are depending on which member of the household is head significantly related to marital status, sex and age. and who owns assets (56). Specifically, quality of life decreases with age, and Although Tanzania is the second country in Africa women experience poorer quality of life and a greater to have a national Ageing policy, after Mauritius, many burden of physical disability than men. Thus, the issues related to older people are not yet fully defined. process of ageing presents a clear public health For example, even in the National Strategy for Poverty challenge in this setting.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 41 Mathew A. Mwanyangala et al.

Table 5. Factors associated with below-median health status responses

Variables Univariate model (OR and 95% CI) p-value Multivariate model (OR and 95% CI) p-value

Sex Men 1 1 Women 1.28 (1.14Á1.44) pB0.001 1.33 (1.15Á1.52) pB0.001

Age group (years) 50Á59 1 1 60Á69 1.56 (1.36Á1.180) pB0.001 1.57 (1.36Á1.81) pB0.001 70Á79 2.98 (2.53Á3.49) pB0.001 2.96 (2.50Á3.51) pB0.001 80 8.95 (6.71Á11.96) pB0.001 8.96 (6.64Á12.09) pB0.001

Education level No formal education 1.74 (1.27Á2.40) p0.001 1.24 (0.88Á1.74) p0.27 Less than or equal to 6 years 1.32 (0.97Á1.82) p0.082 1.25 (0.90Á1.74) p0.180 More than 6 years 1 1

Marital status Now single 1.57 (1.39Á1.77) pB0.001 1.16 (1.00Á1.33) p0.045 In current partnership 1 1

Proportion aged 50 years and over in the same household (%) B25 0.94 (0.74Á1.20) p0.633 1.21 (0.93Á1.58) p0.147 25Á49 1.06 (0.82Á1.37) p0.644 1.21 (0.92Á1.59) p0.162 50Á74 1.11 (0.79Á1.56) p0.558 1.17(0.82Á1.68) p0.384 575% 1 1

Socio-economic quartile Lowest quartile 0.92 (0.78Á1.08) p0.293 1.13 (0.95Á1.34) p0.176 Second quartile 0.72 (0.62Á0.85) pB0.001 0.89 (0.75Á1.60) p0.206 Third quartile 0.84 (0.72Á0.97) p0.022 0.92 (0.78Á1.08) p0.296 Highest quartile 1 1

Acknowledgements 3. Mbamaonyeukwu CJ. Africa’s ageing populations. J Int Instit Ageing 2001; 11: 2Á7. We would like to thank the Kilombero and Ulanga district councils 4. Mba C. Determinants of living arrangements of Lesotho’s older for their support to the Ifakara HDSS. We extend our gratitude to persons female population. J Int Women’s Stud 2002; 3: 1Á22. the leadership of Mlabani village for allowing us to pilot test the 5. United Nations, Population Division, Affairs. DoEaS. Popula- survey tools. We highly appreciate the hard work and commitment tion ageing. New York: United Nations, Population Division, of the HDSS field and data management teams. We are indebted to Affairs; 1999. the respondents who voluntarily offered their time for interviews and 6. URT. The Ministry of labour, youth Development and Sports shared the useful information without which the survey would not National Aging Policy. In: Ministry of labour, Youth Develop- have been possible. We are thankful to the INDEPTH Network and ment and Sports, eds. United Republic of Tanzania Population WHO Survey on Adult Health and Global Ageing (SAGE). and Housing cenus. Dar es Salaam: United Republic of Tanzania; 2003, pp. 3Á7. Conflict of interest and funding 7. URT. Population and housing census. Dar es Salaam: United Republic of Tanzania, National Bureau of Statistics; Funding support for the HDSS was provided by the 2002. Swiss Development Corporation, Norvatis Foundation, 8. Charton KE, Rose D. Nutrition among older adults in Africa: USAID and the Tanzanian Ministry of Health and Social the situation at the beginning of the millennium. J Nutri 2001; Welfare, which is highly appreciated. 131: 245Á85. 9. Kimokoti RW, Hamer DH. Nutrition, health, and aging in sub Saharan Africa. J Nutrition 2008 2008; 66: 611Á23. References 10. Bailey C. Governance of social security schemes: social security documentation. International Social Security Association 2000; 1. UN. World population prospects: the 2000 revision. New York: 21: 71Á113. UN; 2001. 11. Colin G, Turner J, Bailey C, Latulippe D. Social security 2. Mba CJ. Nigeria’s ageing population: a call for attention. J Int pensions: development and reform. Geneva: International Instit Ageing 2001; 12: 15Á24. Labour Office; 2000.

42 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 Health status and quality of life among adults in Tanzania

12. Van de Walle E. African households: censuses and surveys. 33. Saliba D, Orlando M, Wenger NS, Hays RD, Rubenstein LZ. London: Word Bank; 2006. Identifying a short functional disability screen for older persons. 13. Kaseke E. An overview of formal and informal social security J Gerontol A Biol Sci Med Sci 2000; 55: M750Á6. systems in Africa. USA and Johannesburg, South Africa: 34. Saha TD, Compton WM, Pulay AJ, Stinson FS, Ruan WJ, National Academy of Sciences; 2004. Smith SM et al. Dimensionality of DSM-IV nicotine depen- 14. Deaton A, Paxson CH. Patterns of aging in Thailand and Coˆte dence in a national sample: an item response theory application. d’Ivoire. In: Wise DA, ed. Topics in the economics of aging. Drug Alcohol Depend 2009; 108: 21Á8. Chicago, IL: University of Chicago Press; 1992, pp. 163Á206. 35. Revicki DA, Cella DF. Health status assessment for the twenty- 15. Tostensen A. Towards feasible social security systems in sub- first century: item response theory, item banking and computer Saharan Africa. In: Grimm M, ed. Bergen: World Bank; 2004, adaptive testing. Qual Life Res 1997; 6: 595Á600. p. 14. 36. Jenkinson C, Fitzpatrick R, Garratt A, Peto V, Stewart-Brown 16. Fialova D, Onder G. Medication errors in elderly people: S. Can item response theory reduce patient burden when contributing factors and future perspectives. Br J Clin Pharma- measuring health status in neurological disorders? Results col 2009; 67: 641Á5. from Rasch analysis of the SF-36 physical functioning scale 17. Bedard M, Gibbons C, Dubois S. The needs of rural and urban (PF-10). J Neurol Neurosurg Psychiatry 2001; 71: 220Á4. young, middle-aged and older adults with a serious mental 37. Reeve BB, Hays RD, Bjorner JB, Cook KF, Crane PK, Teresi illness. Can J Rural Med 2007; 12: 167Á75. JA et al. Psychometric evaluation and calibration of health- 18. Feachem RGA, Kjellstrom T, Murray CJL, Over M, Phillips related quality of life item banks: plans for the Patient-Reported MA. The health of adults in the developing world. New York: Outcomes Measurement Information System (PROMIS). Med Oxford University Press; 1992. Care 2007; 45: S22Á31. 19. Poullier JP, Hernandez P, Kawabata K. Health systems perfor- 38. Schmidt P, Clouth J, Haggenmuller L, Naber D, Reitberger U. mance assessment: debates, methods, and empiricism. In: Evans Constructing an index for the Subjective Well-being Under ICJLMaDR, ed. National health accounts: concepts, data Neuroleptics scale (SWN), short form: applying structural sources, and methodology. Geneva: World Health Organization; equation modeling for testing reliability and validity of the 2003, pp. 185Á93. index. Qual Life Res 2006; 15: 1191Á202. 20. Tollman S, Doherty J, Mulligan JA. General primary care. In: 39. Gureje O, Kola L, Afolabi E, Olley BO. Determinants of quality Jamison DT, Jamison JG, Measham AR, Alleyne G, Claeson of life of elderly Nigerians: results from the Ibadan study of M, Evans DB, Jha P, Mills A, Musgrove P, ed. Disease control ageing. Afr J Med Med Sci 2008; 37: 239Á47. priorities in developing countries. Washington, DC: Interna- 40. Sen A. Public action and the quality of life in developing tional Bank for Reconstruction and Development and the countries. Oxf Bull Econ Stat 1981; 43: 287Á319. World Bank; 2006, pp. 1193Á209. 41. O’Donovan DK. Angle on disability. Lancet 1989; 2: 866. 21. HelpAge. HelpAge International targets 2005Á2010. Johannes- 42. Sousa RM, Ferri CP, Acosta D, Albanese E, Guerra M, Huang burg: HelpAge; 2007. Y et al. Contribution of chronic diseases to disability in elderly 22. O’ Reilly D, Rosato M, Petterson C. Self reported health and people in countries with low and middle incomes: a 10/66 mortality: ecological analysis on based on electoral wards across dementia research group population-based survey. Lancet 2009; the United Kingdom. BMJ 2005; 331: 988Á9. 374: 1821Á30. 23. Xie J, Mathews FE, Jagger C, Bond J, Brayne C. The oldest old 43. Luciano JV, Ayuso-Mateos JL, Fernandez A, Aguado J, in England and Wales: a descriptive analysis based on MRC Serrano-Blanco A, Roca M et al. Utility of the twelve-item cognitive function and ageing study. Age and Ageing 2008; 37: World Health Organization Disability Assessment Schedule II 396Á402. 24. Xu X, Jensen G. Health effects of managed care among the near (WHO-DAS II) for discriminating depression ‘‘caseness’’ and older persons. J Aging Health 2008; 18: 507. severity in Spanish primary care patients. Qual Life Res 2009; 25. Wilson TR, Alexander DJ, Kind P. Measurement of health- 19: 97Á101. related quality of life in the early follow-up of colon and rectal 44. Filmer D, Pritchett LH. Estimating wealth effects without cancer. Dis Colon Rectum 2006; 49: 1692Á702. expenditure data Á or tears: an application to educational 26. Andrich D. Controversy and the Rasch model: a characteristic enrolments in states of India. Demography 2001; 38: 115Á32. of incompatible paradigms? Med Care 2004; 42: I7Á6. 45. Kinsella K, Phillips DR. Global aging: the challenge of success. 27. Chang C, Reeve B. Item responses theory and its applications to Washington, DC: Population Reference Bureau; 2005. patient-reported outcomes measurements. Eval Health Prof 46. Welfare MoHaS. Tanzania Disability survey. Johannesburg: 2005; 28: 264Á82. World bank; 2009. 28. Kyobutungi C, Ezeh AC, Zulu E, Falkingham J. HIV/AIDS and 47. Goldman N, Glei DA. Sex differences in relationship between the health of older people in the slums of Nairobi, Kenya: DHEAS and health experimental. Gerontology 2007; 42: results from a cross sectional survey. BMC Public Health 2009; 979Á87. 9: 153. 48. Herd P, Goesling B, House JS. Socioeconomic position and 29. Dolan P, Lee H, King D, Metchelfe R. How does NICE value health: the differential effects of education versus income on the health? BMJ 2009; 339: b2577. onset versus progression of health problems. Journal of Health 30. Kahneman D, Krueger AB, Schkade DA, Schwarz N, Stone and Social Behavior 2007; 48: 223Á38. AA. A survey method for characterizing daily life experience: 49. Van de Walle E. African households: censuses and surveys. the day reconstruction method. Science 2004; 306: 1776Á80. London: World Bank; 2006. 31. Salomon JA, Murray CJL. A multi-method approach to 50. United Republic of Tanzania. The Ministry of Labour, Youth measuring health-state evaluations. Health Economics 2003; Development and Sports National Aging Policy. In: Ministry of 13: 281Á90. Labour, Youth Development and Sports, ed. Dar es Salaam: 32. Ware JE Jr., Kosinski M, Bjorner JB, Bayliss MS, Batenhorst A, United Republic of Tanzania; 2003. Dahlof CG et al. Applications of computerized adaptive testing 51. Coster W, Ludlow L, Mancini M. Using IRT variable maps to (CAT) to the assessment of headache impact. Qual Life Res enrich understanding of rehabilitation data. J Outcome Meas 2003; 12: 935Á52. 1999; 3: 123Á33.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 43 Mathew A. Mwanyangala et al.

52. Marmot M, Brunner E. Cohort profile: the Whitehall II Study. 56. Smith V, Goldman N. Socioeconomic differences in health International Journal of Epidemiology 2005; 34: 251Á56. among older adults in Mexico. Social Science and Medicine 53. Ye L, Waite LJ. The impact of childhood and adult SES on 2007; 65: 1372Á85. physical, mental, and cognitive well-being in later life. Journals 57. United Republic of Tanzania national Strategy for Growth of Gerontology Series B: Psychological Science and Social and Reduction of Poverty; 2005. Available from: http://www. Science 2005; 60: S93ÁS101. siteresources.worldbank.org/INTPRS1/Resources/TanzaniaPRSP 54. Meich RA, Hauser RM. Socioeconomic status and health at (June-2005).pdf [cited 3 May 2010]. midlife: a comparison of educational attainment with occupa- tion-based indicators. Annals of Epidemiology 2001; 11: 75Á84. 55. Tavallaii SA, Einollahi B, Azizabadi Farahani M, Namdari M. *Mathew A. Mwanyangala Socioeconomic links to health-related quality of life, anxiety, Ifakara Site Health Institute and depression in kidney transplant recipients. Iran J Kidney P.O. Box 53, Ifakara, Morogoro, Tanzania Dis 2009; 3: 40Á4. Email: [email protected]

44 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2142 INDEPTH WHO-SAGE Supplement The health and well-being of older people in Nairobi’s slums Catherine Kyobutungi1,2*, Thaddaeus Egondi1,2 and Alex Ezeh1,2

1African Population & Health Research Centre, Nairobi, Kenya; 2INDEPTH Network, Accra, Ghana

Background: Globally, it is estimated that people aged 60 and over constitute more than 11% of the population, with the corresponding proportion in developing countries being 8%. Rapid urbanisation in sub- Saharan Africa (SSA), fuelled in part by ruralÁurban migration and a devastating HIV/AIDS epidemic, has altered the status of older people in many SSA societies. Few studies have, however, looked at the health of older people in SSA. This study aims to describe the health and well-being of older people in two Nairobi slums. Methods: Data were collected from residents of the areas covered by the Nairobi Urban Health and Demographic Surveillance System (NUHDSS) aged 50 years and over by 1 October 2006. Health status was assessed using the short SAGE (Study on Global AGEing and Adult Health) form. Mean WHO Quality of Life (WHOQoL) and a composite health score were computed and binary variables generated using the median as the cut-off. Logistic regression was used to determine factors associated with poor quality of life (QoL) and poor health status. Results: Out of 2,696 older people resident in the NUHDSS surveillance area during the study period, data were collected on 2,072. The majority of respondents were male, aged 50Á60 years. The mean WHOQoL score was 71.3 (SD 6.7) and mean composite health score was 70.6 (SD 13.9). Males had significantly better QoL and health status than females and older respondents had worse outcomes than younger ones. Sex, age, education level and marital status were significantly associated with QoL, while slum of residence was significantly associated with health status. Conclusion: The study adds to the literature on health and well-being of older people in SSA, especially those in urban informal settlements. Further studies are needed to validate the methods used for assessing health status and to provide comparisons from other settings. Health and Demographic Surveillance Systems have the potential to conduct such studies and to evaluate health and well-being over time. Keywords: Nairobi; slum settlements; older people; ageing; well-being; quality of life; INDEPTH WHO-SAGE

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘SAGE data’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 17 November 2009; Revised: 27 June 2010; Accepted: 8 July 2010; Published: 27 September 2010

he proportion of older people is increasing world- and over constituted 4.9% of the population under wide. Globally, it is estimated that people aged 60 surveillance. It is estimated that people aged 60 and Tand over currently constitute more than 11% of over in Kenya as a whole constituted 4.0% of the total the population; over 20% in developed nations and about population in 2005 and this proportion is expected to 8% in developing ones. The proportion of older people increase to 4.5% by 2015 and to 9.3% by 2050 (3). Older globally is expected to double to 22% by 2050 (1). In people will therefore form an increasingly important sub- Africa, people aged 60 and over account for only 5% of group in numeric terms in developing nations. the population; this is projected to increase to 11% by Older people have traditionally been held in high 2050 (2). In this study setting, people aged 60 and over esteem in many African societies for their wisdom, role constituted 1.6% of the population, and those aged 50 as heads of families and roles in conflict resolution. More

Global Health Action 2010. # 2010 Catherine Kyobutungi et al. This is an Open Access article distributed under the terms of the Creative Commons 45 Attribution-Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 Catherine Kyobutungi et al.

recently, older people have been involved in the fight As the HIV/AIDS pandemic rages in SSA and as slums against HIV/AIDS, especially in their role as caregivers grow in a rapidly urbanising continent, it is important for HIV-infected family members and orphans left behind that the impact of these processes on older people is by deceased relatives. On the other hand, older people assessed and addressed. The intersection between the have not been spared by the direct effects of HIV/AIDS. HIV/AIDS pandemic, population ageing and uncon- A recent AIDS indicator survey in Kenya shows that the trolled urbanisation in SSA will have far-reaching con- HIV prevalence among the 50Á54 years age group is 8% sequences on the social, economic and health spheres of (similar for both males and females). The prevalence for societies. females is similar to that in the 45Á49 age group while for Despite the evident need to understand issues that males, the prevalence is higher in the 50Á54 age group. affect older people in SSA, relative to other demographic The HIV prevalence in urban areas is also higher than in trends, ageing in Africa has only recently started receiving rural areas (8.9% vs. 7.0%) (4) and even higher (11.4%) in attention in research and policy-making. There is a near the study area according to a recent survey (APHRC, absence of policies and programmes targeting older unpublished data). Apart from HIV/AIDS, older people people in most countries in SSA (15), and Kenya is no are also most affected by chronic degenerative diseases. exception. Health policies and programmes are geared This implies that in Kenya and many other countries in towards the traditional vulnerable groups of women of sub-Saharan Africa (SSA), older people most probably reproductive age and children. The current National bear a dual burden of disease. Health Sector Strategic Plan however recognises that Population ageing is occurring in a context of rapid older people have special needs that are different from urbanisation in SSA. Africa is urbanising at a rate faster other adults and hence spells out specific interventions than any other region in the world and by 2030 more than for older people (16). In addition to regular curative and half of the SSA population will live in urban areas (5). preventive services, such interventions include annual The pace of urbanisation in many SSA countries has not screening and provision of curative services for degen- been matched by economic growth. In fact, in countries erative diseases, and counselling for lifestyle changes. It like Kenya, urbanisation has been rapid amid economic remains to be seen whether these interventions have been stagnation. This has resulted in an increase in the number translated into real programmes that serve older people and size of informal settlements or slums in many cities. in health facilities. It is estimated that more than 70% of urban residents in The fact that older people have been long neglected in SSA live in slum or slum-like conditions. In Kenya, this many policies and programmes in Kenya means that percentage is about 71% (6). The informal nature of these there is a dearth of research on their health and well- settlements means that they are underserved by the public being. This study therefore aims to fill the gap in ageing sector in the provision of basic amenities and services research in Africa by describing the health and well-being including health, education, water and sanitation, and of older people living in two Nairobi slums. garbage collection services. Slums are also characterised by high levels of unemployment, overcrowding, insecur- ity, greater involvement in risky sexual practices, social Methods fragmentation, and high levels of mobility (7Á9). Studies from different SSA countries have shown that slum Study setting residents have worse health outcomes than their rural The study was conducted in two slum communities where counterparts (10Á13). For example, childhood mortality the African Population and Health Research Centre in poor urban areas of Zambia and Malawi is higher than (APHRC) is implementing the longitudinal Nairobi in rural and peri-urban areas (11, 14). Desperate living Urban Health and Demographic Surveillance System conditions and lack of livelihood opportunities could (NUHDSS). The NUHDSS covers large parts of the two predispose residents to risky health-related behaviours slums of Korogocho and Viwandani in Nairobi City, such as high alcohol consumption, unsafe sex, smoking Kenya’s capital and commercial centre. Both commu- and other substance abuse. All these factors have adverse nities are informal settlements located about 5Á10 km effects on health which may be compounded by poor from the city centre. The population under surveillance as access to health services. of 1 January 2007 was 59,513 individuals living in 21,993 Ageing in an urban setting, especially a slum settle- households. ment, poses its own challenges. These include weak social The NUHDSS started after an initial census in August networks, neglect and loss of respect and stature that are 2002. Since January 2003, data on core demographic often accorded older people in more stable communities. events (births, deaths, in- and out-migrations) have been It should be expected, therefore, that older people in slum collected and updated every 4 months during routine settlements have poor or even poorer health outcomes Health and Demographic Surveillance System (HDSS) just like other sub-populations therein. rounds.

46 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 The health and well-being of older people in Nairobi’s slums

Data collection tails on how scores for this study were derived are Data for this study was collected from all residents of provided elsewhere (17). In brief, IRT uses Maximum the NUHDSS who were aged 50 years and over as of Likelihood Estimation methods to model the relationship 1 October 2006. Eligible participants (n2,696) were between a person’s health status and their probability of identified from the most up-to-date NUHDSS database responding to each question in a multi-item scale. Each at the time. Data were collected on 2,072 respondents item is modelled to have a set of parameters which who had complete interviews and only these were describe the relationship between the item and the included in the analysis. Out of the 624 who were not measured construct as well as how the item functions interviewed, 102 refused to be interviewed, 27 had died, within a population. The health score is then transformed 213 had out-migrated and no contact was made with the to a scale of 0 to 100 (where 100 is the best health status). rest for various reasons including absence of a competent More details on the application of the IRT approach to respondent, entire household absent for prolonged peri- computing patient-reported health outcomes are avail- ods and unknown whereabouts. The final response rate able in the paper by Chang and Reeve (18). was 84.4% after omitting the 240 older people later found to have died or out-migrated. Statistical analysis Data were collected in the framework of a larger Descriptive analyses were conducted for both measures of study on the linkages between urbanisation, migration, health. For WHOQoL, mean scores were computed for poverty and health over the life course. An interviewer- different categories of respondents. The different cate- administered questionnaire was used to collect data. gories include: sex (male, female), age (age groups: 50Á59, Interviewers had a minimum education level of Form 4 60Á69, 70Á79, 80), educational level (no formal educa- (12 years of schooling) and were residents in the NUHDSS tion, up to 6 years of formal education, more than 6 years area. They were trained over a five-day period followed by of education), marital status (in current partnership, two days of field testing. Each group of five interviewers never married, separated, divorced and widowed), wealth was supervised by a team leader who manually edited all index (quintiles), whether respondent stays alone (Yes, completed forms, conducted random spot checks on at No) and proportion of people aged 50 years and over in least 5% of forms filled by each field worker under his/her the same household (B25%, 25Á49%, 50Á74%, 75%). In supervision, and offered additional training whenever addition, the proportion of respondents in each category necessary. with a WHOQoL score less than the median was Self-reported health status was assessed using the short computed. For the health status score, mean scores form of the individual SAGE (Study on Global Ageing were also calculated and the proportion of respondents and Adult Health) questionnaire, available as a Supple- falling below the overall median score was calculated for mentary File to this paper. Details of how this tool was each category of respondents. developed, validated and adapted for use in this survey Exploratory analyses were conducted to determine the are described elsewhere (17). In brief, this form has factors associated with poor QoL and poor health status. sections on health status descriptions in eight domains of For both measures of health, respondents who had scores health including mobility, self-care, affect, vision, pain below the median were categorised as having poor QoL and discomfort, sleep/energy, interpersonal activities and or poor health, respectively. cognition. Typically, questions ask about how much In order to investigate the effect of non-response, we difficulty the respondent had had in the preceding fitted a logistic regression model using response status as 30 days with tasks or activities in the eight domains. the outcome and key socioeconomic and demographic Responses range from no difficulty to extreme difficulty characteristics as explanatory variables. A completed on a five-item scale. In addition, the SAGE form has interview was defined as response while an incomplete questions on functioning assessment using items in the interview for a participant determined to be resident in Activities of Daily Living / Instrumental Activities of the study area at any time during the survey was Daily Living (ADL/IADL) tool as well as on Subjective considered non-response. Gender, education and wealth Well-being and Quality of Life (QoL). index were found to be associated with non-response. The This paper focuses on two measures of self-reported predicted probability of responding was calculated for health status: QoL and health status scores. The QoL was every individual in the data using the fitted model. Once assessed using the World Health Organization Quality of the predicted probability was calculated, its inverse Life tool (WHOQoL) score, on a scale from 0 to 100 became the weight for that observation. The computed where 100 is the best QoL. Details of how this is weights were re-adjusted to approximately add up to the computed are described elsewhere (17). Health status sample size. These weights were included in subsequent scores were computed using Item Response Theory (IRT) univariate and multivariable logistic regressions using the parameter estimates in Winsteps†, a Rasch measurement categorical health outcomes described above to adjust for software package (http://www.winsteps.com). More de- non-response. The variables found to be associated with

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 47 Catherine Kyobutungi et al.

non-response were also included in the model as pre- NUHDSS database. Marital status for non-respondents dictors. Results are presented for the models adjusted for could not be established since this variable is not non-response. routinely collected and may change over time. There were no major differences between respondents and non- respondents except for wealth index, where a larger Results proportion of non-respondents fell in the poorest wealth The descriptive characteristics of the study participants quintile compared to respondents, and living arrange- are shown in Table 1. The characteristics of non- ments, whereby a quarter of the respondents were staying respondents are also shown. Demographic characteristics alone compared to more than a third of non-respondents. for non-respondents were obtained from the existing These differences were both statistically significant (pB 0.001). Among both respondents and non-respondents, Table 1. Background characteristics of study subject (re- there were more males than females and the majority of spondents and non-respondents) respondents were in the 50Á59 year age group. A majority of the respondents had at least six or more years of Respondents Non-respondents schooling. The average number of household members Variables (N 2,072) (N 384)   for the respondents was about four members per house- Sex (%) hold compared to about three for non-respondents. Men 1,327 (64.4%) 302 (79.1%) Women 745 (36.0%) 80 (20.9%) The distribution of WHOQoL and health status Mean age (SD) 59.2 (9.06) 57.1 (7.5) scores The distribution of WHOQoL and health status scores is Age group shown in Table 2. The median values used as cut-offs 50Á59 years 1,358 (65.4%) 283(73.9%) were 71.9 for WHOQoL and 67.5 for health status. The 60Á69 years 458 (22.1%) 69 (18.0%) higher the WHOQoL score, the better the QoL, and the 70Á79 years 163 (7.9%) 23 (6.0%) higher the health status scores, the better the health 80 years and over 93 (4.5%) 8 (2.1%) status. The mean WHOQoL score was lower for older Education level (%) No formal education 571 (28.7%) 77 (21.4%) Table 2. Distribution of WHOQoL and Health Status Scores Less than or equal 562 (28.2%) 81 (22.5%) by age and sex to 6 years More than 6 years 858 (43.2%) 202 (56.1%) Variables Men (n1,331) Women (n747)

Marital status (%) Mean WHOQoL score (SD) Now single 662 (32.0%) Á 50Á59 years 73.1 (5.8) 70.9 (6.3) In current partnership 1,410 (68.1%) Á 60Á69 years 71.9 (6.4) 68.3 (6.6) 70Á79 years 71.1 (6.2) 65.7 (7.2) Wealth index (%) 80 years and over 67.3 (9.1) 63.8 (8.5) First quintile (Poorest) 518 (25.0%) 177 (46.3%) Second quintile 206 (10.0%) 6 (1.6%) Proportion of respondents with WHOQoL below the median Third quintile 514 (24.8%) 16(4.2%) 50Á59 years 32.0% 45.8% Fourth quintile 453 (21.9%) 69 (18.1%) 60Á69 years 43.9% 64.6% Fifth quintile (Least poor) 380 (18.4%) 114 (29.8%) 70Á79 years 51.9% 79.8% 80 years and over 71.1% 78.2% Mean number of household 4.12 (3.19) 3.0 (2.5) members (SD) Mean health status score (SD) 50Á59 years 74.7 (13.9) 69.7 (12.5) Proportion of household 0.52 (0.34) 0.62 (0.3) 60Á69 years 71.0 (12.9) 63.9 (10.6) members aged 50 70Á79 years 69.0 (13.3) 60.2 (10.6) years and over (SD) 80 years and over 59.3 (15.9) 56.6 (10.9) Stays alone Proportion of respondents with health status score below the Yes 496 (24.0%) 140 (36.5%) median No 1,576 (76.0%) 244 (63.5%) 50Á59 years 33.3% 50.8% Site of residence (%) 60Á69 years 49.1% 73.5% Korogocho 1,462 (70.6%) 214 (55.7%) 70Á79 years 46.8% 83.3% Viwandani 610 (29.4%) 170 (44.3%) 80 years and over 79.0% 90.9%

48 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 The health and well-being of older people in Nairobi’s slums

respondents but with some sex differences. Female Table 3. Factors associated with poor quality of life respondents had, on average, appreciably lower WHO- QoL scores than their male counterparts in the same age Univariate model Multivariate model group. Similar effects were observed when the propor- Variables (OR and 95% CI) (OR and 95% CI) tions of respondents with a WHOQoL score below the median were considered, with poorer QoL associated Site with women and older age groups. Viwandani 0.59 (0.49Á0.72) 0.85 (0.68Á1.07) A similar pattern to that observed for WHOQoL Korogocho 1.00 1.00 scores was observed with the health status scores. The Sex average health status scores decreased with increasing age Men 0.44 (0.36Á0.53) 0.78 (0.61Á1.01) and females have lower scores than males, indicating Women (Ref) 1.00 1.00 worse health status. The proportion with health status scores below the median increased with age, particularly Age group among females. 50Á59 years 1.00 1.00 The results for the two measures of self-reported health 60Á69 years 1. 97 (1.59Á2.45) 1.55 (1.22Á1.96) status consistently showed that health status and QoL 70Á79 years 3.59 (2.48Á4.95) 2.06(1.40Á3.02) deteriorated in both sexes as people got older and that 80 years and over 5.42 (3.33Á8.81) 2.94 (1.71Á5.02) females had significantly worse health outcomes than Education level males. No formal 3.07 (2.46Á3.82) 1.68 (1.29Á2.18) education Factors associated with poor QoL and poor Less than or equal 1.73 (1.39Á2.16) 1.25 (0.98Á1.60) health status to 6 years Both univariate and adjusted logistic regression results More than 6 years 1.00 1.00 using WHOQoL as the outcome are presented in Table 3. (Ref) Male respondents were significantly less likely to have Marital status poor WHOQoL compared to females in the univariate In current 1.00 1.00 models. However in adjusted models, this effect was partnership (Ref) attenuated and was of borderline statistical significance. Never married 1.63 (1.04Á2.54) 1.17 (0.71Á1.92) An age gradient, consistent with the descriptive results, is Separated 2.12 (1.47Á3.04) 1.55 (1.04Á2.31) observed in the logistic regression models. In adjusted Divorced 2.31 (1.40Á3.80) 1.52 (0.87Á2.64) models, the oldest respondents (80) had almost three Widowed 2.79 (2.20Á3.52) 1.52 (1.12Á2.07) times the risk of having poor QoL as the youngest respondents (50Á59 years). An education gradient was Proportion aged 50 years and over in the same household also observed whereby individuals with no education or B25% 0.96 (0.76Á1.20) 1.03 (0.80Á1.34) less than 6 years of education were more likely to report 25Á49% 0.96 (0.76Á1.21) 1.01 (0.78Á1.31) poor QoL compared to those with more than 6 years of 50Á74% 0.68 (0.53Á0.88) 0.72 (0.54Á0.96) education. This association was significant in both ]75% (Ref) 1.00 1.00 univariate and adjusted models. Marital status was found Wealth Index to be associated with QoL. Respondents who were in First quintile 0.96 (0.73Á1.26) 1.01 (0.74Á1.37) some kind of partnership were least likely to report poor Second quintile 2.18 (1.61Á2.95) 1.37 (0.98Á1.91) QoL. Separated and widowed respondents had signifi- Third quintile 1.46 (1.10Á1.93) 1.22 (0.90Á1.65) cantly worse QoL than those in partnership. There was Fourth quintile 1.29 (0.98Á1.71) 1.06 (0.78Á1.44) no significant relationship between the proportion of Fifth quintile (Ref) 1.00 1.00 older people living in a household and QoL. Wealth index had an inverted-V relationship with QoL. In adjusted models, respondents in the poorest and least poor quintiles had similar odds of reporting poor QoL health as compared to female counterparts (Adjusted while those in the second quintile had higher odds of odds ratio: 0.69, 95% CI: 0.54Á0.89) and the oldest poor QoL. Only the odds ratio for being in the second respondents were close to six times as likely to report quintile approached statistical significance. poor health as the youngest in adjusted models. Indivi- The results on factors associated with poor self- duals with no formal education were more likely to reported heath state are presented in Table 4. Poor health report poor health compared to those with more than status was associated with gender, age, educational level 6 years of education. Individuals who were never married and marital status among older people. As observed with were almost twice as likely to report poor health status QoL, male respondents were less likely to report poor compared to those who were in partnership while

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 49 Catherine Kyobutungi et al.

Table 4. Factors associated with poor health status and could have been exacerbated by the growth of the HIV/AIDS epidemic. As a result, Kenya is a country still Univariate model Multivariate model in the early stages of the health transition. However, as Variables (OR and 95% CI) (OR and 95% CI) non-communicable diseases gain a foothold in SSA, it is unlikely that the country will follow a uni-directional Site path towards the second and third stages of the health Viwandani 0.38 (0.31Á0.46) 0.50 (0.40Á0.63) transition. While there is paucity of data on the Korogocho 1.00 1.00 magnitude of the non-communicable disease burden in Sex the country, studies show that the prevalence of risk Men 0.36 (0.30Á0.43) 0.67 (0.52Á0.86) factors for these illnesses is increasing (20). Within the Women 1.00 1.00 study setting, there is a high mortality burden from HIV/ AIDS (21) but in the absence of morbidity studies, it is Age group hard to quantify the extent to which the country could be 50Á59 years 1.00 1.00 enduring a dual burden of disease characterised by high 60Á69 years 2.32 (1.86Á2.88) 1.83 (1.43Á2.34) mortality and morbidity from both infectious diseases 70Á79 years 3.06 (2.17Á4.31) 1.73 (1.17Á2.60) and non-communicable diseases as has been suggested. 80 years and over 9.47 (5.20Á17.26) 5.66 (3.00Á10.69) The proportion of older people in the study area is Education level lower than the national estimate (3) and this is due to the No formal 3.27 (2.62Á4.08) 1.50 (1.16Á1.96) fact that more young people in the economically produc- education tive age groups migrate and stay in the city to find work Less than or equal to 1.77 (1.42Á2.20) 1.19 (0.94Á1.52) and economic opportunities. For similar reasons, in 6 years all age groups except the population under 15 years, the More than 6 years 1.00 1.00 number of males is more than double that of females in the study area. Since migrants into the NUHDSS Marital status constitute a very large proportion of residents, sex In current partnership 1.00 1.00 differences are even greater at older ages since older (Ref) females are less likely to migrate and historically, more Never married 2.86 (1.79Á4.56) 1.88 (1.10Á3.19) males migrated to cities. These reasons partly explain why Separated 1.91 (1.33Á2.74) 1.24 (0.82Á1.89) we have a high proportion of older people (25%) staying Divorced 2.42 (1.46Á4.01) 1.45 (0.83Á2.53) alone. Other reasons for this observation may include Widowed 3.48 (2.72Á4.43) 1.59 (1.16Á2.18) widowhood especially among females, divorce or separa- Proportion aged 50 years and over in the same household tion or split households where other family members are B25% 1.09 (0.86Á1.37) 1.10 (0.80Á1.43) left in rural areas while the older person works in the city 25Á49% 1.08 (0.85Á1.36) 1.11 (0.85Á1.46) (22). The study area has a sex and age distribution which 50Á74% 0.92 (0.71Á1.18) 0.97 (0.72Á1.29) is unlike the national one but is similar to the distribution ]75% 1.00 1.00 for Nairobi city (Fig. 1). The population pyramids in Fig. 1 (a) and (b) both show a predominance of the 20Á29 Wealth index year age groups among males and females and significant First quintile 0.78 (0.60Á1.03) 1.02 (0.75Á1.40) narrowing of the pyramid after the age of 50 years which Second quintile 1.78 (1.32Á2.40) 1.12 (0.80Á1.57) is more pronounced among females. Third quintile 1.31 (1.00Á1.73) 1.05 (0.77Á1.42) The sex and age distribution is also different between Fourth quintile 1.16 (0.88Á1.52) 0.88 (0.65Á1.19) the two slums because Viwandani slum, being near the Fifth quintile 1.00 1.00 industrial area, is mostly inhabited by migrant male labourers seeking job opportunities in the surrounding widowed individuals were 1.6 times more likely. The industries. Older people who are less likely to find wealth index and proportion of people aged 50 years employment in the industries are therefore less likely to and over in the household were not significantly asso- reside in Viwandani and prefer Korogocho and other ciated with reported health status. slums where they are mostly engaged in informal businesses. Discussion Qualitative research in the Nairobi slums where the Kenya, like many SSA countries, has been hard hit by the study was conducted shows that older people play several HIV/AIDS epidemic. During the 1980s, Kenya’s child important roles in society. They are considered fair mortality declined steadily until the 1990s, when a arbitrators in disputes within families and in the com- reversal in the trend was observed (19). The reversal in munity. They are also considered to have a wealth of childhood mortality coincided with an economic crisis experience and wisdom and hence their advice is sought

50 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 The health and well-being of older people in Nairobi’s slums

b (a) Study site: Korogocho and Viwandani, 2002a (b) Nairobi City, 1999

80+ 80+

70–74 70–74 Males Females

60–64 Males Females 60–64

50–54 50–54

40–44 40–44

30–34 30–34

20–24 20–24

10–14 10–14

0–4 0–4 10.0 8.0 6.0 4.0 2.0 0.0 2.0 4.0 6.0 8.0 10.0 10.0 8.0 6.0 4.0 2.0 0.0 2.0 4.0 6.0 8.0 10.0 Percentage Percentage

(c) National population pyramid for Kenya, 1999b

80+

70–74 Males Females

60–64

50–54

40–44

30–34

20–24

10–14

0–4

10.0 8.0 6.0 4.0 2.0 0.0 2.0 4.0 6.0 8.0 10.0 a Percentage Source: APHRC NUHDSS data.

bSource: Ref. (22).

Fig. 1. Population pyramids for the study area, Nairobi City and the whole of Kenya on various issues. Older people are also perceived as living arrangements. Almost 25% of the respondents live important in community development initiatives where alone and are therefore more likely to be deprived of they provide leadership and counsel though they are also social support structures. The HIV/AIDS epidemic in perceived by some as gatekeepers and impediments to SSA has also led to an increased number of orphans, development. During community crises, they play a most of whom are cared for by grandparents who are leading role in mobilising the community (22). These likely to be older people (23). In the study area, 19.5% roles are in addition to more traditional roles of heads of of respondents were looking after children below the household, breadwinners and care givers for grandchil- age of 15 years. Out of these 1,019 children, 770 were dren. However, older people are also more vulnerable either orphans or their parents’ whereabouts were un- in these settings due to altered family structures and known.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 51 Catherine Kyobutungi et al.

Older people in many parts of SSA have been engaged The association between wealth index and QoL is an in efforts to mitigate the effects of HIV/AIDS due to the inverted V-shape but this variable had no significant increased mortality of people in the reproductive and association with reported health status. This could be more economically productive age groups. The high HIV/ explained by the lower response rates among the poorest AIDS and tuberculosis burden in the study area (20) wealth quintiles compared to other quintiles. On the other means that the chronic ill-health associated with these hand, in an environment with high levels of deprivation, it conditions has led to role reversal whereby older people is possible that differences in wealth are marginal in real are providing care to their ill and dying family members. terms and have no tangible impact on health outcomes. About 7% of respondents were caring for someone with a Self-reported measures of health status have not been prolonged illness at the time of the interview while another widely used in SSA in general nor in Kenya in particular. 6% had cared for someone in the past 3 years. Such Their validity as a measure of health has therefore not responsibilities, coupled with economic adversity, may been established, but the finding of steep age and negatively affect the health and well-being of older people. education gradients with worse female health scores With respect to the specific findings, in both univariate point to a good degree of internal validity. and multivariate analysis for the measure of self-reported It is known that the validity of self-reported measures health, females have worse outcomes than males at all age of health and their reliability are influenced by underlying groups; these deteriorate, as expected, with age. Older socio-cultural factors including basic and health literacy, female disadvantage in health status has been described cultural perceptions of illness, disability and health status in industrialised country settings (24Á26), and so our among others (28, 29). Further studies including vign- findings add to the body of evidence supporting this ettes should investigate the influence of such factors on association. the validity of self-reported health in this population. On Korogocho respondents have significantly worse health the other hand, the longitudinal framework offered by outcomes than Viwandani residents. Other studies in the demographic surveillance sites offers a unique opportu- NUHDSS have shown similar findings in other age nity to validate these measures by assessing their perfor- groups but it is unclear what the underlying reasons are mance against objective measures of health and in since both slums have poor environmental sanitation and predicting mortality. poor access to social services. Viwandani is however The absence of similar studies in the country and in the inhabited by mostly labour migrants seeking employment region makes it hard to interpret some of the findings. in the nearby industrial area and hence there are more However, comparison with findings from other HDSS employment opportunities. In addition, a larger propor- sites may shed more light. Other important research tion of residents in Viwandani stay for short periods and questions include the coping strategies and factors then move on compared to Korogocho. It is possible that associated with resilience and healthy ageing among older residents do not stay long enough to be exposed to the people in resource-deprived settings as well as coping hazardous slum environment or that, in the Viwandani strategies in the absence of strong contributory national cash-based economy, economically unsuccessful migrants, social security funds. who could potentially have worse outcomes, move else- The study adds to the limited body of literature where and leave behind the more successful ones. This is regarding health and well-being of older people in SSA apparent in the characteristics of non-respondents, who and especially those in urban informal settlements. are more likely to be from Viwandani and also more likely to be in the poorest wealth quintile. A migrant Further studies are needed to validate the methods used tracking study that assesses reasons for migration out of for assessing health status and to provide comparisons on the slums and post-migration economic and health status, which the health of the older urban poor can be judged. while logistically extremely challenging, would be helpful in clarifying these issues. As expected, a clear age gradient is observed for both measures of health status; however the gradient is steeper Acknowledgements for the self-reported health status than for QoL. Marital This research uses data partly collected under the Urbanisation, status has a significant effect on health outcomes though Poverty and Health Dynamics (UPHD) Research Programme in the the pattern of the effect differs for the two health outcomes. Nairobi Urban Health and Demographic Surveillance System In both cases, married respondents or those in partnership (NUHDSS). We are also grateful to the WHO-SAGE group for have better health outcomes than other respondents. The availing the SAGE instrument which was used in data collection and relationship between being married and well-being has for their support in the analysis and interpretation of the data. We also wish to acknowledge the contribution of the APHRC’s long been established (26, 27), albeit with other health dedicated field and data management teams, and the residents of outcomes, as has the association between poor health Korogocho and Viwandani for their continued participation in the outcomes and widowhood and never married status. NUHDSS.

52 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 The health and well-being of older people in Nairobi’s slums

Conflict of interest and funding 14. Madise N, Banda E, Benaya K. Infant mortality in Zambia: The UPHD Research Programme is funded by the socioeconomic and demographic correlates. Soc Biol 2003; 50: 148Á66. Wellcome Trust UK (grant number GR078530AIA). 15. Kollapan J. The rights of older people Á African perspectives. Work in the NUHDSS has been supported by grants Available from: http://www.globalaging.org/elderrights/world/ from the William and Flora Hewlett Foundation and the 2008/africa.pdf [cited 3 March 2009]. Rockefeller Foundation. We acknowledge funding from 16. Ministry of Health Á Kenya. Reversing the trends: The Second the National Institutes of Health which enabled us to National Health Sector Strategic Plan-NHSSP-II, 2005-2010. Nairobi, Kenya: Ministry of Health; 2005. collect the data on health status assessment. 17. Kowal P, Kahn K, Ng N, Naidoo N, Abdullah S, Bawah A, et al. Ageing and adult health status in eight low-income countries: the INDEPTH WHO-SAGE collaboration. Global References Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302. 18. Chang C, Reeve B. Item response theory and its applications to patient-reported outcomes measurement. Eval Health Prof 1. United Nations Department of Economic and Social Affairs/ 2005; 28: 264Á82. Population Division. World population ageing 2007. New York: 19. Central Bureau of Statistics, Ministry of Health, Macro R. United Nations Department of Economic and Social Affairs/ Kenya Demographic and Health Survey 2003. Calverton, MD: Population Division; 2007. CBS, MoH and ORC Macro; 2004. 2. United Nations Department of Economic and Social Affairs/ 20. Kyobutungi C, Ziraba AK, Ezeh A, Ye Y. The burden of disease Population Division. Population ageing and development. profile of residents of Nairobi’s slums: results from a Demo- Available from: http://www.un.org/esa/population/publications/ graphic Surveillance System. Popul Health Metr 2008; 6: doi: ageing/ageing2009.htm [cited 24 June 2010]. 10.1186/1478-7954-1186-1181. 3. United Nations Department of Economic and Social Affair / 21. de Laat J, Archambault C. Child well-being, social amenities, Population Division. World population prospects: the 2008 and imperfect information: shedding light on family migration revision. New York: United Nations Department of Economic to urban slums. New York: PAA; 2007. and Social Affair/Population Division; 2008. 22. Ministry of Finance and Planning, Central Bureau of Statistics. 4. National AIDS and STD Control Program MoHK. Kenya Kenya Population Census 1999. Nairobi: Central Bureau of AIDS Indicator Survey 2007: preliminary report. Nairobi: Statistics; 2001. Ministry of Health; 2008. 23. Arber S, Cooper H. Gender differences in health in later life: the 5. United Nations Department of Economic and Social Affairs/ new paradox? Soc Sci Med 1999; 48: 61Á76. Population Division. World urbanization prospects: the 2007 24. Lahelma E, Martikainen P, Rahkonen O, Silventoinen K. revision. New York: United Nations Department of Economic Gender differences in ill health in Finland: patterns, magnitude and Social Affairs/Population Division; 2008. and change. Soc Sci Med 1999; 48: 7Á19. 6. United Nations Human Settlements Programme (UN-HABITAT). 25. Matthews S, Hertzman C, Ostry A, Power C. Gender, work roles The challenge of slums: global report on human settlements. and psychosocial work characteristics as determinants of health. London and Sterling, VA: Earthscan Publications Ltd.; 2003. Soc Sci Med 1998; 46: 1417Á24. 7. African Population and Health Research Centre. Population 26. Coombs RH. Marital status and personal well-being: a litera- and health dynamics in Nairobi informal settlements. Nairobi: ture review. Fam Relat 1991; 40: 97Á102. African Population and Health Research Centre; 2002. 27. Johnson NJ, Backlund E, Sorlie PD, Loveless CA. Marital 8. Lamba D. The forgotten half; environmental health in Nairobi’s status and mortality: the national longitudinal mortality study. poverty areas. Environ Urban 1994; 6: 164Á8. Ann Epidemiol 2000; 10: 224Á38. 9. Todaro M. Urbanization and rural-urban migration: theory and 28. Carr AJ, Gibson B, Robinson PG. Measuring quality of life: is policy. Economic development in the Third World. New York: quality of life determined by expectations or experience? BMJ Longman; 1989, pp. 263Á89. 2001; 322: 1240Á3. 10. Brockerhoff M, Brennan E. The poverty of cities in developing 29. Carr AJ, Higginson IJ. Are quality of life measures patient countries. Popul Dev Rev 1998; 24: 75Á114. centred? BMJ 2001; 322: 1357Á60. 11. Madise N, Diamond I. Determinants of infant mortality in Malawi: an analysis to control for death clustering within families. J Biosoc Sci 1995; 27: 95Á106. *Catherine Kyobutungi 12. Magadi M, Zulu E, Brockerhoff M. The inequality of maternal African Population & Health Research Center health care in urban sub-Saharan Africa in the 1990s. Popul Longonot Road, Upper Hill Stud (Camb) 2003; 57: 347Á66. P.O. Box 10787, GPO 00100, Nairobi, Kenya 13. Taffa N. A comparison of pregnancy and child health outcomes Tel: 254 20 2720400 between teenage and adult mothers in the slums of Nairobi, Fax: 254 20 2720380 Kenya. Int J Adolesc Med Health 2003; 15: 321Á9. Email: [email protected]

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2138 53 æINDEPTH WHO-SAGE Supplement Self-reported health and functional limitations among older people in the Kassena-Nankana District, Ghana Cornelius Debpuur1,2*, Paul Welaga1,2, George Wak1,2 and Abraham Hodgson1,2

1Navrongo Health and Demographic Surveillance System, Navrongo, Ghana; 2INDEPTH Network, Accra, Ghana

Background: Ghana is experiencing significant increases in its ageing population, yet research on the health and quality of life of older people is limited. Lack of data on the health and well-being of older people in the country makes it difficult to monitor trends in the health status of adults and the impact of social policies on their health and welfare. Research on ageing is urgently required to provide essential data for policy formulation and programme implementation. Objective: To describe the health status and identify factors associated with self-rated health (SRH) among older adults in a rural community in northern Ghana. Methods: The data come from a survey on Adult Health and Ageing in the Kassena-Nankana District involving 4,584 people aged 50 and over. Survey participants answered questions pertaining to their health status, including self-rated overall health, perceptions of well-being and quality of life, and self-reported assessment of functioning on a range of different health domains. Socio-demographic information such as age, sex, marital status and education were obtained from a demographic surveillance database. Results: The majority of older people rated their health status as good, with the oldest old reporting poorer health. Multivariate regression analysis showed that functional ability and sex are significant factors in SRH status. Adults with higher levels of functional limitations were much more likely to rate their health as being poorer compared with those having lower disabilities. Household wealth was significantly associated with SRH, with wealthier adults more likely to rate their health as good. Conclusion: The depreciation in health and daily functioning with increasing age is likely to increase people’s demand for health care and other services as they grow older. There is a need for regular monitoring of the health status of older people to provide public health agencies with the data they need to assess, protect and promote the health and well-being of older people. Keywords: self-reported health status; functional limitations; older people; INDEPTH WHO-SAGE; adult health; Kassena- Nankana District; Ghana

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘‘SAGE data’’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 27 November 2009; Revised: 4 June 2010; Accepted: 8 July 2010; Published: 27 September 2010

lthough population ageing is often associated population in the developing world. It is expected that the with industrialised societies such as Europe, proportion of older people will grow rapidly in many AAmerica and Japan, the phenomenon is gradu- parts of the developing world, including sub-Saharan ally gaining attention in the developing world. Advances Africa (1, 2). The rapid growth of the aged population in public health and the associated improvement in life poses various challenges. Chronic diseases and disability expectancy has increased the proportion of the aged are disproportionately high among older people. Thus, a

Global Health Action 2010. # 2010 Cornelius Debpuur et al. This is an Open Access article distributed under the terms of the Creative Commons 54 Attribution-Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 Self-reported health and functional limitations among older people in Ghana

growing elderly population will increase the demand for of the country and shares a boundary with Burkina Faso health care and other social services. Due to their low to the north. Since 1993, the Navrongo Health Research economic development, inadequate health infrastructure Centre (NHRC) has been operating a demographic and limited social security programmes, meeting the surveillance system in this area. The district lies between needs of older people in the developing world and latitudes 10.5 and 11.08 N and longitudes 1.0 and 1.58 W especially in sub-Saharan Africa is and will be difficult. (4). The land is relatively flat and covers an area of 1,675 This is likely to be compounded by the erosion of the km2, with altitude of between 200 and 400 m above mean traditional family support systems for older people. sea level. Located in the Guinea savanna belt, the ecology Policies and programmes that address the health and of the study area is typically Sahelian, with a short rainy other needs of the growing aged population are urgently season from April to September and a prolonged dry needed to ensure successful ageing and functional in- season from October to March. The mean annual rainfall dependence of the aged. is about 1,300 mm, with the heaviest rains occurring in However, health research in developing countries August. Monthly temperatures range from 20 to 408C, (including Ghana) has been and continues to be heavily with the mean annual minimum and maximum being 22.8 focused on younger population groups. As such, the and 34.48C, respectively. extent of ageing, the health needs of the ageing popula- Data from demographic surveillance estimated the tion, as well as the implications of national policies for population of the district as at end of June 2007 to be the health and welfare of the aged are poorly understood 147,536 with females constituting 53%, giving a M:F and yet to be well appreciated. Questions about changing ratio of 0.89. About 38% of the population is under health over the life course and compression of morbidity 15 years old, while those aged 65 and over constitute demand an empirical basis for analysis, particularly in the 4.7%. This gives a dependency ratio of 74.5%. The district context of planning and preparing social protection is largely rural with dispersed settlements. There are two mechanisms (health and pension systems) to meet the main ethnic groups Á the Kassenas and the Nankanas Á demands of this growing population group. with other ethnic groups forming about 5% of the The ongoing World Health Organization’s global Study population. Although mortality and fertility are high, on Adult Health and Ageing (SAGE) provides an there have been declines since the 1990s. For instance, the important platform for generating empirical data on crude death rate declined from 18.7 to 10.4 per 1,000 ageing and health transition for policy formulation and between 1997 and 2007, while the crude birth rate fell programme implementation, especially in sub-Saharan from 29.4 to 26.2 per 1,000 and the total fertility rate Africa. Four African countries Á Ghana, South Africa, from 5.0 to 4.0 during the same period. Tanzania and Kenya Á are participating in this pro- The economy of the district is largely agrarian with gramme of research, and have conducted various surveys about 90% of the population dependent on subsistence on ageing and adult health using comparable instruments. agriculture. Major crops grown are cereals such as millet, This article draws on data from a survey conducted in the maize, sorghum and rice. The Tono irrigation dam as well Kassena-Nankana District of northern Ghana as part of as several dug-out dams in various communities facilitate this global programme of research on ageing. irrigated farming and dry-season gardening. Rearing of We describe the health status of older people based on animals like cattle, goats, sheep and poultry form part of their own reports on various aspects of their health. We the agricultural activities. Due to the dependence on then examine factors associated with self-rated health agriculture and declining agricultural yields, poverty is (SRH) among older people. In particular, we examine endemic in the area. The district has a poor road network whether perceived disability in various activities of living and transportation in many parts is limited to bicycles influences rating of one’s health status. The social, and occasional vehicles. Typically, movement within ecological and economic circumstances of the district communities is by foot and use of bicycles. Recently, are more representative of the northern ecological zone of however, there has been an increase in the use of motor Ghana as well as other Sahelian populations to the north bikes, especially in the urban part of the district. of Ghana than of the southern and coastal zones of the Health facilities in the district include one hospital country (3). The results of this study therefore have (located in Navrongo), six health centres, three clinics and relevance for our understanding of the health of older several chemist’s shops. In addition to these static health people in Ghana and beyond. facilities, community health officers have been deployed to several communities to offer door-to-door services to Methods the people. As part of recent efforts to promote access to basic health services, a national health insurance scheme The setting has been instituted and district mutual health insurance The Kassena-Nankana District1 (KND) in the Upper schemes are operational in all districts of the country. East region of Ghana is located at the northern-most part The main causes of morbidity in the study area are

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 55 Cornelius Debpuur et al.

malaria, gastroenteritis and acute respiratory infection. Status Descriptions included overall rating of health, Periodic outbreaks of epidemic meningococcal meningitis questions on eight domains of health (mobility, self-care, have been recorded in the district. Service provision data pain and discomfort, cognition, interpersonal activities, suggest an increasing prevalence of hypertension and sleep/energy, affect and vision), as well as functional diabetes, and there is need for more systematic docu- assessment questions. Vignettes for health status descrip- mentation of the type and prevalence of non-communic- tions were included in the Full SAGE survey but not in able diseases among adults. The Adult Health and the Summary version. Under the Subjective Well-being Ageing study being implemented in the district as part and Quality of Life section, respondents were asked of the INDEPTH WHOÁSAGE initiative will contribute questions on their thoughts about their life situation. towards highlighting the health situation of adults and Almost all the questions in the questionnaire had 5-point inform health care delivery in the district. The survey scale response categories. Background information on reported here is the first district-wide population-based age, education, marital status of each respondent as well survey of adults to collect information on self-reported as household information were obtained from the routine health status among persons aged 50 and over, and thus HDSS data. provides baseline data for monitoring and evaluating Standardised self-reported surveys of health have adult health. contributed immensely to the understanding of the health status of elderly people in the developed world Data and Asia. However, such studies (particularly those The data for this study come from the summary version focusing on older people) are rare in sub-Saharan of the INDEPTH WHO-SAGE Adult Health and Africa. The data reported in this article will contribute Ageing Survey implemented by the NHRC. The Adult towards bridging the knowledge gap on the health status Health and Ageing Survey is an INDEPTH Network of older people in sub-Saharan Africa and the develop- multi-site activity in collaboration with the World Health ing world at large. Organization’s Study on global AGEing and Adult Health (SAGE). The survey forms part of efforts by Outcome variables the INDEPTH Network to establish a longitudinal The primary outcome of interest in this study is overall database on older people to inform policies related to SRH status. This is based on respondents’ assessment of their well-being. Ethical approval for the study was their current health status on a 5-point scale in response obtained from the ethics committee of the Ghana Health to the question: ‘In general, how would you rate your Service as well as the institutional review board of the health today?’ Response categories were: very good, NHRC. Community approval was obtained from the good, moderate, bad and very bad. Barely 10% of chiefs and elders. Written consent from individuals was respondents rated their health as very good and few obtained before interview. rated their health either as ‘bad’ (4.8%) or ‘very bad’ The summary version of the SAGE study primarily (0.2%). Almost half (49.4%) reported their health as targeted older people (50 and over), although smaller ‘good’, while 36.6% rated their health as moderate. samples of adults 18Á49 years were also included. A From this we created a dichotomous measure coded 0 if single-stage simple random sample of 6,074 older people response was ‘very good’ or ‘good’ and 1 if response was (50 years and over) and 1,360 younger adults (18Á ‘moderate’, ‘bad’ or ‘very bad’. This simple measure of 49 years) in the Kassena-Nankana District was drawn health status has been used in population-based epide- using the Health and Demographic Surveillance System miological research, and has been identified as a powerful (HDSS) database as a sampling frame. The data collec- predictor of morbidity and mortality (5Á7). In dichot- tion was integrated into the routine HDSS data collection omising SRH in our analysis, we follow the lead round that took place between January and April 2007. of previous researchers who adopted a similar approach Trained HDSS interviewers visited households and con- (5,7Á9) and the observation by Manor et al. (10) that such ducted face-to-face interviews with selected individuals. dichotomisation does not make any difference. The questionnaire was written in English although Other indicators of health status examined in this study the interviews were conducted in the local languages are overall health status and self-reported functional of respondents. Translation of the questions in to Kassim limitations. The overall health status of individuals was and Nankam Á the two principal languages in the dis- assessed based on responses to questions in eight trict Á (and back translation from the local languages into domains of health covering affect, cognition, interperso- English) as well as pre-testing of the questionnaire was nal activities and relationships, mobility, pain, self-care, done as part of interviewer training. sleep/energy, and vision. At least two questions were The questions asked in the survey were grouped under asked in each domain, thus providing more robust two sections Á Health Status Descriptions, and Subjective assessments of individual health levels and reducing Well-being and Quality of Life. Items under Health measurement error for any single self-reported item. An

56 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 Self-reported health and functional limitations among older people in Ghana

overall health status score (HSS) for each respondent was Analysis derived from responses to these various items using Item The analysis is in two parts. First, we describe the health Response Theory (IRT) parameter estimates in Winsteps, status of older people based on three indicators: overall a Rasch measurement software package (http://www.win SRH, an index of self-reported functional ability (WHO- steps.com). The health score is then transformed to a DASi) and an overall HSS. In the second part of the scale of 0Á100 (where 0 represents the worst health and analysis, we explore factors related to poor SRH using 100 the best health status). logistic regression. In this analysis we are particularly Based on self-reports of difficulty in carrying out interested in the influence of reported functional ability various activities contained in the health status descrip- (WHODASi) on self-related health status. Functional tions section of the questionnaire, an index of overall ability is an important dimension of health and an disability (WHO Disability Assessment Scale Á WHO individual’s assessment of ability to perform basic daily DAS) was constructed. Self-reported functioning was activities is likely to influence SRH. However, the assessed through the standardised 12-item WHODAS, magnitude of the influence of functional limitations on Version 2 (11). On a 5-point scale, respondents rated their SRH may be mitigated by factors such as the cause and level of difficulty in carrying out various activities. These duration of disability, awareness of co-morbidity and responses were used to create a score of overall disability; access to assistive devices. Generally, we expect that adults the score was then transformed to a scale ranging from 0 with greater functional disability will rate their health (no disability) to 100 (greatest disability). In effect, poorer than those with lower disability. We controlled for WHODAS is an overall summary of one’s perceived confounders such as age, ever attended school, marital difficulties in carrying out various functions of daily status, relationship to household head, socioeconomic living. A higher score indicates greater perceived diffi- status and proportion of household members aged 50 or culty in carrying out daily functions, while a lower score over. These factors have been identified as significant indicates lower perceived difficulty in functioning. In factors in self-reported health, as have age and gender order to make this score conceptually consistent with the differences (5, 12). Similarly, marital status, education, HSS, it was inverted to a score designated here as socioeconomic status and social support have been WHODASi, so that a higher score (on a 0Á100 scale) identified as relevant factors in health status (13). We represents better functioning. In the analyses we grouped include relationship to household head and proportion of WHODASi into quintiles to represent levels of functional older people in household as crude indicators of social ability. support.

Socio-demographic variables Results Socio-demographic information on respondents was Although a sample of adults aged 18Á49 years were obtained from routine demographic surveillance data interviewed using the summary version of the SAGE including sex, age, education, marital status, relationship Adult Health Survey, our analysis in this article is limited to head of household, number of older people in the to older participants in the survey. Of the 6,074 older household and household economic status. Age was people targeted for survey, 4,584 were successfully inter- categorised into three subgroups: 50Á59, 60Á69 and viewed (a response rate of 75.5%). A major reason for 70. Marital status was categorised as married or non-participation was the inability of the interviewers to unmarried. Educational status was categorised as never meet the targeted respondent after at least three visits to attended school or ever attended school. In the analysis the household. Other reasons include migration, death those who have never attended school are referred to as and inaccurate information. In Table 1 we compare having no formal education, while those who have ever respondents and non-respondents in terms of back- attended school are described as having some formal ground characteristics (sex, age, education, marital status, education. The socioeconomic status of households was relationship to household head, socioeconomic quintile assessed in terms of wealth quintiles based on possessions of household, average household size and proportion of and housing characteristics. The five quintiles represent household members aged 50 years and over). The data poorest, poorer, poor, less poor and least poor house- indicate that compared with respondents, non-respon- holds. In terms of relationship to the head of household, dents were largely male, slightly younger, unmarried, respondents were described as head, spouse of head, more educated and from relatively less poor households. parent of head or other relation to head of household. These are likely to be more active and mobile and hence The number of older people in the household was are more likely to be away from home during the survey. expressed as a proportion of the total number of people To the extent that our respondents are not representative in the household and grouped into quartiles for the of the older population of the district, our results may analysis. have limited generalisability.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 57 Cornelius Debpuur et al.

Table 1. Background characteristics of 4,584 adult respon- The percentage of older people reporting poor health dents and 1,437 non-respondents aged 50 and over in increased with age among both men and women. How- northern Ghana ever, the greatest differentials in SRH were observed in terms of levels of functional disability. The proportion of Respondents Non-respondents respondents reporting poor health was substantially Variables (n4,584) (n1,437) higher among those also reporting low functional ability, both in men and women. Whereas less than one in five Sex (%) participants in the highest category of functional ability Men 39.0 44.9 reported poor health, more than three in four of those in Women 61.0 55.1 Mean age (SD) 62.5 (9.1) 61.4 (9.0) the lowest category of functional ability reported poor

Age group (%) Table 2. Proportions reporting poor self-rated health among 50Á59 years 43.0 50.2 4,584 adults aged 50 and over in northern Ghana 60Á69 years 35.9 30.8 70Á79 years 16.6 14.8 Variables Men (%) Women (%) All (%) 80 years and over 4.5 4.2 Sex Education level (%) Men 35.2 No formal education 90.7 85.2 Women 45.2 Less than or equal to 3.9 3.5 6 years Age group (years) More than 6 years 5.4 11.5 50Á59 26.7 37.7 33.5 60Á69 33.7 48.8 43.4 Marital status (%) 70 years and over 50.7 58.8 54.9 Now single 46.3 50.5 In current partnership 53.7 49.5 Education level No formal education 37.1 46.3 42.9 Socioeconomic quintile (%) Some formal education 25.5 36.4 29.8 First quintile 27.5 23.5 Second quintile 24.4 18.8 Marital status Third quintile 21.9 20.4 Now single 41.1 48.5 47.3 Fourth quintile 18.7 21.5 In current partnership 33.9 40.6 36.6 Fifth quintile 7.4 15.7 Relationship to household head Relationship to household head (%) Head 34.7 44.9 38.2 Head 51.0 52.3 Spouse 14.8 41.5 40.8 Spouse 21.1 15.4 Parent 38.1 52.2 51.3 Parent 13.8 10.9 Other relation 43.5 47.0 46.1 Other relation 14.2 21.4 Proportion of household members aged 50 and over (%) Mean number of house- 6.6 (4.6) 6.2 (5.1) B25 32.9 45.4 40.3 hold members (SD) 25Á49 37.2 44.3 41.4 50Á74 32.6 44.4 40.4 Mean proportion of 0.4 (0.2) 0.4 (0.3) 575 46.4 53.2 50.7 household members aged 50 and over (SD) Socioeconomic quintile Poorest quintile 36.7 45.7 41.8 Second quintile 36.8 52.0 45.7 The majority of respondents (61%) were female, and Third quintile 37.6 46.3 43.1 the average age was 62.5 years, with men (average age of Fourth quintile 30.1 41.3 37.5 63.7 years) being older than the women (average age of Least poor quintile 29.1 34.8 32.7 61.7 years). Nearly three-quarters of the respondents WHODASi quintile were aged below 70, and less than 10% had ever attended Highest ability quintile 13.8 20.0 16.8 school. About half of the respondents were married, Second quintile 23.3 28.3 26.4 while a similar proportion were heads of households. Third quintile 35.4 44.1 40.8 Overall, less than half of the respondents (42%) rated Fourth quintile 54.9 57.6 56.7 their overall health as poor, with slightly more women Lowest ability quintile 75.6 77.6 76.9 (45%) than men (35%) reporting poor health (Table 2).

58 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 Self-reported health and functional limitations among older people in Ghana

health (Fig. 1). Other differentials were also observed in These three indicators measure different dimensions of terms of education, marital status and household socio- health, and although SRH, WHODASi and HSS are economic status. The proportion of household popula- related, none is completely determined by the others. tion aged 50 and over is included to indicate social SRH and WHODASi are positively related with correla- support within the household. Households with more tion of 0.49, while SRH and HSS are similarly correla- than half of members aged 50 or over have a greater ted (0.50). The highest correlation is found between proportion of elderly dependents and possibly less social WHODASi and HSS (0.84). support, hence SRH in such households could be poorer On the basis of SRH, WHODASi and HSS, reported than in less-dependent households. health status declined with age and was slightly worse Table 3 shows results of WHODASi and HSS for older among women than men. We explored the association men and women in the Kassena-Nankana District by age between functional disability and SRH among older category. A higher WHODASi score indicates a higher people while controlling for selected socio-demographic level of functional ability compared to a lower score. The factors such as sex, age, education, marital status, mean WHODASi for the sample is 70.9 (73.7 for men and relationship to head of household, proportion of people 69.1 for women), with the mean score decreasing with age aged 50 and over in the household and socioeconomic such that the oldest respondents had lowest functional quintile of the household. ability. Thus, the reported level of functional ability Table 4 presents logistic regression results with poor decreased with age. This pattern was observed among SRH as the outcome variable. In the univariate model both men and women (Fig. 2). Generally, the WHODASi most factors (except proportion of household members aged 50 and over, and household socioeconomic status) score was lower among women compared to men of had a significant association with SRH status. Individuals comparable age. The ageÁsex pattern in functional limita- with lower functional ability levels were more likely to tions is evident in the proportion of older people whose report poor health than their colleagues with better WHODASi scores were below the median for the overall functional ability. Men appeared less likely to report sample. Higher proportions of participants in the older poor health than women. Other researchers have sug- age groups had scores below the median than their gested that women’s poorer rating of their health may be younger counterparts. Similarly, more women in each age indicative of greater sensitivity to health conditions rather group had WHODASi scores below the median (72.2) than a female health disadvantage (7). The oldest old compared to men. were much more likely to rate their health poorly than For the overall HSS a higher score indicates better those 50Á59 years old. Similarly those with no education health than a lower score. The mean HSS score for the were more likely to report poor health than those with sample was 64.0 with men scoring slightly higher (65.8) some education; being single was associated with reports than women (62.8) as shown in Table 3. In terms of age, of poor health. In terms of relationship to the household younger age groups tended to report better health (higher head, those who were parents of or otherwise related to mean HSS) than their older counterparts; while more the head appeared more likely to report poor health women than men in each age group reported HSS below compared to the household heads themselves. the median (63.5). In the multivariate model, the effects of WHODASi remained significant, with respondents in the higher disability quintiles much more likely to report poor 80 health status than those in the lowest disability quintile. In other words, adults with greater functional limitations 70 were more likely to rate their health as poor compared to those with less functional limitations. The other factors 60 that had significant effect on SRH were sex and house- hold wealth quintile. Women were more likely than men

50 to rate their health as poor, while older people in the two higher wealth quintiles were less likely to rate their health as poor compared to their counterparts in the least Mean WHODASi functional ability score 40 wealthy quintile. The effects of age were barely significant 50–54 55–59 60–64 65–69 70–74 75–79 80 + Age group after allowing for WHODASi, although older adults

Poor self-rated health Good self-rated health appeared more likely to report poor health than their younger colleagues. Fig. 1. Mean WHODASi functional ability score, by age These results suggest that functional disability is the group and self-rated health, among 4,584 adults aged 50 and primary factor associated with overall SRH among older over in northern Ghana. people in the Kassena-Nankana District. The influence

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 59 Cornelius Debpuur et al.

Table 3. Distribution of WHODASi functional ability score and health status score by age and sex among 4,584 adults aged 50 and over in northern Ghana

Variables Men (n1,789) Women (n2,795) All (n4,854)

Mean WHODASi score (SD) 50Á59 years 79.7 (15.7) 74.8 (15.0) 76.6 (15.4) 60Á69 years 74.5 (16.9) 67.6 (16.8) 70.1 (17.2) 70 years and over 63.6 (20.9) 58.1 (18.7) 60.7 (20.0) All ages 73.7 (18.7) 69.1 (18.7) 70.9 (18.1)

Proportion of respondents with WHODASi less than median 50Á59 years 29.4 40.1 36.2 60Á69 years 41.8 57.7 51.9 70 years and over 63.7 75.2 69.7 All ages 42.5 53.0 48.9

Mean health status score (SD) 50Á59 years 68.4 (9.4) 65.2 (7.4) 66.4 (8.3) 60Á69 years 65.9 (8.7) 62.1 (7.4) 63.5 (8.1) 70 years and over 61.7 (9.3) 58.3 (7.3) 59.9 (8.5) All ages 65.8 (9.6) 62.8 (7.8) 64.0 (8.7)

Proportion with health status score less than median 50Á59 years 27.2 39.1 34.8 60Á69 years 39.6 57.5 51.0 70 years and over 59.9 77.2 68.8 All ages 39.9 52.8 47.8 of functional limitations on SRH observed in this study is other studies, it appears that these factors are not consistent with findings from other studies (9, 14, 15). important determinants of SRH in this population. Other significant determinants of SRH were sex and household wealth quintile. Although age, education and Discussion marital status appeared to be significant in the univariate Data on adult health status, particularly the health of analysis, their significance eroded when other variables older people in sub-Saharan Africa, are required to were controlled for in the multivariate analysis. Unlike monitor trends in the health status of adults and the extent to which social and health policies impact on older people. One relatively easy way of generating such data is through population surveys of self-reported health. The

80 implementation of such surveys has contributed immen- sely to the understanding of ageing and transitions in health with age in the developed world and Asia. 70 Although self-reported health is subjective, it has been found to be a good predictor of future health care use and 60 mortality. In 2007, the NHRC conducted a survey on ageing and adult health in the Kassena-Nankana District

50 of Ghana as part of the INDEPTH WHO-SAGE Adult Health Study. The survey collected information on self- reported health among adults in the district. Data from Mean WHODASi functional ability score 40 this survey have been analysed to describe the health 50–54 55–59 60–64 65–69 70–74 75–79 80+ Age group status as well as identify factors associated with SRH Men Women status among older people in this rural setting. Our results indicate that the majority of older people Fig. 2. Mean WHODASi functional ability score, by age rated their overall health as good. However, women were group and sex, among 4,584 adults aged 50 and over in more likely than men to rate their health as poor. northern Ghana. A similar pattern was observed with regard to reported

60 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 Self-reported health and functional limitations among older people in Ghana

Table 4. Factors associated with poor self-rated health as poor compared to those with lower disability. Multi- among 4,584 adults aged 50 and over in northern Ghana variate regression results showed that levels of functional disability, sex and household wealth quintile had sig- Univariate model Multivariate model nificant influence on SRH status. Variables (OR and [95% CI]) (OR and [95% CI]) The findings in this study are comparable with the results of previous studies in various parts of the WHODASi quintile world. Earlier studies have noted the existence of socio- Highest ability 1.00 1.00 demographic differentials in SRH. Research evidence quintile suggests that men generally report fewer diseases and Second quintile 1.78 [1.43Á2.22]** 1.65 [1.32Á2.07]** fewer limitations in activities of daily living at older ages Third quintile 3.42 [2.72Á4.23]** 3.18 [2.56Á3.96]** than their female counterparts. Women are more likely to Fourth quintile 6.51 [5.18Á8.17]** 5.76 [4.54Á7.32]** rate their health poorer and to report more functional Lowest ability 16.56 [13.1Á20.9]** 14.23 [11.1Á18.3]** limitations and disability than men (16Á20). Irrespective quintile of sex, however, older age is related to higher odds of Sex reporting health problems and various studies have Men 1.00 1.00 observed that older adults tend to rate their health Women 1.54 [1.37Á1.75]** 1.40 [1.38Á1.73]** poorer than their younger colleagues (16, 21, 22). Lower Age group (year) socioeconomic status is associated with worse morbidity, 50Á59 1 1 mortality and self-reported health in older persons (23). 60Á69 1.52 [1.32Á1.74]** 1.12 [0.96Á1.31] Other factors such as marital status, socioeconomic 70 years and over 2.41 [2.06Á2.82]** 1.24 [1.02Á1.51]* status and education are also known to affect health status (24, 25), although marital status and education did Education level not appear significant in our analysis. No formal education 1.00 1.00 Older people in this district face considerable health Some formal 0.56 [0.46Á0.69]** 0.92 [0.71Á1.17] challenges like their colleagues elsewhere. As our results education indicate, there is considerable increase in functional Marital status limitations and poor health with age, with more women Now single 1.56 [1.32Á1.75]** 0.94 [0.78Á1.13] tending to report health problems than their male In current partnership 1 1 counterparts. Unfortunately however, older people in the Kassena-Nankana District do not only have to deal Relationship to household head with functional limitations, but also have to deal with Head 1 1 infectious diseases such as malaria and gastroenteritis. Spouse 1.11 [0.96Á1.30] 0.94 [0.75Á1.19] What is more, they grapple with these health challenges in Parent 1.70 [1.43Á2.03]** 1.00 [0.80Á1.26] a context of inadequate health care and weak social Other relation 1.38 [1.15Á1.66]** 1.06 [0.85Á1.33] support systems. Public policy and health interventions Proportion of household members aged 50 and over (%) that promote healthier lifestyles and improve access to B25 1.00 1.00 health care are required to improve the health and quality 25Á49 1.04 [0.90Á1.20] 1.00 [0.85Á1.18] of life of older people. In spite of increasing urbanisation, 50Á74 1.00 [0.85Á1.18] 0.96 [0.80Á1.17] the majority of Ghana’s older people live in rural areas 575 1.52 [1.23Á1.88]** 1.49 [1.16Á1.92]** where health and social services are inadequate. Educa- tion and information on healthy living need to be made Socioeconomic quintile available to the general population to enhance prevention Poorest quintile 1.00 1.00 and control of chronic conditions. Programmes need to Second quintile 1.17 [0.99Á1.38] 1.12 [0.93Á1.35] focus attention on promoting healthy ageing. Bold policy Third quintile 1.05 [0.88Á1.25] 0.95 [0.78Á1.15] decisions are also needed to integrate ageing and adult Fourth quintile 0.83 [0.70Á1.00] 0.74 [0.60Á0.90]** health issues into all aspects of national planning and Least poor quintile 0.67 [0.52Á0.87]** 0.65 [0.48Á0.88]** development. Some observers have noted that the con- *pB0.05; **pB0.001. cerns of older people remain marginalised in Ghana’s social and economic debates (21). There is the need to marshal evidence on the health situation of older people functional disability and overall health score. Functional in the country and to use this evidence to advocate for disability was higher among women compared to men. programmes and policies to address the health care and Among both men and women, older adults were more other needs of older people. This study has highlighted likely to report functional disability. Adults with higher the situation of older people in one of the rural districts in functional disability were more likely to rate their health Ghana, and it is hoped that this will broaden the evidence

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 61 Cornelius Debpuur et al.

on the health status of older Ghanaians and contribute baseline for monitoring trends in the health status of towards effective policy formulation in the country. older people in the Kassena-Nankana District. Results from the national SAGE study conducted in the country around the same time as our study will provide a Acknowledgements broader national picture on the health status of older people. For the purposes of monitoring the health status The authors would like to thank the people of the Kassena-Nankana of older people, such studies need to be conducted District, especially all the men and women who agreed to be periodically and in a variety of settings. interviewed, for their support and participation in the study. We This initial survey has demonstrated the feasibility of are grateful to the NHRC staff who collected and processed the data conducting population-based health surveys of adults for this study. in rural Ghana. The results of the analyses are generally consistent with other studies and indicate Conflict of interest and funding the scope for monitoring population health using self- This project was supported by a grant from the assessments of health. There is the opportunity for INDEPTH WHO-SAGE study and the INDEPTH follow-up and longitudinal analysis anchored on the Network. The authors would like to acknowledge the HDSS platform existing in the district. Future analyses INDEPTH Network for their financial support. will explore the relationship between SRH and mor- bidity and mortality in this population. The INDEPTH Note WHO-SAGE Adult Health Research platform (of 1. In 2008 the Kassena-Nankana District was split which this study is a part) is uniquely placed to into two districts Á Kassena-Nankana and Kassena- contribute towards an understanding of the relationship Nankana West districts. In this article we use the between SRH and subsequent morbidity and mortality original name of the district to refer to the two in the region. Subsequent analysis of SRH and districts. mortality from INDEPTH Network sites will contri- bute to the literature on this topic, which is currently References under-researched in sub-Saharan Africa. 1. United Nations Department of Economic and Social Affairs (DESA). Population Division. World population prospects: the Conclusion 2006 revision, highlights. United Nations. New York, NY: As in other developing countries, the population of DESA; 2007. older people in Ghana is increasing steadily. Despite the 2. Rosenberg M. Overview of life expectancy. Available from: increasing number of older people in the country, http://geography.about.com/od/populationgeography/a/lifeexpec however, very little is known about their health status, tancy.htm [cited 26 August 2008]. 3. Adongo BP, Phillips JF, Kajihara B, Fayorsey C, Debpuur C, especially for those in rural areas. This lack of knowl- Binka FN. Cultural factors constraining the introduction of edge impedes development and implementation of family planning among the Kassena-Nankana of northern policies and programmes as well as evaluation of the Ghana. Soc Sci Med 1997; 45: 1789Á804. impact of social and health policies on older people. 4. Nyarko P, Wontuo P, Nazzar A, Phillips J, Ngom P, Binka F. In: Ghana is participating in the WHO multi-country Sankoh O, Kahn K, Mwageni E, Ngom P, Nyarko P, eds. Population and health in developing countries. Population, SAGE. The data presented in this study form part of health, and survival at INDEPTH sites, Vol. 1. Ottawa: this global study. Our results suggest that the ageing International Development Research Centre; 2002, pp. 247Á56. process in this district is consistent with what has been 5. Idler EL, Benyamini Y. Self-rated health and mortality: a review observed in other parts of the world. SRH declines with of twenty-seven community studies. J Health Soc Behav 1997; age among both men and women. It appears that with 38: 21Á37. 6. Frankenberg E, Jones NR. Self-rated health and mortality: does increasing age there is a decline in health which is the relationship extend to a low income setting? J Health Soc manifest in increasing functional disability. This depre- Behav 2004; 45: 441Á52. ciation in health and daily functioning increases the 7. Kuhn R, Rahman O, Menken J. Survey measures of health: how demand for health care and other services by older well do self-reported and observed indicators measure health people. Therefore, steps need to be taken to address the and predict mortality? In: Cohen B, Menken J, eds. Committee on population, division of behavioral and social sciences and health care and other needs of older people. Health education. Washington, DC: The National Academies Press; policies and programmes that improve functional capa- 2006, pp. 314Á41. city and well-being for older people are particularly 8. Deeg DJ, Kriegsman DM. Concepts of self-rated health: urgent. There is also the need for regular monitor- specifying the gender difference in mortality risk. Gerontologist ing and assessment of the health status of older people 2003; 43: 376Á86. 9. Murata C, Kondo T, Tamakoshi K, Yatsuya H, Toyoshima H. to provide public health agencies with the data they 2006. Determinants of self-rated health: could health status need to assess, protect, and promote the health and well- explain the association between self-rated health and mortality? being of older people. The present study will serve as a Arch Gerontol Geriatr 2006; 43: 369Á80.

62 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 Self-reported health and functional limitations among older people in Ghana

10. Manor O, Matthews S, Power C. Dichotomous or categorical 18. Knurowski T, Lazic D, van Djik JP, Geckova AM, Tobiasz- response? Analysing self-rated health and lifetime social class. Adamczyk B, van den Heuvel WJA. Survey of health status and Int J Epidemiol 2000; 29: 149Á57. quality of life of the elderly in Poland and Croatia. Croat Med J 11. World Health Organization. Measuring health and disability: 2004; 45: 750Á6. a manual for the World Health Organization Disability 19. Olsen KM, Dahl SA. Health differences between European Assessment Schedule (WHODAS 2.0). Geneva: World Health countries. Soc Sci Med 2007; 64: 1665Á78. Organization; 2009. 20. Zimmer Z. Poverty, wealth inequality, and health among older 12. Zimmer Z, Natividad J, Lin HS, Chayovan N. A cross national adults in rural Cambodia. Soc Sci Med 2008; 66: 57Á71. examination of the determinants of self-assessed health. J 21. Mba CJ. The health condition of older women in Ghana: a case Health Soc Behav 2000; 41: 465Á81. study of Accra City. J Int Women’s Stud [online] 2006; 8: 13. Kuate-Defo B. Interactions between socioeconomic status and 176Á189. Available from: http://www.bridgew.edu/soas/jiws/Nov living arrangements in predicting gender-specific health status 06/index.htm. [cited 16 February 2009]. among the elderly in Cameroon. In: Cohen B, Menken J, eds. 22. Omar MR, Arthur JB. Self-reported health among older Committee on population, division of behavioral and social Bangladeshis: how good a health indicator is it? Gerontologist sciences and education. Washington, DC: The National Acade- 2003; 43: 856Á63. mies Press; 2006, pp. 276Á313. 23. Lawlor DA, Sterne JAC. Socioeconomic inequalities in health. 14. Idler EL. Gender differences in self-rated health, in mortality, BMJ 2007; 334: 963Á4. and in the relationship between the two. Gerontologist 2003; 43: 24. Waite L, Gallagher M. The case for marriage: why married 372Á5. people are happier, healthier, and better off financially. New 15. Damian J, Ruigomez A, Pastor V, Martin-Moreno JM. Deter- York: Doubleday; 2000. minants of self assessed health among Spanish older people 25. Williams K, Umberson D. Marital status, marital transitions, living at home. J Epidemiol Community Health 1999; 53: 412Á6. and health: a gendered life course perspective. J Health Soc 16. Shi J, Liu M, Zhang Q, Lu M, Quan H. Male and female adult Behav 2004; 45: 81Á98. population health status in China: a cross-sectional national survey. BMC Public Health 2008; 8: 277. Available from: http:// www.biomedcentral.com/content/pd/1471-2459-8-277.pdf [cited *Cornelius Debpuur 6 April 2009]. Navrongo Health Research Centre 17. Tajvar M, Arab M, Montazeri A. Determinants of health- P.O. Box 114 related quality of life in elderly in Tehran, Iran. BMC Central Navrongo, UER, Ghana public Health 2008; 8: 323. Available from: http://www.biomed Tel: 233 74222310 central.com/content/pdf/1471-2458-8-323.pdf [cited 6 April Fax: 233 74222320 2009]. Email: [email protected]

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2151 63 INDEPTH WHO-SAGE Supplement Patterns of health status and quality of life among older people in rural Viet Nam Hoang Van Minh1,2*, Peter Byass3#, Nguyen Thi Kim Chuc1,2 and Stig Wall3#

1Faculty of Public Health, Hanoi Medical University, Hanoi, Viet Nam; 2INDEPTH Network, Accra, Ghana; 3Department of Public Health and Clinical Medicine, Umea˚ Centre for Global Health Research, Umea˚ University, Umea˚ , Sweden

Background: To effectively and efficiently respond to the growing health needs of older people, it is critical to have an indepth understanding about their health status, quality of life (QoL) and related factors. This paper, taking advantage of the INDEPTH WHO-SAGE study on global ageing and adult health, aims to describe the pattern of health status and QoL among older adults in a rural community of Viet Nam, and examine their associations with some socio-economic factors. Methods: The study was carried out in the Bavi District, a rural community located 60 km west of Hanoi, the capital, within the Epidemiological Field Laboratory of Bavi (FilaBavi). Face-to-face household interviews were conducted with people aged 50 years and over who lived in the FilaBavi area. The interviews were performed by trained surveyors from FilaBavi using a standard summary version SAGE questionnaire. Both descriptive and analytical statistics were used to examine the patterns of health status and QoL, and associations with socio-economic factors. Results: Higher proportions of women reported both poor health status and poor QoL compared to men. Age was shown to be a factor significantly associated with poor health status and poor QoL. Higher educational level was a significant positive predictor of both health status and QoL among the study subjects. Higher economic status was also associated with both health status and QoL. The respondents whose families included more older people were significantly less likely to have poor QoL. Conclusion: The findings reveal problems of inequality in health status and QoL among older adults in the study setting by sex, age, education and socio-economic status. Given the findings, actions targeted towards improving the health of disadvantaged people (women, older people and lower education and economic status) are needed in this setting. Keywords: older people; health status; quality of life; rural; Viet Nam; INDEPTH WHO-SAGE

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘‘SAGE data’’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 5 March 2010; Revised: 13 May 2010; Accepted: 8 July 2010; Published: 27 September 2010

uring the past few decades, under the forces population who reach middle age and beyond is increas- of a demographic transition characterised by ing sharply (1Á3). Developing countries are currently Ddeclining fertility rates and increasing life ageing much faster than industrialised countries (3, 4). expectancy, the proportion of people in the world In 2002, almost 400 million people aged 60 and over lived in the developing world. By 2025, it may rise to 840 million representing 70% of all older people world- # Deputy Editor, Peter Byass, Chief Editor, Stig Wall, have not wide (2, 3). participated in the review and decision process for this paper.

Global Health Action 2010. # 2010 Hoang Van Minh et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 64 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2124 WHO-SAGE study on older adults in rural community of Viet Nam

Population ageing in low- and middle-income countries and Demographic Surveillance System (HDSS), sup- has special implications for many public services, especially ported by Sida/SAREC, was established in 1999 with a for health care, in these countries. The health care systems sample of around 50,000 individuals from the Bavi of many developing countries are still focused on District. People aged 50 and over accounted for about childhood and infectious diseases as well as reproductive 17% of the total population under surveillance. The health services. But the ageing population leads to surveyed population includes three distinct groups: those increasing demands for care that addresses chronic health in mountainous areas, highlands, and riverside or island conditions (2). Nowadays, in all countries, and in low- and dwellers (10). The FilaBavi HDSS is a member of the middle-income countries in particular, measures to help INDEPTH network (11). older people remain healthy and active are urgently needed (2, 4). To effectively and efficiently respond to the growing Data collection health needs of older populations, it is critical to have an Face-to-face household interviews were planned for all indepth understanding about their health conditions, people aged 50 years and over who lived in the FilaBavi quality of life (QoL) and related socio-economic factors. area between the end of 2006 and the beginning of 2007. Viet Nam, a developing country in South-East Asia, is The interviews were done by trained surveyors from undergoing demographic transition and experiencing FilaBavi using a summary version of the SAGE ques- rapid population ageing. The proportion of people aged tionnaire (available as a Supplementary File to this 50 years and over rose from 12.6% in 2000 to 14.1% in paper). Further details of the study methodology are 2005 and will account for 18.9% of the total population in available separately (9). The questionnaire was translated 2015 (3). In Viet Nam, the number of older people living into the local language and pre-tested before official use. in rural areas is about 3.5 times higher than those living Spot-checks and re-checks on sample data were con- in urban areas (5). About 44% of older Vietnamese are ducted by supervisors for quality control. working, but mostly in agricultural activities which provide low and unstable incomes. Other sources of Measurements income, including pension and social assistance benefits, Outcome variable are significant factors to reduce risks for older people. Self-reported health status and QoL among the study However, the coverage of the current social protection subjects were outcome variables. Health status scores were system in Viet Nam is not adequate (6). calculated based on self-reported health levels in eight Life expectancy in Viet Nam reached 72.2 years in health domains covering: affect, cognition, interpersonal 2005, a relatively high level compared to the nation’s activities and relationships, mobility, pain, self-care, sleep/ economic conditions. However, the average healthy life energy, and vision. Each domain included at least expectancy was far lower, at 58.2 years and ranked 116 two questions. Asking more than one question about among 174 countries in the world (7). Health care for the difficulties in a given domain provides more robust older people in Viet Nam has been improved, but the assessments of individual health levels and reduces mea- accessibility for vulnerable and low-income older people surement error for any single self-reported item. Health is still low, and the poorer shoulder greater burdens of status scores were computed by using Item Response health care costs in terms of percentage of household Theory (IRT) parameter estimates in Winsteps†, a Rasch expenditure (8). measurement software package (http://www.winsteps. As in other developing countries, little empirical com). Higher health status scores within a 0Á100 scale research has been conducted in Viet Nam on the health imply better health status. Respondents who had health status, QoL and related socio-economic status among status scores below the median were categorised as having older people. This article, therefore, taking advantage of poor health status. QoL was assessed by using the eight- the INDEPTH WHO-SAGE study on global ageing and item version of the World Health Organization Quality of adult health (9), aims to describe the patterns of health Life instrument (WHOQoL). Results from the eight items status and QoL among older adults in a rural community were summed to get an overall WHOQoL score which was of Viet Nam, and examine their associations with some then transformed into a 0Á100 scale. The higher the socio-economic factors. WHOQoL score, the better the QoL. Respondents who had WHOQoL scores less than the median were Methods considered as having poor QoL. More details on how scores for this study were derived are given elsewhere (9). Study design and setting This was a population-based cross-sectional study, car- Explanatory variables ried out in Bavi District, a rural community located 60 Explanatory variables included sex (male, female), age km west of Hanoi, the capital, within the Epidemiological (grouped as 50Á59, 60Á69, 70Á79, 80 years), educational Field Laboratory of Bavi (FilaBavi). The FilaBavi Health level (no formal education, up to 6 years of formal

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2124 65 Hoang Van Minh et al.

education, more than 6 years of education), marital Table 1. Background characteristics of study subjects (res- status (in current partnership, never married, separated, pondents and non-respondents) divorced and widowed), wealth index (quintiles), whether respondent stays alone (yes or no) and proportion Respondents Non-respondents of people aged 50 years and over in the same household (n8,535) (n339) (B25%, 25Á49%, 50Á74%, 75%). The wealth quintiles were constructed by using principal component analysis Gender techniques (12). Variables used included household in- Male (%) 3,469 (40.6) 140 (42.6) come, the area of land owned, type of house, materials of Female (%) 5,066 (59.4) 189 (57.4) roof and floor, toilet facilities, electricity and water Age (years) supplies, and ownership of a range of durable assets for 50Á59 (%) 3,221 (37.7) 148 (45) each household. 60Á69 (%) 2,258 (26.5) 87 (26.3) 70Á79 (%) 2,086 (24.4) 45 (13.8) Data analysis 80 and over (%) 970 (11.4) 49 (14.9) Both descriptive and analytical statistics were performed. Mean age (SD) 65.3 (10.7) 63.7 (19.2) The data analysis began with calculation of frequencies and percentages of the variables of interest. Multivariable Education logistic regressions were then carried out to examine No formal education (%) 878 (10.3) 85 (25.9) the association between health status and QoL with Primary orB6 years (%) 4,190 (49.1) 112 (33.9) the selected explanatory variables. More than 6 years (%) 3,467 (40.6) 132 (40.2) Marital status Ethical considerations In current partnership (%) 5,895 (69.1) 215 (65.5) The protocol of this study was approved by the Scientific Now single (%) 2,640 (30.9) 114 (34.5) Board of FilaBavi. All subjects in the study were asked Economic status of household for their written informed consent before collecting data, Poorest quintile 1,209 (14.2) 41 (12.4) and they had complete right to withdraw from the study Second quintile 1,548 (18.2) 52 (15.7) at any time without disadvantage. Third quintile 1,787 (21) 65 (19.9) Fourth quintile 1,996 (23.4) 81 (24.6) Results Least poor quintile 1,976 (23.2) 90 (27.4) Mean number of 4.2 (2) 4.3 (1.9) Characteristics of the study populations household members (SD) Of the total 8,874 people aged 50 and over living in the Proportion of household 50.7 (28.9) 49.5 (28.8) study setting at the time of the survey, there were 8,535 members aged 50 and who participated in the study (amounting to 96%). Four over (SD) percent of subjects were unable to participate as they were away (2.3%) or were not healthy enough to take part in the survey (1.7%). The background characteristics of same age group. Overall, the proportion of respondents potential study subjects (respondents and non-respon- with poor QoL among men and women was 38.4 and dents) are described in Table 1. There were no significant 52.4%, respectively (Table 3). differences in socio-economic characteristics between the Fig. 1 shows the distribution of the study subjects by respondents and the non-respondents. health status and QoL. Women were shown to have poorer health status than men. About 25.9% of men and Distribution of health status and WHOQoL scores 40% of women reported having both poor health status Table 2 presents the distribution of health status scores of and poor QoL. the study population by age and sex. The overall mean health status score was 66.2 and median 65.0. In both Factors associated with poor health status and poor sexes, the average health status scores decreased with age. quality of life (QoL) Men had higher health status scores than women of the Multivariate logistic regression analyses of the association same age group. Overall, the proportion of respondents between poor health status and poor QoL, and socio- with below-median health status among men and women economic status are shown in Table 4. Men were shown to was 39.1 and 58.3%, respectively. be significantly less likely to have poor health status A similar pattern was observed for QoL. The overall compared to women. Older respondents had poorer health mean WHOQoL score was 61.2, median 62.5. In both status than those younger. People with lower educational sexes, the average WHOQoL score decreased with age. levels had a significantly higher probability of having poor Women had lower WHOQoL scores than men of the health status than those with higher educational levels.

66 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2124 WHO-SAGE study on older adults in rural community of Viet Nam

Table 2. Distribution of health status scores by age and sex Men Women among 8,535 adults aged 50 years and over in northern rural 60.0

Viet Nam 50.0 48.3

40.0 Men Women 40.0 Variables (n3,469) (n5,066) 29.3 30.0 (%) 25.9 Mean health status score (SD) 20.0 18.3 50Á59 years 72.5 (11.5) 68.8 (9.4) 13.2 12.5 12.4 60Á69 years 68.8 (9.9) 64.8 (7.9) 10.0 70Á79 years 65.3 (9.2) 61.7 (8.2) 80 years and over 60.1 (8.8) 57.6 (8.2) – Poor health and poor Poor health and non- Non- poor health and Non- poor health and All ages 60.1 (8.8) 57.6 (8.2) quality of life poor quality of life poor quality of life non- poor quality of life

Proportion of respondents 68.9 (11.0) 64.4 (8.4) Fig. 1. Distribution (%) of study subjects by health status with health status score and quality of life, among 8,535 adults aged 50 years and below the median over in northern rural Viet Nam. 50Á59 years (%) 24.4 35.4 60Á69 years (%) 33.1 51.7 educational level, without any marital partnership, with 70Á79 years (%) 50.1 67.1 fewer older people in the family and with lower economic 80 years and over (%) 70.0 81.5 status were more likely to have both poor health status All ages (%) 35.9 54.2 and poor QoL.

Discussion Study subjects in the lowest wealth quintile were more This article describes the pattern of health status and likely to have poor health status than those belonging to QoL among older adults in a rural community of Viet the highest wealth quintile. There was no significant Nam. It reveals socio-economic inequalities in health association between poor health status and marital status status and QoL among older adults in the study setting. or the proportion of older people living in a household. We found that a higher proportion of women reported Similar to the pattern of health status, poor QoL was both poor health status and poor QoL compared to men. shown to be significantly associated with women, older The findings are in line with recent studies on health ages, lower educational levels and lower economic status. status from other Asian countries such as Pakistan (13), Table 4 shows that the respondents whose families had Bangladesh (14) and Singapore (15). Gender inequality in more older people (]75% people aged 50 years and over health has been well documented in the international in the same household) were significantly less likely to literature (16). The findings are also consistent with have poor QoL. Women, older people, those with lower results from previous studies on QoL that reported female disadvantages in both emotional and subjective Table 3. Distribution of WHOQoL scores by age and sex well-being (17Á20). One likely explanation could be that among 8,535 adults aged 50 years and over in northern rural women are more likely to suffer from conditions that are Viet Nam debilitating but not fatal. The paradox is that women report poorer health but live longer, and this is true in Variables Men (n3,469) Women (n5,066) almost every country in the world (21). Age was shown to be a factor significantly associated Mean QoL score (SD) with poor health status and poor QoL. This has been 50Á59 years 65.7 (12.7) 62.2 (12.3) consistently shown in previous studies (13Á15). In our 60Á69 years 64.1 (13.2) 60.9 (12.6) setting, chronic diseases were shown to be more prevalent 70Á79 years 61.9 (14.1) 57.7 (13.2) among women and older people (22). 80 years and over 56.6 (14.1) 53.8 (14.0) We found that higher educational levels were signifi- All ages 63.7 (13.5) 59.5 (13.2) cant positive predictors of both health status and QoL among the study subjects. Education is well known as an Proportion of respondents with WHOQoL score below median important factor for health, both among men and 50Á59 years (%) 32.0 45.1 women, particularly in rural areas. The findings are 60Á69 years (%) 37.3 48.6 consistent with previous studies (13Á15). Education is 70Á79 years (%) 44.5 58.3 assumed to have a positive effect on health status 80 years and over (%) 60.9 66.2 since persons with more education are assumed to be All ages (%) 38.4 52.4 better informed about health matters, diet and disease

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2124 67 Hoang Van Minh et al.

Table 4. Factors associated with poor health status and poor quality of life among 8,535 adults aged 50 years and over in northern rural Viet Nam

OR with 95% CI

Variables Poor health Poor QoL Poor health and poor QoL

Gender Men 0.7 (0.6Á0.8)a 0.8 (0.7Á0.9)a 0.8 (0.7Á0.9)a Women 1 1 1

Age group 50Á59 years 1 1 1 60Á69 years 1.5 (1.3Á1.7)a 1.0 (0.9Á1.1) 1.3 (1.1Á1.5)a 70Á79 years 2.4 (2.1Á2.8)a 1.2 (1.1Á1.4)a 1.9 (1.6Á2.2)a 80 years and over 4.6 (3.7Á5.7)a 1.8 (1.5Á2.1)a 3.0 (2.4Á3.6)a

Educational level No formal education 2.7 (2.2Á3.3)a 2.1 (1.7Á2.5)a 2.3 (1.9Á2.9)a Less than or equal to 6 years 1.6 (1.4Á1.7)a 1.5 (1.4Á1.7)a 1.6 (1.4Á1.8)a More than 6 years 1 1 1

Marital status Now single 1 1 1 In current partnership 0.9 (0.8Á1.0) 0.9 (0.8Á1.0) 0.8 (0.7Á0.9)a

Proportion of people aged 50 years and over in the same household B25% 1.2 (1.0Á1.4) 1.6 (1.4Á1.9)a 1.4 (1.2Á1.6)a 25Á49% 1.1 (1.0Á1.3) 1.6 (1.4Á1.9)a 1.4 (1.2Á1.6)a 50Á74% 1.2 (1.0Á1.4) 1.5 (1.3Á1.7)a 1.4 (1.2Á1.6)a ]75% 1 1 1

Socio-economic quintile Poorest quintile 1.7 (1.4Á2.0)a 3.2 (2.7Á3.8)a 2.5 (2.1Á3.0)a Second quintile 1.2 (1.0Á1.4) 2.0 (1.8Á2.4)a 1.6 (1.4Á1.9)a Third quintile 1.2 (1.0Á1.4) 1.7 (1.5Á2.0)a 1.5 (1.3Á1.8)a Fourth quintile 1.1 (1.0Á1.3) 1.6 (1.4Á1.9)a 1.5 (1.3Á1.7)a Least poor quintile 1 1 1 aSignificant results (95% CI does not include 1). prevention measures leading to better health conditions, to understand causal mechanisms that resulted in consequently leading to higher QoL. particular heath status and QoL outcomes among the Similarly, improvements in economic status are also study population. Secondly, low educational level and the likely to raise both health status and QoL. In addition to presence of impaired cognition in older people might providing means for purchasing health care, higher have led to inaccuracies in the self-reported data. Our economic status can provide better nutrition, housing careful training and field supervision would have over- and recreational opportunities. The findings are consis- come this problem to some extent. tent with previous studies in Asia (13Á15), in Europe (23) In summary, this study provides cross-sectional evi- and the Americas (24). dence on patterns of health status and QoL among older This study also revealed a positive effect of having more adults in rural Viet Nam. The findings reveal problems of older people living in the same family. This positive effect inequality in health status and QoL among older adults may be the result of mutually practised health beliefs and in the study setting by sex, age, education and economic behaviours, shared physical environments and interperso- status. Given these findings, actions to enhance the health nal relations between the older people in the same family. of disadvantaged people (women, the elderly, less edu- We need to note some limitations of this study. Firstly, cated and lower economic status) are needed in this the cross-sectional nature of the data limited our ability setting.

68 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2124 WHO-SAGE study on older adults in rural community of Viet Nam

Acknowledgements 13. Ahmad K, Jafar T, Chaturvedi N. Self-rated health in Pakistan: results of a national health survey. BMC Public Health 2005; 5: This research was supported by FAS, the Swedish Council for Social 51. and Work Life Research, Grant No. 2003-0075. We would like to 14. Rahman MO, Barsky AJ. Self-reported health among older thank INDEPTH WHO-SAGE for support and contribution of the Bangladeshis: how good a health indicator is it? Gerontologist SAGE instrument. 2003; 43: 856Á63. 15. Lim W-Y, Ma S, Heng D, Bhalla V, Chew S. Gender, ethnicity, health behaviour & self-rated health in Singapore. BMC Public Conflict of interest and funding Health 2007; 7: 184. The authors have not received any funding or benefits 16. Vlassoff C. Gender inequalities in health in the third world: from industry to conduct this study. uncharted ground. Soc Sci Med 1994; 39: 1249Á59. 17. Netuveli G, Blane D. Quality of life in older ages. 2008; ldn003. Br Med Bull 2008; 85: 113Á26. References 18. Smith A, Sim J, Scharf T, Phillipson C. Determinants of quality of life amongst older people in deprived neighbourhoods. 1. Omran AR. The epidemiological transition theory revisited Ageing Soc 2004; 24: 793Á814. thirty year later. World Health Statistic Quart 1998; 51: 99Á119. 19. Kabir ZN, Tishelman C, Agu¨ero-Torres H, Chowdhury AMR, 2. World Health Organization. Active aging: a policy framework. Winblad B, Ho¨jer B. Gender and ruralÁurban differences in Geneva: World Health Organization; 2002. reported health status by older people in Bangladesh. Arch 3. United Nations. World population prospects: the 2006 revision Gerontol Geriatrics 2003; 37: 77Á91. and world urbanization prospects; 2006. Available from: http:// 20. Srapyan Z, Armenian HK, Petrosyan V. Health-related quality esaunorg/unpp [cited 10 December 2009]. of life and depression among older people in Yerevan, Armenia: 4. Shrestha LB. Population aging in developing countries. Health a comparative survey of retirement home and household Aff (Millwood) 2000; 19: 204Á12. residents aged 65 years old and over. Age Ageing 2006; 35: 5. UNFPA Viet Nam. How can Vietnam respond to population 190Á3. ageing? Hanoi: UNFPA; 2009. 21. Mascitelli L, Pezzetta F, Sullivan JL. Why women live longer 6. Giang TL, Dfau WD. Demographic changes and the long-term than men: sex differences in longevity. Gend Med 2006; 3: 341. pension finances in Vietnam: a stochastic actuarial assessment. 22. Hoang Van M, Dao Lan H, Kim Bao G. Self-reported chronic J Popul Ageing 2009; 1: 125Á51. diseases and associated sociodemographic status and lifestyle 7. Ministry of Health of Vietnam. Joint annual health review 2008: risk factors among rural Vietnamese adults. Scand J Public health financing in Vietnam. Hanoi: Ministry of Health of Health 2008; 36: 629Á34. Vietnam; 2008. 23. Aberg-Yngwe M, Diderichsen F, Whitehead M, Holland P, 8. Giang TL. Health status and accessibility to health facilities of Burstrom B. The role of income differences in explaining social the elderly in Viet Nam. Soc Insur Rev 2008; 7: 37Á43. inequalities in self-rated health in Sweden and Britain. J 9. Kowal P, Kahn K, Ng N, Naidoo N, Abdullah S, Bawah A, Epidemiol Community Health 2001; 55: 556Á61. et al. Ageing and adult health status in eight low-income 24. Smith KV, Goldman N. Socioeconomic differences in health countries: the INDEPTH WHO-SAGE collaboration. Global among older adults in Mexico. Soc Sci Med 2007; 65: 1372Á85. Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5302 10. Chuc NTK, Diwan VK. FilaBavi, a demographic surveillance site, an epidemiological field laboratory in Vietnam. Scand J *Hoang Van Minh Public Health 2003; 31: 3Á7. Faculty of Public Health 11. INDEPTH Network. Population and health in developing Hanoi Medical University countries. Volume 1: population, health, and survival. Ottawa: No 1, Ton That Tung, Dong Da INDEPTH Network, IDRC; 2002. Hanoi, Viet Nam 12. Vyas S, Kumaranayake L. Constructing socio-economic status Tel: 84 48523798 indices: how to use principal components analysis. Health Policy Fax: 84 45745070 Plan 2006; 21: 459Á68. Email: [email protected]

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2124 69 æINDEPTH WHO-SAGE Supplement Socio-demographic differentials of adult health indicators in Matlab, Bangladesh: self-rated health, health state, quality of life and disability level Abdur Razzaque1,2*, Lutfun Nahar3, Masuma Akter Khanam4 and Peter Kim Streatfield1,2 1Matlab Health and Demographic Surveillance System, ICDDR,B, Mohakhali, Dhaka, Bangladesh; 2INDEPTH Network, Accra, Ghana; 3Department of Social Science, East West University, Dhaka, Bangladesh; 4Chronic Disease Unit, ICDDR,B, Dhaka, Bangladesh

Background: Mortality has been declining in Bangladesh since the mid- twentieth century, while fertility has been declining since the late 1970s, and the country is now passing through the third stage of demographic transition. This type of demographic transition has produced a huge youthful population with a growing number of older people. For assessing health among older people, this study examines self-rated health, health state, quality of life and disability level in persons aged 50 and over. Data and methods: This is a collaborative study between the World Health Organization Study on global AGEing and adult health and the International Network for the Demographic Evaluation of Populations and Their Health in developing countries which collected data from eight countries. Two sources of data from the Matlab study area were used: health indicator data collected as a part of the study, together with the ongoing Health and Demographic Surveillance System (HDSS) data. For the survey, a total of 4,000 randomly selected people aged 50 and over (HDSS database) were interviewed. The four health indicators derived from these data are self-rated health (five categories), health state (eight domains), quality of life (eight items) and disability level (12 items). Self-rated health was coded as dummy while scores were calculated for the rest of the three health indicators using WHO-tested instruments. Results: After controlling for all the variables in the regression model, all four indicators of health (self-rated health, health state, quality of life and disability level) documented that health was better for males than females, and health deteriorates with increasing age. Those people who were in current partnerships had generally better health than those who were single, and better health was associated with higher levels of education and asset score. Conclusions: To improve the health of the population it is important to know health conditions in advance rather than just before death. This study finds that all four health indicators vary by socio-demographic characteristics. Hence, health intervention programmes should be targeted to those who suffer and are in the most need, the aged, female, single, uneducated and poor. Keywords: adult health; self-rated health; health state; quality of life; disability; Matlab; Bangladesh; INDEPTH WHO-SAGE

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘SAGE data’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 10 December 2009; Revised: 3 June 2010; Accepted: 8 July 2010; Published: 27 September 2010

ortality has been declining in Bangladesh since country is now passing through the third stage of the mid-twentieth century, while fertility has demographic transition (1). This type of demographic Mbeen declining since the late 1970s, and the transition has produced a huge youthful population and

Global Health Action 2010. # 2010 Abdur Razzaque et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 70 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.4618 Socio-demographic differentials of adult health indicators in Matlab

a growing number of older people. Due to such an age socio-demographic characteristics as well as the inter- structure, the population is now experiencing a double relationship of these health indicators. disease burden; over 50% of deaths in Matlab are now due to non-infectious diseases (2). Methods Bangladesh is one of the 20 developing countries with the largest numbers of older people, and by 2025 Setting Bangladesh, along with four other Asian countries, will Data for this study come from Matlab Upazila (sub- account for about half of the world’s older population (3). district) where the International Centre for Diarrhoeal In fact, population increase among those aged 65 and Disease Research, Bangladesh (ICDDR,B) has main- over was negligible in Bangladesh during the first half tained a field station since 1963. Matlab is a rural area of the twentieth century, but it increased substantially located about 55 km south-east of Dhaka. The area is a during the second half (2.4 million) and it is projected to low-lying deltaic plain intersected by the tidal river Gumti increase by 20.8 million during the first half of the and its numerous canals. In the past, major modes of twenty-first century (4). transport within the area were walking, country boat and As social security is almost non-existent for older people in some cases small steamer or launch. However, in recent in Bangladesh (pension for government and semi- years most of the villages have become accessible by government employees 5%, and governmental support rickshaw. Farming is the dominant occupation, except in for elderly people 10%), older people usually live in a few villages where fishing is the means of livelihood (13). extended households and depend primarily on adult Most of the farmers are in marginal situations with less children for economic support and personal care (5). than a hectare of land and 40% of them are landless. For However, the traditional family support system for older many families, sharecropping and work on others’ land on people is under pressure due to the increasing out- a daily wage basis have become the main sources of migration of household members to cities, and women’s livelihood. Some people work in mills and factories in labour force participation outside the home, causing different towns and cities but their families live in the study vulnerability for older people. area. RuralÁurban out-migration is about 5% in recent In Bangladesh about 50% of the population fall below years, while it is about 1% for international migration; the poverty line, and so older people are likely to be in ill however, these rates were much lower in the 1980s (3.3% vs. health, in social isolation and in poverty (6). Moreover, 0.3%). Women are largely restricted to activities in the the majority of the older people live in rural areas where home, with relatively little opportunity to venture outside there is no specialised care service for older people the home, although these restrictions have decreased in recent years. Rice constitutes the staple food and is in health facilities (Upazila Health Complex). Based harvested three times annually. Rates of illiteracy are on Matlab data, it was documented earlier that the high and are higher among older people. prevalence of chronic morbidity was 75% among older Since 1966 the ICDDR,B has maintained a Health and people (last 3 months) while it was about 50% (last Demographic Surveillance System (HDSS) in the Matlab 1 month) for acute morbidity (7); 2.1% of older males and area covering about 225,000 people. The surveillance 3.6% of females could not use a toilet without help. system collects data on births, deaths, migrations, mar- As costs associated with assessing health status of a riages, divorces and household divisions (14), and also population are high, there is a need for low-cost health collects cross-sectional socio-economic data which are indicators, particularly for developing countries. Currently, available for 1974, 1982, 1996 and 2005. The HDSS data some low-cost health indicators are available for developed are of high quality because they have been collected countries that are good predictors of mortality and during regular household visits (every 2 weeks until 1997, functional ability (8Á11), but such indicators are rare for every month between 1998 and 2006 and every 2 months the developing countries. Based on the Matlab Health and since then) by the Community Health Research Workers Socio-economic Status Survey of Bangladesh, (12) it was (CHRWs). reported that adults of this community can effectively Since October 1977, half of the surveillance area has assess their own health even with poor education and low been exposed to Maternal and Child Health and Family levels of interaction with the modern health system. Planning (MCH-FP/ICDDR,B service area) services while The current study has collected data on four indicators the other half is a comparison area (15, 13). These two areas of health using a summary version (SAGEÁINDEPTH) of are almost similar in socio-economic conditions but differ the full WHO-SAGE questionnaire: self-rated health, in access to the MCH-FP programme. Beginning in health state, quality of life and disability level. The study 1996, the community-based maternity care service of will examine these four health indicators for people aged the ICDDR,B service area was gradually phased out 50 and over, and their relationship with various and replaced by a facility-based strategy of sub-centres.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.4618 71 Abdur Razzaque et al.

However, these health services are targeted mainly at face-to-face interview and contact with absentees was mother and child health and not to older population, attempted three times. As a quality check, about 2% of except for services for diarrhoeal diseases. In fact, treat- samples were re-interviewed by an independent field ment for diarrhoea has been provided from the Matlab worker/supervisor and feedback was incorporated field hospital since the beginning and such service is open accordingly. to all irrespective of place of residence. Based on the survey data, four health indicators were The history of modern medicine is rather short in calculated: self-rated health, health state, quality of life Bangladesh, since it did not reach the rural population and disability level. Self-rated health was a categorical until after World War II. During 1947Á1970 the physical variable (five categories), health state was measured infrastructure for delivering health services by the then through eight domains (affect, cognition, interpersonal government was mainly urban-based, and such services activities and relationship, mobility, pain, self-care, sleep/ were more curative than preventive in nature. The energy, and vision), quality of life was measured through government accepted primary health care as a national eight items and disability was assessed through 12 items health objective in 1978, since when the health care (17). Self-rated health was coded as a dummy while system has been reoriented to provide essential care to the scores were calculated using the WHO-tested instruments general mass of the population. Funding for the health for health status, quality of life and disability level. All sector increased significantly from the early 1980s, with three of these scores were transformed into new facilities including Maternal and Child Welfare 0Á100 scales on which higher scores indicate better Centres in urban and sub-urban areas, Upazila Health outcomes [better health status, better quality of life Complexes at Upazila level and Family Welfare Centres (WHOQoL) and better functional ability (WHODASi)]. at Union level (16). In Matlab town the government runs Analyses were undertaken using both bivariate a 31-bed free general hospital with nine doctors (Upazila and multivariate methods. The dependent variable was Health Complex) along with several Family Welfare dichotomous for self-rated health and involved contin- Centres, each with a sub-assistant Community Medical uous scores for health state, quality of life and disability Officer and a Family Welfare Visitor. Except for the level. The independent variables were age of respondent, service from Upazila Health Complex, all other services sex, marital status, proportion of people aged above 50 in are targeted to maternal and child health. Finally, there the household, education level and asset quintiles. are across the country both private practitioners (quali- Age was grouped into four (50Á59, 60Á69, 70Á79 and 80 fied and unqualified), private clinics (in big cities) and and over), completed years schooling into three (none, 1Á traditional practitioners (Ayurvedic, Unani and Homoeo- 5 and 6 years or more), marital status into two (now pathy); these services cover the population across all age single and in current partnership) and proportion of groups. people aged above 50 in household into four groups (B0.25, 0.25Á0.49, 0.50Á0.74 and 0.75 or more). Asset Data and methods index was calculated based on a number of consumer This is a multi-country study between the World Health items (radio, watch, etc.), dwelling characteristics (wall Organization Study on global AGEing and adult health and roof material) and type of drinking water and toilet (SAGE) and the International Network for the Demo- facilities (18). For this study we have studied first to fifth graphic Evaluation of Populations and Their Health in quintiles as poorest to richest. developing countries (INDEPTH), and collected data For examining the interrelationship between two from eight countries of Africa and Asia. Two sources of variables, self-rated health was grouped into two categories data from the Matlab study area were used: survey data (very good, good, moderate1 and bad/very bad0); collected as a part of the study and the ongoing HDSS health status (IRT health 555.20and55.21); quality data. For the survey, questionnaires were received from of life (WHOQoL 580.00 and80.01); disability level (WHODASi 81.0 0 and 81.0 1); x2-tests the SAGEÁINDEPTH and piloted in the field after 5    translating into local languages. A total of 4,000 people were performed for significance level. 50 years and older, out of 31,400, were selected randomly from the HDSS database (ICDDR,B-service area); a Results sample from half of the HDSS area was selected to About two-fifths of the sample belonged to the age group minimise travel time to visit the sample households. 50Á59 years while about one-fifth were aged 70 and over The survey was conducted by a team of college-graduate (Table 1). Educational level was low, with about 55% females with data collection experience. Interviewers illiterate and only about 15% had six or more years of received extensive training on data collection, particularly schooling. About 25% of people were single, 30% of about asking questions on sensitive topics and on the data household members were 50 years or older and mean collection tools designed for the survey. The interviews household size was slightly over 5. Sample households are were conducted at the residence of the respondent by not equally distributed across quintiles, with more from

72 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.4618 Socio-demographic differentials of adult health indicators in Matlab

Table 1. Background characteristics (%) of the study popu- Table 2. Distribution of health indicators by age and sex for lation in Matlab, Bangladesh 4,037 adults aged 50 and over in Matlab, Bangladesh

Respondents Non-respondents Men Women Variables (N3,990) (N31,425) Indicators (N2,016) (N2,021)

Sex Self-rated health (Percentage of very good/good/moderate) Men 49.9 47.4 50Á59 years 87.2 77.4 Women 50.1 52.6 60Á69 years 77.9 60.1 70Á79 years 64.4 42.9 Age group (years) 80 years and over 48.9 24.2 50Á59 45.3 44.0 60Á69 33.8 34.0 Mean health status (score) 70Á79 17.1 17.3 50Á59 years 65.7 57.7 80 and over 3.8 4.7 60Á69 years 62.2 55.4 70Á79 years 59.2 51.3 Education level 80 years and over 55.6 50.7 No formal education 56.3 57.4 Less than or equal to 6 years 28.7 28.7 Mean quality of life (score) More than 6 years 14.9 13.9 50Á59 years 80.3 77.3 60Á69 years 79.0 74.7 Marital status 70Á79 years 77.8 72.1 Now single 23.8 29.7 80 years and over 76.4 71.4 In current partnership 76.2 70.3 Mean functional ability level (score) Socio-economic quintile 50Á59 years 84.0 62.1 First quintile 15.2 13.6 60Á69 years 76.1 54.5 Second quintile 16.6 16.8 70Á79 years 66.3 45.8 Third quintile 17.5 20.3 80 years and over 54.6 42.0 Fourth quintile 23.2 23.5 Fifth quintile 27.4 25.9 Mean number of household 5.4 4.9 Table 3 shows multivariate relationships for self- members rated health and health status by socio-demographic Percentage of household 18.6 16.6 characteristics. After controlling for all other variables members aged 50 years in the regression model (logistic), males reported sig- and over nificantly better health (2.19 times) than females; health got significantly worse as age increased (7.70 times better for age group 50Á59 and reduced to 2.07 times for age the fourth and fifth quintiles, because the quintiles are group 70Á79 compared to age group 80 years and over); population-based. In fact, sample characteristics are educated people had significantly better health than comparable to the population characteristics. uneducated (0.74 times for those with no formal educa- All four measures of health indicator (self-rated health, tion and 0.87 times for those less or equal to 6 years health state, quality of life and disability level) indicated compared to those with six or more years of education); that health was better for males than females irrespective and health got significantly worse as socio-economic of age categories and health deteriorated gradually as age status declined (0.74 times for first quintile to fifth increased (Table 2). For self-rated health, the proportion quintile). with good health declined from 87.2% to 48.9% for males For health status, after controlling for all other variables and 77.4% to 24.2% for females between age groups 50Á in the regression model (linear regression), the score for 59 and 80 years and over; while for health status, the males increased by 7.07 per unit change in the female score; mean score declined from 65.7 to 55.6 for males and 57.7 for age group 50Á59, the score increased by 8.76 per unit to 50.7 for females between these two age groups. For change and 2.51 times per unit change for age group 70Á79 quality of life, the mean score decreased from 80.3 to 76.4 compared to those in age group 80 years and over; for no for males and 77.3 to 71.4 for females between age groups formal education the score declined by 1.22 per unit change 50Á59 and 80 years and over; while for functional ability and by 0.74 per unit change for those with less or equal to level, the mean score decreased from 84.0 to 54.6 for six years compared to those with more than six years of males and 62.1 to 42.0 for females between these two age schooling; for single persons the score declined by 0.08 per groups. unit change of those in a current partnership; and for first

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.4618 73 Abdur Razzaque et al.

Table 3. Multivariate models of factors associated with self-rated health (logistic regression) and health state (linear regression) for 4,037 adults aged 50 and over in Matlab, Bangladesh

Variables Self-rated health (Exponent of b and 95% CI) Health status (b coefficient and 95% CI)

Sex (ref: women) Men 2.19 (1.83, 2.62)** 7.07 (6.48, 7.66)** Age group (ref: 80 years and over) 50Á59 years 7.70 (5.34, 11.09)** 8.76 (7.44, 10.07)** 60Á69 years 4.06 (2.84, 5.08)** 5.95 (4.64, 7.25)** 70Á79 years 2.07 (1.43, 2.99)** 2.51 (1.15, 3.87)**

Education level (ref: more than 6 years) No formal education 0.74 (0.57, 0.95)* 1.22 (1.98, 0.46)** Less or equal to 6 years 0.87 (0.67, 1.13) 0.74 (1.51, 0.03)***

Marital status (ref: in current partnership) Now single 0.97 (0.79, 1.18) 0.08 (0.78, 0.63)

Proportion aged 50 years and over in the household (ref: ]0.75) 0.25 1.06 (0.82, 1.38) 0.03 (0.93, 0.87) 0.25Á0.49 0.97 (0.75, 1.25) 0.08 (0.95, 0.79) 0.50Á0.74 0.83 (0.63, 1.10) 0.55 (1.51, 0.41)

Socio-economic quintile (ref: Fifth quintile) First quintile 0.74 (0.58, 0.94)* 1.04 (1.85, 0.23)* Second quintile 0.81 (0.64, 1.02)*** 1.33 (2.10, 0.57)** Third quintile 0.78 (0.62, 0.98)* 0.90 (1.64, 0.15)* Fourth quintile 0.93 (0.76, 1.15) 0.56 (1.24, 0.11)

*PB0.05; **PB0.01; ***PB0.10. socio-economic quintile the score declined by 1.04 per unit those in age group 70Á79 compared to those 80 years or change compared to those in fifth quintile. more; for no formal education the score decreased by Table 4 shows the multivariate relationship of quality 4.31 per unit change and by 2.66 per unit change for of life (WHOQoL) and disability level (WHODASi) by those with less or equal to 6 years compared to those with socio-demographic characteristics. After controlling for six or more years of schooling; and for the first socio- all other variables in the regression model (linear regres- economic quintile the score decreased by 2.32 per unit sion), the WHOQoL score for males increased by 2.01 change compared to those in fifth quintile. per unit change in female score; for age group 50Á59 the All four health indicators (self-rated health, health score increased by 3.42 per unit change and by 0.87 per state, quality of life and disability level) show that males, unit change for age group 70Á79 compared to those in age those who were younger, educated and those in higher group 80 years or more; for single persons the score socio-economic groups reported better health, compared decreased by 4.04 per unit change of those in a current to females, older age groups, illiterates and those in lower socio-economic groups. partnership; for no formal education the score decreased Table 5 shows the interrelationship of different health by 0.81 per unit change, and by 0.31 per unit change for indicators. Results show that all four health indicators are those with less or equal to 6 years compared to those with highly significantly related to each other. six years or more schooling; and for the first socio- economic quintile the score decreased by 2.95 per unit change and by 0.93 per unit change for those in the Discussion fourth quintile compared to those in the fifth quintile. Bangladesh is currently passing through the third stage For functional ability level, after controlling for all of demographic transition, where both fertility and other variables in the regression model (linear regression), mortality rates are at relatively low levels. Such as the score for males increased by 20.17 per unit change in demographic transition has produced a huge youthful the female score; for age group 50Á59 the score increased population with a growing number of older people (4), by 25.49 per unit change and by 8.96 per unit change for where disease patterns are changing from infectious to

74 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.4618 Socio-demographic differentials of adult health indicators in Matlab

Table 4. Multivariate models (linear regression) of factors associated with quality of life and functional ability level for 4,037 adults aged 50 and over in Matlab, Bangladesh

Variables Quality of life (b coefficient and 95% CI) Functional ability level (b coefficient and 95% CI)

Sex (ref: women) Men 2.01 (1.68, 2.34)** 20.17 (18.82, 21.52)** Age group (ref: 80 years and over) 50Á59 years 3.42 (2.69, 4.16)** 25.49 (22.50, 28.48)** 60Á69 years 2.07 (1.34, 2.80)** 18.08 (15.00, 21.06)** 70Á79 years 0.87 (0.11 1.63)* 8.96 (5.86, 12.06)** Education level (ref: more than 6 years) No formal education 0.81 (1.23, 0.38)** 4.31 (6.05, 2.57)** Less or equal to 6 years 0.31 (0.75, 0.11) 2.66 (4.42, 0.89)** Marital status Now single (ref: in current partnership) 4.04 (4.43, 3.64)** 0.19 (1.42, 1.82) Proportion aged 50 years and over in the household (ref: ]0.75) 0.25 0.23 (0.74, 0.26) 0.34 (1.70, 2.40) 0.25Á0.49 0.04 (0.53, 0.44) 0.80 (1.19, 2.80) 0.50Á0.74 0.04 (0.57, 0.50) 0.37 (2.57, 1.83) Socio-economic quintile (ref: least poor quintile) Poorest quintile 2.95 (3.41, 2.50)** 2.32 (4.17, 0.48)* Second quintile 2.29 (2.71, 1.86)** 2.04 (3.76, 0.30)* Third quintile 1.40 (1.82, 0.98)** 1.46 (3.16, 0.23) Fourth quintile 0.93 (1.31, 0.55)** 0.72 (2.27, 0.81)

*PB0.05; **PB0.01. non-infectious (2). Traditionally, older people are viewed quality of life and ability level) deteriorated with in this society as an integral part of the family and used to increasing age. The finding is in agreement with a recent enjoy absolute authority over the younger generation; study from Matlab that the prevalence of chronic disease however, the status of older people is under pressure due increased with age (20). It is likely that this population will to demographic, social and economic change (19). need more support (physical/co-residence, social and As a result of mortality decline during the past few economic) as the number of older people is increasing decades, life span has increased significantly in Bangladesh rapidly along with an increase in chronic diseases. but it is not known whether health status has improved In Bangladesh, older females survive better than males during the increased life span. The study found that all (2) but health indicators from the current study (self- four health indicators (self-rated health, health state, rated health, health state, quality of life and disability

Table 5. Inter-relationship of different health indicators in order persons, Matlab, Bangladesh

Quality of life Disability level Health state Self-rated health

Quality of life Disability level x2 526.7 PB0.001 Health state x2 355.8 x2 645.6 PB0.001 PB0.001 Self-rated health x2 313.3 x2 303.8 x2 499.2 PB0.001 PB0.001 PB0.001

Note: Health indicators (categories): self-rated health (very good, good, moderate1 and bad/very bad0); Health status (IRT health 55.21 and 555.20); Quality of life (WHOQoL 580.00 and80.01); Disability level (WHODASi 581.00 and81.01).

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.4618 75 Abdur Razzaque et al.

level) demonstrate that females are worse-off than males 4. Razzaque A. Population growth of Bangladesh: past, present during old age. However, it was reported that the health and future. South Asian J Popul Health (submitted for disadvantage for women reflect their ‘greater sensitivity’ publication). 5. Chaklader H, Haque M, Kabir M. Socio-economic situation to health conditions (12). In this society, where women of urban elderly population: evidence from a micro study. In: continue to be valued less than men as documented in the Sattar MA, Kabir M, Elahi KM, Abedin S, eds. The elderly: past (21), older women’s health reflects their lifelong contemporary issues. Dhaka: Dhaka Presidency Press; 2003, experience of discrimination, deprivation and neglect (6). pp. 1Á13. Traditionally, older women also own fewer assets and 6. Chaklader H, Kabir M. What we know about elderly have less control over family income, and a recent study population’s health problems: evidences from rural Bangladesh villages. In: Sattar MA, Kabir M, Elahi KM, Abedin S, eds. from Matlab reported that females experience more The elderly: contemporary issues. Dhaka: Dhaka Presidency chronic disease than their male counterparts (20). Press; 2003,. p. 134Á53. All four health indicators documented that health 7. Mostafa G, Streatfield PK. Health implications of an aging is better among educated/rich than uneducated/poor Bangladeshi population. In: Sattar MA, Kabir M, Elahi KM, people. The finding is also in agreement with mortality Abedin S, eds. The elderly: contemporary issues. Dhaka: Dhaka patterns, in which educated/rich people had lower Presidency Press; 2003,. p. 33Á53. 8. Idle EI, Benyamini, Y. Self-rated health and mortality: a review mortality than uneducated/poor (2). Some years ago, it of 20-seven community studies. J Health Soc Behav 1987; 38: was reported (22) that socio-economic differentials in 21Á37. mortality indicate that a degree of success has been 9. Idle EI, Kasl SV. Health perceptions and survival: do global achieved in one section of the community that has not evaluations of health status really predict mortality? J Gerontol been achieved in others. In Matlab (20), it has been 1991; 46: 555Á65. documented that some chronic diseases (stroke, heart 10. Idle EI, Kasl SV. Self-ratings of health: do they also predict disease, diabetes) increase with increased education while changes in functional ability? J Gerontol Ser B Psychol Sci Soc Sci 1995; 50: S344Á53. others (joint pain, pulmonary, hypertension, cancer) 11. Frankenberg E, Jones N. Self-rated health and mortality: does decrease. the relationship extend to a low-income setting? Minneapolis, All four health indicators were found to be interrelated MN: Population Association of America; 2003. and these indicators also showed similar patterns 12. Kuhn R, Rahman O, Menken J. Survey measures of health: how by socio-demographic characteristics. This indicates that well do self-reported and observed indicators measure health these health indicators, although measuring different and predict mortality? In: Cohen B, Menken J, eds. Aging in Sub-Saharan Africa: recommendations for furthering research. dimensions of health, had some common characteristics. Washington, DC: National Academies Press; 2006, pp. 314Á41. Preliminary analysis of the same dataset show that these 13. Van Ginneken J, Bairagi R, de Francisco A, Sarder AM, four health indicators are also predictors of subsequent Vaughan P. Health and demographic surveillance in Matlab: mortality (23). past, present and future, special publican No. 72. Dhaka: To improve the health of the population, it is im- ICDDR,B; 1998. portant to know their health status in advance rather 14. Razzaque A, Streatfield PK. Matlab demographic surveillance system, Bangladesh. In: Sankoh OA, Kahn K, Mwageni E, than just before death. The findings of this study have Ngom P, Nyarko P, eds. Population and health in deve- policy implications in terms of assessing the overall loping countries. Vol. 1. Ottawa: IDRC & INDEPTH; 2002, burden of diseases and effectiveness of health systems. pp. 287Á95. Moreover, the study indicates that health intervention 15. Bhatia S, Mosley WH, Faruque ASG, Chakraborty J. The programmes should be targeted to those who suffer and Matlab family planning-health services project. Stud Fam Plann need most: the older, female and uneducated/poor people. 1980; 11: 202Á11. 16. DGHS (Directorate General of Health Services). Bangladesh Health Services-1989. Government of the People’s Republic of Conflict of interest and funding Bangladesh: DGHS; 1990. 17. Kowal P, Kahn K, Ng N, Naidoo N, Abdullah S, Bawah A, The authors have not received any funding or benefits et al. Ageing and adult health status in eight low-income from industry to conduct this study. countries: the INDEPTH WHO-SAGE collaboration. Global Health Action Supplement 2, 2010: DOI: 10.3402/gha.v3i0. 5302. References 18. Razzaque A, Streatfield PK, Gwatkin D. Does health interven- tion improve socioeconomic inequalities of neonatal, infant 1. Bairagi R, Datta AK. Demographic transition in Bangladesh: and child mortality? Evidence from Matlab, Bangladesh. Int J what happened in the twentieth century and what will happen Equity Health 2007; 6: 4. DOI: 10.1186/1475-9276-6-4. next? Asia Pac Popul J 2001; 16: 3Á16. 19. Kabir ZN. Traditional norms vs reality: the case of elderly. 2. Razzaque A, Carmichael GA, Streatfield PK. Adult mortality in Paper presented at the women’s health conference, Singapore, 15 Matlab: levels, trends and determinants. Asian Popul Stud 2009; May 1999. 5: 85Á100. 20. Razzaque A, Ali Ashraf, et al. Self-reported common chronic 3. EastÁWest Centre. The future of population of Asia. Honolulu: diseases and their relationship with body mass index: evidence EastÁWest Centre; 2002. from ICDDR,B field sites; 2010 (unpublished).

76 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.4618 Socio-demographic differentials of adult health indicators in Matlab

21. Chen LC, Huq E, D’Souza S. Sex bias in the family allocation of food and health care in rural Bangladesh. Popul Dev Rev 1981; *Abdur Razzaque 7: 55Á70. HDSU 22. Antonovsky A, Bernstein J. Social class and infant mortality. ICDDR,B Soc Sci Med 1977; 11: 453Á70. Mohakhali, Dhaka-1212 23. Razzaque A, Mustafa AHMG, et al. Self-rated health and Bangladesh subsequent mortality: evidence from Matlab, Bangladesh; 2010 Email: [email protected] (unpublished).

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.4618 77 æINDEPTH WHO-SAGE Supplement Health and quality of life among older rural people in Purworejo District, Indonesia Nawi Ng1,2,3*#, Mohammad Hakimi2,3, Peter Byass1#, Siswanto Wilopo2,3 and Stig Wall1#

1Department of Public Health and Clinical Medicine, Centre for Global Health Research, Epidemiology and Global Health, Umea˚ University, Umea˚ , Sweden; 2Purworejo Health and Demographic Surveillance System, Faculty of Medicine, Gadjah Mada University, Yogyakarta, Indonesia; 3INDEPTH Network, Accra, Ghana

Introduction: Increasing life expectancy and longevity for people in many highly populated low- and middle- income countries has led to an increase in the number of older people. The population aged 60 years and over in Indonesia is projected to increase from 8.4% in 2005 to 25% in 2050. Understanding the determinants of healthy ageing is essential in targeting health-promotion programmes for older people in Indonesia. Objective: To describe patterns of socio-economic and demographic factors associated with health status, and to identify any spatial clustering of poor health among older people in Indonesia. Methods: In 2007, the WHO Study on global AGEing and adult health (SAGE) was conducted among 14,958 people aged 50 years and over in Purworejo District, Central Java, Indonesia. Three outcome measures were used in this analysis: self-reported quality of life (QoL), self-reported functioning and disability, and overall health score calculated from self-reported health over eight health domains. The factors associated with each health outcome were identified using multivariable logistic regression. Purely spatial analysis using Poisson regression was conducted to identify clusters of households with poor health outcomes. Results: Women, older age groups, people not in any marital relationship and low educational and socio- economic levels were associated with poor health outcomes, regardless of the health indices used. Older people with low educational and socio-economic status (SES) had 3.4 times higher odds of being in the worst QoL quintile (OR3.35; 95% CI2.73Á4.11) as compared to people with high education and high SES. This disadvantaged group also had higher odds of being in the worst functioning and most disabled quintile (OR1.67; 95% CI1.35Á2.06) and the lowest overall health score quintile (OR1.66; 95% CI1.36Á2.03). Poor health and QoL are not randomly distributed among the population over 50 years old in Purworejo District, Indonesia. Spatial analysis showed that clusters of households with at least one member being in the worst quintiles of QoL, functioning and health score intersected in the central part of Purworejo District, which is a semi-urban area with more developed economic activities compared with other areas in the district. Conclusion: Being female, old, unmarried and having low educational and socio-economic levels were significantly associated with poor self-reported QoL, health status and disability among older people in Purworejo District. This study showed the existence of geographical pockets of vulnerable older people in Purworejo District, and emphasized the need to take immediate action to address issues of older people’s health and QoL.

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘‘SAGE data’’ as its subject, detailing how you propose to use the data, to [email protected]

Keywords: adult health; health status; clustering; quality of life; disability; ageing; Purworejo; Indonesia; INDEPTH WHO-SAGE

Received: 3 November 2009; Revised: 28 June 2010; Accepted: 8 July 2010; Published: 27 September 2010

#Editor, Nawi Ng, Deputy Editor, Peter Byass, Chief Editor, Stig Wall, have not participated in the review and decision process for this paper.

Global Health Action 2010. # 2010 Nawi Ng et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 78 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 Health of older people in Indonesia

dvances in public health and medical technolo- Indonesian older people play an important role in their gies have driven population growth in the last families and their society. In traditional Javanese society, Acentury. Increasing life expectancy and longevity older parents typically co-reside with one of their young- in many highly populated low- and middle-income est children, usually a daughter (extended family), who countries has led to an increased number of older people. accepts responsibility to take care of them until they die. In 2006, about 500 million people (7.5%) of the 6.5 billion Well-off older persons provide key intergenerational world population were aged 65 years and over, and this support for families (9), have high social status and are number is projected to double by 2030, to represent respected in their communities. Javanese people highly 12.5% of the global population (1, 2). During 2007Á2050, respect older people because of the value placed on the population in low- and middle-income countries is lineage. Though many of the older people in Indonesia, projected to increase by 61% (3). By 2030, the population particularly those who are widowed, live in poverty, they aged 65 years and over is projected to increase by 140% in also contribute significantly to the rural economy; many developing countries (1, 4). About 53% of this older engage actively in agricultural industries as non-skilled population lives in Asia, home to 61% of the world’s labour. Most of them, particularly women, have low population. education. Older people who are still working are less Indonesian population structure has shifted signifi- economically dependent on their next-of-kin (10). cantly towards an ageing population since 1950. The total Elderly care and intergenerational relationships have fertility rate (TFR) has decreased from 5.5 in 1950Á1955 become an emerging issue, particularly for those who to 2.4 children per woman in 2000Á2005. Life expectancy live in urban areas, as societal values change from has increased from 37.5 to 68.6 years during the same extended family to nuclear family structures, and younger generations become more mobile in search of better period. As a consequence, the population aged 60 years career opportunities. and over increased from 6.2% in 1950 to 8.4% in 2005 A significant amount of research and literature on and is projected to increase to 23.7% in 2050 (5). The older people in Indonesia is available, mainly from Indonesia National Socio-Economic Survey in 2004 anthropological studies focusing on the socio-cultural showed variation in the proportion of older people across aspects of ageing, intergenerational relationships and the provinces in Indonesia ranging from 2% in Papua to changes in family structure and support for older people 12.8% in Yogyakarta. The proportion of older people in (10Á13). However, studies on health status and QoL Central Java was about 9.5%. The survey also showed among the older population are largely lacking, and very that about one-third of those over 60 years reported an little is known about morbidity among Indonesia’s older illness during the month prior to the survey with no population (6, 14). differences between rural and urban areas (6). Self-reported health has been identified as a strong The expected growth in the ageing population in predictor of morbidity and subsequent mortality (15, 16). Indonesia poses significant challenges to the health While evidence has mainly come from developed coun- system and government. Currently, the health system tries, it can also be extended to low-income settings such focuses more on battling infectious diseases such as as Indonesia, as shown by Frankenberg and Jones in the malaria, tuberculosis, diarrhoea and dengue fever. Re- panel data analysis of the Indonesia Family Life Survey sources have not been allocated proportionally to the (IFLS) in 1993, 1997 and 2000. The IFLS data shows that larger and increasingly threatening burden of chronic individuals who perceived their health as poor are more non-communicable diseases such as heart diseases, stroke, likely to die, and the association remains even after being diabetes, cancer and hypertension (7). Changing family adjusted for physical function, physical illness and structure and patterns of work and retirement pose depression, weight, height and indicators of high blood immediate economic challenges, particularly to the social pressure (17). An understanding of older people’s health insurance systems. The pensions and social insurance and well-being will provide important information on any system only cover a small percentage of the Indonesian special health care needs and demand for services, and population who work in the formal sector, which excludes this knowledge can be used to guide planning of health most of the older population. Indonesian social insurance interventions and programmes (18). schemes, which are limited to covering formal workers in The primary objectives of this study are to describe productive age groups and poor population sectors, are patterns of socio-economic and demographic factors that not designed to anticipate an ageing population (8). The determine the health status of older people in Indonesia. lack of a social safety net increases the vulnerability of The secondary objective is to identify the clustering older people to poor health and quality of life (QoL), pattern of poor health among them. Knowledge on the mostly due to the threat of chronic illness from non- determinants of health status and spatial distribution of communicable diseases, and lack of financial support for poor health will help to improve our understanding of accessing health care. older people’s health, thus providing evidence for the

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 79 Nawi Ng et al.

district authorities in promoting better health status and also linked to data from the household socio-economic developing targeted interventions for disadvantaged po- survey conducted in 2004. The socio-economic survey pulations in their specific geographical areas. collected data on household characteristics and owner- ship of non-disposable and disposable goods, and socio- Methods economic status (SES) quintiles were derived through principal component analysis (21). The final merged Study area and participants dataset was converted into STATA data format for data The WHO Study on global AGEing and adult health analysis. (SAGE) (19) was conducted in a functioning Health and Demographic Surveillance System (HDSS) site in Pur- worejo District, Central Java, Indonesia. The Purworejo Data analysis HDSS is a member of the INDEPTH Network, which Three outcome measures were used in the analysis: self- consists of 38 HDSS sites in Africa, Asia and Oceania reported QoL, self-reported problems in functioning (http://www.indepth-network.org). The district is located and disability, and overall health status. Each of those between longitudes 1098 and 1108E and latitude 78S, measures was developed as composite indices from series about 60 km from Yogyakarta City. It covers an area of of validated questions (22). 1,035 km2, spanning a diverse geographical area from the The composite index for self-reported QoL was coast in the south to the mountains in the north. The adapted from the WHO Quality of Life (WHOQoL) district has 750,000 inhabitants (26% under 15 years old, tool (23). The index was derived from eight questions 63% in the economically productive age group and 11% assessing respondent’s thoughts about their life and life over 65 years old). Eleven percent of those over six years situation, satisfaction with themselves and their health, of age have had no formal education. About 89% of the ability to perform daily living activities, personal relation- population live in rural areas. The Purworejo HDSS has ships, living conditions and overall life. Answers to the been running since 1994 covering a total of 600,000 Likert scale were summed up and later transformed to a person-years of observation (20). In 2006, the total 0Á100 scale with 0 representing the worst QoL and 100 population under surveillance was 55,000 (13,443 house- representing the best QoL. holds living in 128 enumeration areas). The study was Questions to assess problems in functioning and conducted between January and June 2007. We identified disability were adapted from the WHO Disability Assess- and invited all adults aged 50 years and over to ment Schedule (WHODAS) 12-item instrument (24). The participate in the study, a total of 14,958 people. series of questions assessed any difficulties faced by the respondents in performing different daily life activities Instruments due to their health conditions. The responses were This study used the modified and shortened version of collected on the Likert scale and different weights were the INDEPTH WHO-SAGE questionnaire (19), con- assigned to responses from different questions. The total sisting of subjective well-being and QoL, function and score was then inverted to transform it to an index disability, and health status description modules. All the between 0 and 100, with 0 representing extreme problems questionnaires were translated into Bahasa Indonesia and or complete disability and 100 representing a total were pilot-tested during NovemberÁDecember 2006. absence of disability, termed WHODASi. The use of WHODAS in the INDEPTH WHO-SAGE study has Data collection and management been described elsewhere (22). Household visits were conducted by trained surveyors Overall health status was measured using self-reported who administrated the survey questionnaire. Supervisors health derived from eight health domains, including conducted spot-checks and revisits to 5% of the partici- affect, cognition, interpersonal relationships, mobility, pants to ensure the quality of data obtained. All pain, self-care, sleep/energy, and vision (19). Two ques- questionnaires were checked and validated by field tions in each domain, which measured the difficulties supervisors and then sent to the central office in Gadjah faced by the respondents in performing activities, were Mada University, Yogyakarta, for data entry. Data entry put to the respondents and responses were collected using was conducted in D-Entry software and the SAGE data a five-response scale. Item response theory with partial was linked to the surveillance database. Double entry was credit model was used to generate a composite health also conducted on 5% of total questionnaires. Demo- status score. Following each item calibration using chi- graphic variables (such as age, highest level of education squared fit statistics to evaluate its contribution to the completed, marital status, household size and proportion composite health score, the raw composite score was of person over 50 years old within household) and transformed through Rasch modelling into a continuous geographic coordinates of each household were extracted cardinal scale, with 0 representing worst health and from the surveillance database. The SAGE dataset was 100 representing best health (22). The psychometric

80 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 Health of older people in Indonesia

properties of the health score have been evaluated else- Over half of the study participants were women (54%), where (25). and the majority (84%) had less than 6 years of All the three continuous indices (WHOQoL, WHO education. Only 7.2% of the study participants were DASi and overall health status) were later categorised aged 80 years and over. The data showed that 29% of the into quintiles, independently. The three outcome mea- participants were not in a marital relationship but most sures were defined as being in the worst quintiles for of the participants did not live alone. The average number QoL, functioning and disability, and overall health, of household members was 3.5. As the study covered as defined by the three indices, respectively. Socio- all older people in the surveillance area, the house- demographic and economic factors associated with being hold socio-economic quintiles presented in this study in the worst quintile of each health outcome were identified through multivariable logistic regression. The SES quintiles were later regrouped into low (first and Table 1. Background characteristics of respondents and second), middle (third) and high (fourth and fifth) non-respondents among adults aged 50 years and over in quintiles. Educational levels were defined as low (no Purworejo, Indonesia formal education), medium (less than 6 years of educa- tion) and high (at least 6 years of education). As there Respondents Non-respondents was moderate correlation between educational level and Variables (N11,753) (N2,564) socio-economic groups, we combined the educational and Sex, n (%) socio-economic groups into five categories in the analysis. Men 5,420 (46.1) 1,285 (50.1) The regression analysis was performed separately for each Women 6,333 (53.9) 1,278 (49.9) outcome measure. All analyses were conducted using Age, mean (standard 64.1 (9.4) 65.5 (11.5) STATA statistical software version 10.0. deviation) The SAGE data containing individual observations on the health outcomes was transformed into household Age group, n (%) level data, by counting the number of individuals in each 50Á59 years 4,344 (36.9) 928 (36.2) household belonging to the worst quintile of each index. 60Á69 years 4,045 (33.3) 709 (27.7) This household level data was later merged with the 70Á79 years 2,644 (22.7) 595 (23.2) geographical coordinates in the surveillance area. The 80 years and over 720 (7.2) 331 (12.9) purely spatial analysis using Poisson probability model- Education level, n (%) ling was conducted to identify clusters of households with No formal education 3,440 (29.6) 659 (27.4) at least one member being in the worst quintile of QoL, Less than or equal to 6,459 (54.7) 1,257 (52.2) disability and health score, independently. The total 6 years number of people aged 50 years and over in each More than 6 years 1,854 (15.7) 491 (20.4) household was used as the population in the analysis. Monte Carlo hypothesis testing was used with 999 Marital status, n (%) replications and a significance level of 0.05. The risk In current partnership 8,400 (71.0) 1,925 (77.6) estimates for each cluster were identified. The analysis Being single 3,353 (29.0) 556 (22.4) was conducted using SaTScanTM software, version 7.0 Socio-economic quintile, n (%) (26). First quintile 2,394 (20.4) 225 (17.1) The Research Ethics Committee at Gadjah Mada Second quintile 2,317 (19.8) 259 (19.6) University and Purworejo District Health Offices ap- Third quintile 2,390 (20.3) 248 (18.8) proved the SAGE study in Purworejo District, Indonesia. Fourth quintile 2,387 (20.3) 303 (23.0) Documented informed consent was obtained from each Fifth quintile 2,265 (19.2) 285 (21.6) individual prior to the study. Number of household 3.5 (1.7) 3.5 (1.8) Results member, mean (standard deviation) A total of 14,958 individuals aged 50 years and over were visited, with data obtained from 12,459 individuals Proportion of household member aged 50 years and over, n (%) (83%). Cleaned and complete data from 11,753 indivi- B25% 995 (8.6) 324 (12.9) duals were available for analysis. The background char- 25Á49% 3,288 (28.0) 646 (25.6) acteristics of the respondents and the non-respondents 50Á74% 3,853 (32.6) 733 (29.1) (n2,564) were presented in Table 1. Reasons for not ]75% 3,617 (30.9) 818 (32.5) participating in the study included: could not be reached after two visit attempts (81%), refusal (8.3%), died (5%) Note: All figures were weighted to the Purworejo HDSS popula- and out-migration (5.7%). tion in 2007.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 81 Nawi Ng et al.

reflected the quintiles in the whole surveillance popula- Table 2. Distribution of health indices by age-group and sex tion (Table 1). among 11,753 adults aged 50 years and over in Purworejo Table 2 presents summary statistics of three different District, 2007 health indices of WHOQoL, WHODASi and overall health status scores across different age groups and sexes. Indices Men Women Overall, a higher proportion of women aged over 50 years in Purworejo District were categorised in the worst WHO Quality of Life (QoL) score quintiles of health indices as compared to men. These Mean score (95% CI) patterns were observed consistently in all age groups. A 50Á59 years 75.5 (75.3Á75.7) 75.1 (74.9Á75.3) larger discrepancy in functioning and disability and 60Á69 years 74.6 (74.3Á74.8) 73.9 (73.7Á74.1) health status was observed across age groups in men 70Á79 years 73.3 (72.9Á73.6) 72.6 (72.3Á72.9) and women. The differences of QoL index were, however, 80 years and over 71.7 (70.9Á72.4) 71.5 (70.7Á72.3) less prominent across age groups in men and women. The Percentage in the worst quintile (95% CI) results showed that function, QoL and overall health 50Á59 years 11.8 (10.4Á13.2) 14.7 (13.2Á16.1) status decreased substantially among the oldest age- 60Á69 years 17.3 (15.5Á19.1) 22.0 (20.3Á23.7) group, with more than 50% of those over 80 belonging 70Á79 years 25.9 (23.5Á28.4) 32.0 (29.6Á34.4) to the worst function and disability and overall health 80 years and over 37.4 (32.4Á42.3) 42.9 (37.7Á48.1) status quintiles. Being in the older age group, having low education and WHO Disability Assessment Schedule (WHODASi) score being in a low socio-economic group, and not being in a Mean score (95% CI) marital relationship were significantly associated with 50Á59 years 93.2 (92.8Á93.6) 91.2 (90.8Á91.7) higher odds of being in the worst quintiles for QoL, 60Á69 years 88.4 (87.7Á89.0) 84.2 (83.6Á84.9) 70 79 years 81.0 (80.0 81.9) 76.2 (75.2 77.2) functioning and disability, and overall health, respec- Á Á Á 80 years and over 70.9 (68.7 73.1) 66.4 (64.1 68.7) tively. The multivariable analysis showed that respon- Á Á dents aged over 80 years were more than 3.3 times more Percentage in the worst quintile (95% CI) likely to be in the worst quintile of QoL compared to 50Á59 years 5.5 (4.5Á6.5) 8.8 (7.7Á10.0) those aged between 50 and 59 years. They were 12.6 and 60Á69 years 14.3 (12.6Á15.9) 23.4 (21.6Á25.1) 10.6 times more likely to be in the worst functioning and 70Á79 years 28.0 (25.4Á30.5) 40.1 (37.5Á42.6) overall health score quintiles, respectively. The education 80 years and over 52.0 (46.8Á57.1) 59.2 (54.0Á64.3) and socio-economic gradient was also prominent for QoL Overall health score reporting, with individuals in the low SES group who had Mean score (95% CI) a low level of education being 3.4 times more likely be in 50Á59 years 77.3 (76.9Á77.8) 74.7 (74.3Á75.1) the worst quintile of QoL compared to those with high 60Á69 years 73.0 (72.5Á73.5) 69.9 (69.5Á70.3) education in the high SES group (Table 3 and Fig. 1). The 70Á79 years 68.4 (67.9Á69.0) 66.0 (65.6Á66.5) overall effects of low SES and education were less 80 years and over 64.1 (63.2Á65.1) 62.9 (61.9Á63.8) prominent, though statistically significant, for being in the worst disability and overall health status quintiles. Percentage in the worst quintile (95% CI) The spatial analysis revealed clusters of households 50Á59 years 6.0 (5.0Á7.0) 10.6 (9.4Á11.9) with at least one member being in the worst quintile of 60Á69 years 15.6 (13.9Á17.2) 27.2 (25.4Á29.1) QoL, functioning and disability, and overall health, 70Á79 years 30.7 (28.1Á33.3) 43.6 (41.0Á46.1) respectively (Fig. 2). Clusters of households with a 80 years and over 50.4 (45.3Á55.5) 60.8 (55.7Á65.9) member being in the worst quintile of self-reported QoL were identified in the northern part of the district, Note: All figures were weighted to the Purworejo HDSS popula- which is a mainly hilly and mountainous area. This area tion in 2007. is less developed, less urbanised and contains many households categorised in the poorest socio-economic 1,700 inhabitants per km2. Most households in these quintile. In contrast, the clusters of households with areas fall within the richest socio-economic quintile with at least one member being in the worst quintile of the majority of people over 50 having had at least six overall health status were identified in the mid-southern years of education. part of the district, mainly highly populated semi-urban and coastal areas. This part of Purworejo District is Discussion mainly low land covering four main sub-districts of In addition to risks for the oldest old, our study showed Bayan, Banyuurip, Kutoarjo and Purworejo. These are that people with low levels of education and SES the four most populated sub-districts in Purworejo had higher odds of having poorer self-reported QoL District with a population density ranging from 918 to and health. Economic instability during old age may

82 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 Health of older people in Indonesia 2.03) 1.58) 1.66) 1.68) 1.02) 1.10) 1.09) 1.07) 1.76) 6.16) 3.11) 1.66) 12.9) Á Á Á Á Á Á Á Á Á Á Á Á Á 7 (95% CI) Adjusted ORs ghted to the Purworejo HDSS 16.4) 10.6 (8.69 5.06) 1.66 (1.36 2.18) 1.30 (1.07 2.71) 1.37 (1.12 2.79) 1.39 (1.15 0.71)0.98) 0.89 (0.78 1.05 (1.00 0.82) 0.90 (0.73 1.19) 0.81 (0.60 3.06) 1.56 (1.38 7.48) 5.34 (4.63 3.52) 2.73 (2.38 1.85) 1.50 (1.35 Á Á Á Á Á Á Á Á Á Á Á Á Á score quintile as outcome (95% CI) Model 3: Being in the worst health status Unadjusted ORs 2.06) 4.21 (3.50 1.57) 1.81 (1.51 1.77) 2.25 (1.87 1.66) 2.33 (1.95 1.07)1.13) 0.64 (0.57 0.96 (0.93 1.13) 0.73 (0.65 1.02) 1.02 (0.87 1.77) 2.79 (2.55 6.44) 6.55 (5.73 3.17) 3.09 (2.71 1.55) 1.69 (1.55 15.5) 13.6 (11.4 Á Á Á Á Á Á Á Á Á Á Á Á Á (95% CI) Adjusted ORs quintile as outcome 5.03) 1.67 (1.35 2.15) 1.27 (1.04 2.87) 1.44 (1.18 2.79) 1.36 (1.12 0.76)1.01) 0.93 (0.81 1.08 (1.02 0.89) 0.92 (0.75 7.75) 5.54 (4.76 3.55) 2.75 (2.38 1.73) 1.39 (1.25 19.4) 12.6 (10.3 Á Á Á Á Á Á Á Á Á Á 1.25) 0.76 (0.57 3.01) 1.56 (1.38 Á Á Á (95% CI) Model 2: Being in the worst WHODASi Unadjusted ORs 4.11) 4.15 (3.42 3.01) 1.77 (1.46 2.16) 2.36 (1.95 1.68) 2.31 (1.92 1.09) 0.98 (0.95 1.10) 0.68 (0.61 1.27) 0.80 (0.71 1.13) 1.06 (0.9 1.49) 2.74 (2.5 4.01) 16.1 (13.3 2.39) 6.74 (5.86 1.59) 3.09 (2.69 1.26) 1.57 (1.44 Á Á Á Á Á Á Á Á Á Á Á Á Á (95% CI) Adjusted ORs 6.21) 3.35 (2.73 3.41) 2.47 (2.03 2.71) 1.77 (1.44 2.16) 1.37 (1.12 0.99) 1.04 (0.99 0.82) 0.96 (0.84 0.94) 1.05 (0.87 1.05) 0.85 (0.64 5.22) 3.32 (2.75 3.04) 2.09 (1.83 1.83) 1.41 (1.25 1.42) 1.13 (1.02 Á Á Á Á Á Á Á Á Á Á Á Á 2.05) 1.32 (1.16 quintile as outcome Á (95% CI) Model 1: Being in the worst QoL 2.81 (2.32 2.22 (1.82 1.78 (1.46 Unadjusted ORs Three different models in assessing factors associated with poor health indices among 11,753 adults aged 50 years and over in Purworejo District, 200 74% 0.73 (0.65 49% 0.84 (0.75 . Á Á 79 years 2.69 (2.38 59 years69 years 1 1.62 (1.45 1 1 1 1 1 Á Á Á 75% 1 1 1 1 1 1 25% 0.88 (0.74 Low SES, low education 5.11 (4.21 Low SES, middle-high Middle SES, all education High SES, high educationHigh SES, low-middle 1 1 1 1 1 1 50% 25% In current partnership 1 1 1 1 1 1 Being single 1.86 (1.7 80 years and over 4.38 (3.69 70 50 60 Men111111 Women 1.30 (1.19 education levels education ] B Note: WHOQoL, World Healthpopulation Organization in Quality 2007. of Life; WHODASi, World Health Organization Disability Assessment Schedule. All analyses were wei Education and SES Family size 0.96 (0.93 Percentage aged 50 years and over in the household Marital status Age group Sex Table 3 Variables

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 83 Nawi Ng et al.

Fig. 1. The odds ratio for poor health among different education and socio-economic groups among 11,753 adults aged 50 years and over in Purworejo District, 2007. potentially be more of a threat to the urban older than to intergenerational financial transfer does not necessarily their rural counterparts. The majority of older Javanese in influence parent’s labour supply (27). our study were still engaged in agricultural production Family and local community support for older people and were typically more economically productive and is still reliable in rural Java. Only a very small proportion stable compared to their urban counterparts. Our data of older Indonesians receive a pension as their source of reaffirmed the results from the IFLS conducted in 1993 income (about 13% of males and 4% of females in 1985 that showed older Indonesian men and women often with no significant change since then). Those who receive remain economically active; males and younger age a pension are mainly urban dwellers who had worked in groups were more active than women and older age government sectors, the military or industries. Pensions groups. The IFLS data indicated that older men who co- are not paid to urban poor or traditional agricultural resided still worked about 30 hours per week, while those workers (14). The National Social Security Law for poor who did not co-reside worked about 38 hours per week. people, proposed by the government in 2004, has yet to The IFLS data also showed that the availability of be agreed by the legislative body and operationalised by

Cluster of households with at least one member being in the worst quintiles of quality of life Centre: 109.972 °E, 7.640 °S Radius: 13.2 km RR: 1.51 (p<0.001)

Cluster of households with at least one member being in the worst quintiles of functioning Centre: 109.955 °E, 7.741°S Radius: 7.9 km RR: 1.45 (p<0.001)

Cluster of households with at least one member being in the worst quintiles of health score Centre: 109.890 °E, 7.742 °S Radius: 12.4 km RR: 1.40 (p<0.001)

Fig. 2. Spatial distribution of poor health indices among 11,753 adults aged 50 years and over in Purworejo District, 2007.

84 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 Health of older people in Indonesia

the government. This delay has resulted in inefficient and has lower socio-economic development and a higher inappropriate distribution of funds to the needy. The proportions of the population with low levels of educa- Indonesian Government has fostered community parti- tion, who might have poor perceptions of health, well- cipation in the provision of care and economic support being and illnesses, and thus have higher morbidity and to the elderly, mainly to reduce community dependence lower use of health care services. on the insufficient resources provided directly by the The hardworking Javanese population views ‘life as a government. continuous series of misfortunes, calamities and hard- Living arrangements are an important influence on ships which a human being has to experience and to care of older people. Our respondents who were not in endure readily’. The Javanese lead an active life through any marital relationship reported significantly worse QoL constant endeavour (ichtiyar) in activities relating to (OR1.32), worse functioning (OR1.56) and worse agricultural production, economic life and social and health status (OR1.56) compared to those in a marital family matters. Despite their view of life (ichtiyar), the relationship. Traditional Indonesian values mean that Javanese peasant population accepts what comes (nrimo) children are supposed to stay together and to take care of and accepts fate willingly (ingkang nrimah), an attitude their parents, especially when they are no longer econom- which helps them to avoid disappointment or emotional ically productive. However, changes in social values over upset when things go wrong. When discussing the the past few decades have led to increasing migration to burdens of life, the Javanese typically surrender and larger cities for better career opportunities, changes that accept fate (pasrah lan sumarah). Older Javanese people may have affected elder care practices (10). The IFLS in appear to be content to await death, hopefully sur- 1993 showed that 60% of those over 60 years shared a rounded by their children and grandchildren when their house with their children. The data shows that older time comes (29). The Javanese acceptance of fate in their household heads are less likely to co-reside. The pre- life might explain why the majority reported good health dictors for co-residence include households with large and QoL when asked in the study. numbers of children, households where the family head is QoL is one of a number of complex components of currently married, households in urban areas, or areas successful ageing covering life expectancy, life satisfac- with expensive housing. None of these factors were tion, mental and psychological health, cognitive function, identified as significant predictors for the transition to physical health and functioning, income, living condi- co-residence with a child in the follow-up survey in 1997 tions and arrangements, social support and social net- (28). Data from this SAGE survey can be linked to the works. Measuring QoL is also a complex exercise, longitudinal HDSS data to assess how changes in living especially among older people (14). In addition to the arrangements and migration over time could potentially above-mentioned aspects, QoL is also very bound by affect older people’s welfare and well-being in the study culture and may represent different constructs in different population. settings. Results from health status assessments can Our findings show clusters of households with poor usually be used to predict QoL for older populations; self-reported health, functioning and QoL among older however, it is not uncommon to observe discrepancies people. This points to the existence of health inequality in between these two measures (30). the study area and signifies the need to identify factors Service provision for older people, particularly health determining the distributions of poor health outcomes in promotion and social services, is generally lacking in this rural population. Knowledge of the epidemiological Indonesia. Most older people care institutions are based burden of poor health outcomes and their associated in urban areas and there is no alternative care for older factors is an important prerequisite for the government to people in rural settings. Since the mid-1980s the Ministry develop health promotion and intervention programmes of Health in Indonesia has promoted services to older for the older population in Purworejo District and people through ‘the Integrated Health Post Service for throughout Indonesia (6). Elderly People’ programme (Posyandu Lansia) (31). This Results from this study that show clusters of house- is a community-organised health promotion centre at holds with poor self-reported health outcome may poten- village level supervised by staff from the nearest primary tially indicate areas with higher risk of subsequent health care centre. The concept of the Integrated Health mortality. Assessing the future morbidity and mortality Post Service was initially developed to address maternal patterns longitudinally using this SAGE survey as base- and child health issues and later expanded to cover the line data could prove this hypothesis. Results from the ageing population. However, the programme lacks a multivariable analysis are supported by the results from strong health promotion dimension, and puts a lot of the spatial analysis that showed clusters of households focus on the often-inadequate therapeutic aspects of with poor QoL located in the northern part of Purworejo older people’s illnesses. Activities to promote healthy District, which is a largely remote, mountainous area that ageing and healthy life-styles to enhance older people’s is not easily accessed by public transportation. This area well-being are mainly lacking in the programme. As

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 85 Nawi Ng et al.

Indonesia is predicted to have the world’s fastest growing and (c) Farzad Mostashari of the New York City Department of older population during the 1990Á2025 period (14), the Health and Mental Hygiene. government needs to immediately address the issues through policy and action to promote the well-being Conflict of interest and funding and health of its older population. This research has been supported by special grants from the Swedish Council for Social and Work Life Research (FAS), Grant No. 2003-0075. Coordination for preparing Limitations of the study this article has been supported by the Umea˚ Centre for While assessments of self-reported health and QoL Global Health Research, with support from FAS, the have been extensively researched in many countries, Swedish Council for Working Life and Social Research corresponding methodological developments, particu- (Grant No. 2006-1512). larly in low-middle income country settings, are still challenging. Efforts have been undertaken to derive cross-culturally comparable instruments, yet researchers References are still attempting to validate instruments across different settings through development of new validation 1. National Institute on Aging. Why population aging matters: a global perspective. Washington, DC: National Institute on techniques such as vignettes (32, 33). This article has not Aging, National Institutes of Health; 2007. addressed the issue of comparability of our results to 2. World Population Prospects. 2008 Revision [database on the other settings. Further analyses that take the rich data on Internet]. United Nations; 2009. Available from: http://esa. vignettes into account might provide better insights on un.org/unpp/index.asp?panel2 [cited 23 June 2009]. how our data on older people’s well-being can be 3. Population Reference Bureau. 2007 World population data sheet. Washington, DC: Population Reference Bureau; 2007. compared to data from other settings. 4. Byass P. Towards a global agenda on ageing. Global Health We observed that there were considerably more posi- Action 2008, 1:1908. DOI: 10.3402/gha.v1i0.1908. tive responses of data obtained from the Likert-scale 5. United Nations Department of Economic and Social Affairs questions used in the study, and while this might reveal PD. World Population Ageing 2007. New York: United Nations; the true levels of health and QoL in our study population, 2007. 6. Abikusno N. Older population in Indonesia: trends, issues it might also reflect how this rural population valued and policy responses. Bangkok: UNFPA Indonesia and their health and life, regardless of the true levels of their Country Technical Services Team for East and South-East Asia; health and QoL. Good protocols and periodic training of 2007. interviewers hopefully reduced the possibility of social 7. Ng N. Chronic disease risk factors in a transitional country: desirability bias in our study. the case of rural Indonesia. Umea˚: Umea˚ University; 2006. Available from: http://urn.kb.se/resolve?urn=urn:nbn:se:umu: iva-775 [cited 1 July 2009]. Conclusion 8. Heller P. Is Asia prepared for an aging population? Washington, DC: International Monetary Fund; 2006. Being female, old, unmarried and having a low education 9. Schroder-Butterfill E. Inter-generational family support pro- and socio-economic level are significant predictors of vided by older people in Indonesia. Ageing Soc 2004; 24: self-reported poor QoL and health status, and disability 497Á530. among older people in Purworejo District. This study 10. Keasberry IN. Elder care and intergenerational relationships in rural Yogyakarta, Indonesia. Ageing Soc 2001; 21: 641Á65. shows the existence of geographic pockets of vulnerable 11. Niehof A. Ageing and the elderly in Indonesia: identifying key older people in Purworejo District, and emphasises the issues. Bijdragen tot de Taal-, Land-en Volkenkunde 1995; 151: need to take immediate action to address issues on older 422Á37. people’s health and QoL. Lack of care and services for 12. Van Eeuwijk P. Old-age vulnerability, ill-health and care support older people has to be addressed, and the Indonesian in urban areas of Indonesia. Ageing Soc 2006; 26: 61Á80. 13. Kreager P, Schroder-Butterfill E. Indonesia against the trend? health system, through its Posyandu Lansia, should Ageing and inter-generational wealth flows in two Indonesian increase the balance of ‘curing sick older people’ and communities. Demogr Res 2008; 19: 1781Á810. ‘caring for healthy older people and promoting their 14. Phillips DR. Ageing in the Asia-Pacific region. London and health and well-being’. New York: Routledge; 2005. 15. Jylha M. What is self-rated health and why does it predict mortality? Towards a unified conceptual model. Soc Sci Med 2009; 69: 307Á16. Acknowledgements 16. Idler EL, Benyamini Y. Self-rated health and mortality: a review of twenty-seven community studies. J Health Soc Behav 1997; The authors would like to acknowledge Dr. Somnath Chatterji, Dr. 38: 21Á37. Paul Kowal, and Ms. Nirmala Naidoo of WHO and the INDEPTH 17. Frankenberg E, Jones NR. Self-rated health and mortality: does Adult Health working group for their support in data analysis and the relationship extend to a low income setting? J Health Soc interpretation of the data. SaTScanTM is a trademark of Martin Behav 2004; 45: 441Á52. Kulldorff. The SaTScanTM software was developed under the joint 18. Schroder-Butterfill E, Marianti R. A framework for under- auspices of (a) Martin Kulldorff, (b) the National Cancer Institute standing old-age vulnerabilities. Ageing Soc 2006; 26: 9Á35.

86 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 Health of older people in Indonesia

19. World Health Organization. WHO Study on global AGEing arrangements. Melbourne: Department of Economics, Univer- and adult health (SAGE). Geneva: World Health Organization; sity of Melbourne; 2001. 2009. Available from: http://www.who.int/healthinfo/systems/ 28. Frankenberg E, Beard V. The Kindred Spirit: the ties that bind sage/en/index.html [cited 1 July 2009]. Indonesian children and their parents. Southeast Asian J Soc Sci 20. Wilopo SA. Key issues on the research design, data collection, 1999; 27: 65Á86. and management in Community Health and Nutrition Research 29. Koentjaraningrat. Javanese Culture. Oxford: Oxford University Laboratory (CHN-RL) Purworejo Demographic Surveillance Press; 1985. System (DSS). Jogjakarta: CHN-RL Faculty of Medicine, 30. Covinsky KE, Wu AW, Landefeld CS, Connors AF Jr, Phillips Gadjah Mada University; 1997. RS, Tsevat J, et al. Health status versus quality of life in older 21. Vyas S, Kumaranayake L. Constructing socio-economic status patients: does the distinction matter? Am J Med 1999; 106: indices: how to use principal component analysis. Health Policy 435Á40. 2006: 21: 459Á68. 31. Ministry of Health Republic of Indonesia. Pedoman pembinaan 22. Kowal P, Kahn K, Ng N, Naidoo N, Abdullah S, Bawah A, kesehatan usia lanjut bagi petugas kesehatan [Health officer’s et al. Ageing and adult health status in eight lower-income guideline of health care for older people]. Jakarta: Ministry of countries: the INDEPTH WHO-SAGE collaboration. Global Health Republic of Indonesia; 2000. Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0. 32. Salomon JA, Tandon A, Murray CJ, Groups World Health 5302. Survey Pilot Study Collaborating Groups. Unpacking health 23. World Health Organization. WHO Quality of life BREF perceptions using anchoring vignettes. In: Murray CJ, Evans (WHOQOL-BREF). Geneva: World Health Organization; DB, eds., Health systems performance assessment: debates, 2009. Available from: http://www.who.int/substance_abuse/re methods and empiricism. Geneva: World Health Organization; search_tools/whoqolbref/en/ [cited 31 May 2009]. 2003: 401Á7. 24. World Health Organization. WHO Disability Assessment 33. King G, Murray CJ, Salomon JA, Tandon A. Enhancing the Schedule II (WHO DAS II); 2009. Available from: http:// validity and cross-cultural comparability of measurement in www.who.int/classifications/icf/whodasii/en/index.html [cited 31 survey research. Am Political Sci Rev 2004; 98: 191Á207. March 2009]. 25. Moussavi S, Chatterji S, Verdes E, Tandon A, Patel V, Ustun B. *Nawi Ng Depression, chronic diseases, and decrements in health: results Department of Public Health and Clinical Medicine from the World Health Surveys. Lancet 2007; 370: 851Á8. Centre for Global Health Research, Epidemiology and Global Health 26. Kulldorff M. A spatial scan statistic. Commun Stat Theory Umea˚ University, SE-901 85 Umea˚ , Sweden Methods 1997; 26: 1481Á96. Tel: 46 90 7851391 27. Cameron LA, Cobb-Clark DA. Old-age support in developing Fax: 46 90 138977 countries: labor supply. Intergenerational transfers and living Email: [email protected]

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2125 87 æINDEPTH WHO-SAGE Supplement Social gradients in self-reported health and well-being among adults aged 50 and over in Pune District, India Siddhivinayak Hirve1,2*, Sanjay Juvekar1,2, Pallavi Lele1,2 and Dhiraj Agarwal1,2

1Vadu Rural Health Program, KEM Hospital Research Center, Pune, Maharashtra, India; 2INDEPTH Network, Accra, Ghana

Background: India’s older population is projected to increase up to 96 million by 2011 with older people accounting for 18% of its population by 2051. The Study on Global Ageing and Adult Health aims to improve empirical understanding of health and well-being of older adults in developing countries. Objectives: To examine age and socio-economic changes on a range of key domains in self-reported health and well-being amongst older adults. Design: A cross-sectional survey of 5,430 adults aged 50 and over using a shortened version of the SAGE questionnaire to assess self-reported assessments (scales of 1Á5) of performance, function, disability, quality of life and well-being. Self-reported responses were calibrated using anchoring vignettes in eight key domains of mobility, self-care, pain, cognition, interpersonal relationships, sleep/energy, affect, and vision. WHO Disability Assessment Schedule Index and WHO health scores were calculated to examine for associations with socio-demographic variables. Results: Disability in all domains increased with increasing age and decreasing levels of education. Females and the oldest old without a living spouse reported poorer health status and greater disability across all domains. Performance and functionality self-reports were similar across all SES quintiles. Self-reports on quality of life were not significantly influenced by socio-demographic variables. Discussion: The study provides standardised and comparable self-rated health data using anchoring vignettes in an older population. Though expectations of good health, function and performance decrease with age, self-reports of disability severity significantly increased with age, more so if female, if uneducated and living without a spouse. However, the presence or absence of spouse did not significantly alter quality of life self- reports, suggesting a possible protective effect provided by traditional joint family structures in India, where older people are social if not financial assets for their children. Keywords: ageing; self-reported health; well-being; quality of life; INDEPTH WHO-SAGE

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘SAGE data’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 3 November 2009; Revised: 9 May 2010; Accepted: 8 July 2010; Published: 27 September 2010

ndia’s population is rapidly moving towards an older age. The 2001 Census accounts for 7.5% of the popula- age structure consequent on declining mortality and tion being aged 60 years and over i.e. more than 76 Ihigh fertility in the twentieth century, followed by a million, a sharp increase from 25 million (5.63%) in 1961; rapid decline in fertility and access to better health care in 33 million (6%) in 1971; 43 million (6.49%) in 1981; and recent times as successively larger cohorts step into old 57 million (6.76%) in 1991 (1). Life expectancy at birth

Global Health Action 2010. # 2010 Siddhivinayak Hirve et al. This is an Open Access article distributed under the terms of the Creative Commons 88 Attribution-Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2128 Social gradients in self-reported health status

has likewise increased from 50.5 in 1971 to 60.8 years for related to impending morbidity and mortality. However, males (49 to 62.5 years for females) in 2001. Kerala and health valuation is multi-faceted and influenced not only Maharashtra amongst others have taken the lead in by disease experience and disease perception but also by ushering in this demographic transition in India (2). health expectations which in turn are influenced by the This demographic evolution seen in recent decades has socio-cultural context of the individual (15). Conse- major consequences on the economy, disease burden quently, there arises a need to standardise the ways in facing society and important implications for govern- which individuals report their health status, as people mental health, economic and social policies such as from varying socio-cultural backgrounds may rate their health care for the elderly, retirement benefits, old age health differently. As self-assessments of health play an homes, food and personal security, economic growth, etc. increasing role in measurement of health outcomes, an Elderly women face a double burden Á not only because approach using ‘anchoring vignettes’ can improve the of their advancing age and the prevailing societal gender utility of SRH by addressing issues of comparability differential, but also because they survive without their amongst individuals and populations. life partners (approximately 50% widows amongst elderly The Study on Global Ageing and Adult Health women compared to 15% widowers amongst elderly (SAGE) aims to improve the empirical understanding men). of health and well-being of older adults, and ageing, in India’s older population is projected to increase to 96 developing countries. This paper explores the socio- million (8.2%) by the next census in 2011, with older demographic gradients of older people’s health with a people accounting for approximately 18% of its popula- focus on physical performance and function, using the tion by 2051. This calls for a shift from demographically short SAGE version implemented at the Vadu, India, based programmes and policies to economically oriented Health and Demographic Surveillance System (HDSS). policies and programmes which would take care of the economic, health and social security and quality of life concerns of older people, so that they can lead a dignified Methods life in their closing years without adding to the millions below the poverty line (3). Study area and study sample The Madrid International Plan of Action on Aging The SAGE is designed as a longitudinal data platform in 2002 prioritises Advancing Health and Well-being into six countries including India, based on methodological old age as a central theme. There is not enough evidence advances created by the WHO’s World Health Survey to say whether longer life expectancy is accompanied by programme (16). The shortened version of SAGE has been implemented by the INDEPTH Network in eight of improved health or simply more years of poor health Á especially in the context of changing familial norms its member DSS sites (Agincourt in South Africa, Ifakara in Tanzania, Nairobi in Kenya, Navrongo in Ghana, towards small families and altered social and personal Filabavi in Viet Nam, Matlab in Bangladesh, Purworejo support systems (4). in Indonesia and Vadu in India), each site having an Of the different patterns of living among older people initial enrolment target of 5,000 adults (except the urban such as living with a spouse, or with children or in old age slum-based site of Nairobi with a target of 2,000) aged 50 homes, living alone or with a spouse tends to be most and over. Of these, Agincourt, Navrongo and Vadu stable for those aged 65 years and over, whereas living implemented both the shorter and longer version to with a child or grandchild is the most stable living complement the national SAGE implementation in their arrangement for the oldest old (5). Financial dependence respective countries. The Vadu HDSS monitors demo- has increased, leisure time and social cohesion have graphic trends in its population of some 80,000 people decreased, and life styles have changed for older people spread over 22 villages in Pune district in Maharashtra, with a gradual breakdown of the traditional joint family India. The SAGE short version was administered in 2007 system (6Á8). by trained graduate field-based researchers, to a ran- Ageing research in India has focused mostly on disease domly selected sample of 6,000 individuals aged 50 and states and risk factors. Evidence on elderly health, over. physical performance and disability is limited to under- standing the psycho-social or socio-behavioural risk factors (9Á13). There is a shift from the traditional SAGE tool assessment of health based on risk factors, mortality The SAGE tool has been adapted from the WHO’s World and utilisation of health care services to an assessment Health Survey implemented in 70 countries, from 16 that focuses on functioning and disability in multiple other cross-sectional and longitudinal studies on ageing health and related domains of daily life (14). Self-rated including the US Health and Retirement Study and health (SRH) has often been used in large survey settings English Longitudinal Study on Ageing, and cognitive to rapidly assess health status, and has been shown to be testing of the draft tool in South Africa and Viet Nam in

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2128 89 Siddhivinayak Hirve et al.

2004. The resulting SAGE tool was piloted in India, expectations for health Á in the same way that the self- Ghana and Tanzania in 2005. ratings do for the respondent’s own levels of health (17). The long SAGE tool comprises three main question- The average score for each health domain for each naires: household, individual and proxy (http:// respondent was calculated. As an example, if the respon- www.who.int/healthinfo/systems/sage/en/index1.html). dent had mild difficulty in washing/bathing or dressing The household questionnaire includes the household (score of 2) and no difficulty in taking care of or roster, and details of housing, family support networks maintaining general appearance (score of 1) and mild and transfers, household assets and income and household difficulty in staying by himself for a few days (score of 2), expenditure. The individual questionnaire includes socio- then the average score for the respondent for the self-care demographic characteristics, and information on work domain was calculated as 1.67. Though the self-assessment history and benefits, health state descriptions, anthropo- ratings were categorical, the summary score average metry, physical and cognitive performance tests and becomes a continuous variable with a narrow range from biomarkers, risk factors and preventive health behaviours, 1 to 5. As a result, most of these average scores did not have chronic conditions and health service coverage, health care normal distributions and hence the average summary score utilisation, social cohesion, subjective well-being and was re-coded as categorical (1 to 5) with cut points 0Á1, 1.1 2, 2.1 3, 3.1 4 and 4.1 5. quality of life, and impact of care giving on older people. Á Á Á Á A mean WHO Quality of Life score was calculated The proxy questionnaire was for a proxy respondent if the based on eight self-assessment ratings addressing satis- interviewer felt that the subject selected did not faction with various health domains. The mean WHO- possess the cognitive ability to complete the individual QoL score ranges from 1 to 5 (where 5 indicates poor questionnaire. satisfaction with quality of life) and this was transformed The shortened version of the SAGE instrument, used into a 0 to 100 scale, in which a higher score indicates a for this study, includes only the salient two to four self- higher quality of life. assessment ratings per domain from the longer SAGE A WHO Disability Assessment Schedule (WHODAS) tool, and covers eight different health domains of index was calculated based on standard weights applied mobility, self-care, pain and discomfort, cognition, parti- to 12 self-assessment ratings of limitations of function in cipation in interpersonal activities, sleep/energy, affect various health domains. The index ranges from 0 to 100 and vision. The shortened version comprises three main (where 100 indicates extreme disability), and was then sections Á the first section is a self-assessment of health inverted into a score designated WHODASi, with a range state descriptions, function and disability in these eight from 0 to 100 in which a higher score indicated a higher domains supplemented by summary self-assessment rat- functional ability. ings of overall health and function. The second section is Health status scores were derived using Item Response a self-assessment of overall subjective well-being and Theory (IRT) parameter estimates in Winsteps, a Rasch quality of life. The third section includes four sets of 20 measurement software. IRT uses Maximum Likelihood vignettes each, applied in rotation to different respon- Estimation (MLE) which combines the pattern of re- dents. Each vignette set covers two of the eight health sponses as well as the characteristics of each specific item domains; with five vignettes for each domain question. for the multiple health questions (each with multiple For each self-assessment question, the respondent is response categories) to produce the final health scores asked to rate his/her own health, function and disability (18).The health status score was then transformed to a on a 5-point categorical scale (1 to 5) where the score 1 scale of 0 to 100, with higher scores representing better denotes the best health (categories range from very good health status. to very bad) or least difficulty in a function or the least These three 0 to 100 scores thus represent different disability (categories range from none to extreme diffi- aspects of self-reported health, but all follow a 0 to 100 culty or cannot do). The SRH measurement is supple- scale in which higher scores represent better outcomes. mented by age, sex, education, socio-economic status The distribution of self-reported responses to each of quintiles, and marital status information collected on all the health domains was compared across age groups, sex, individuals every 6 months as part of routine demo- marital status, socio-economic status quintiles and edu- graphic surveillance in the Vadu HDSS site. cational levels for significant differences between the The anchoring vignette serves to describe a concrete lowest and highest categories of the function and level in a given health domain that the respondent performance-rating variables (Kolmogorov Smirnov evaluates using the same question and response categories equality of distribution test). used for self-assessment on that domain. The vignettes are ‘fixed’ across all respondents so that any variation in Results self-assessment can be attributed to differences in re- We analysed data on 5,430 individuals aged 50 and over, sponse category cut-points that reflect the respondent’s with adequate cognitive ability to complete the survey,

90 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2128 Social gradients in self-reported health status

Table 1. Socio-demographic profile of 5,475 adults aged 50 and over in Vadu, India

Males (n2,850) Females (n2,625) Test of significance

51.6% 48.4% NS Mean age (SD) years 63.1 (8.9) 62.5 (8.9) NS

Age group (years) 50Á59 (%) 39.5 39.7 NS 60Á69 (%) 36.1 38.9 70Á79 (%) 19.1 16.5 80 and over 5.1 4.8

Education x2 632.8 No formal education (%) 36.9 8.1 pB0.001 56 years (%) 55.6 79.6 6 years (%) 7.5 12.2

Marital status Now single (%) 9.5 35 pB0.001

Socio-economic status Poorest quintile (%) 10.5 12.7 NS Second quintile (%) 15.6 15.1 Third quintile (%) 21.2 22.7 Fourth quintile (%) 22.3 19.7 Least poor quintile (%) 30.2 29.6 Mean number of household members (SD) 6.9 (3.5) 6.8 (3.6) NS Mean number of people aged 50 years and over in household (SD) 1.77 (0.78) 1.77 (0.78) NS and who could be linked to the DSS database. Table 1 both the mobility questions, thus validating the use gives the socio-demographic profile of subjects. As of anchoring vignettes for comparison of self-ratings of expected, older women were significantly less educated mobility between individuals. The average self-rating of than younger ones. A significantly higher proportion of mobility by older adults aged 50 and over lay somewhere older women were widows (35%) compared to older men between the level of mobility described by vignette 1 (‘xxx as widowers (9.5%). There was no significant sex has no problems with walking, running or using her differential for any other socio-demographic variable. hands, arms and legs. S/he jogs 4 kms twice a week’) and Fig. 1 shows an example comparing self-ratings with 2(‘XXX is able to walk distances of up to 200 metres anchored vignettes (ordered in increasing levels of without any problems but feels tired after walking 1 km difficulty) for two mobility questions. There was good or climbing up more than one flight of stairs. S/he has no response consistency in ratings of the five vignettes used problem with day-to-day physical activities such as for describing different levels of difficulty in mobility for carrying food from the market’).

Fig. 1. Self-assessments and vignette ratings for two mobility questions among 5,475 adults aged 50 and over in Vadu, India.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2128 91 Siddhivinayak Hirve et al.

Fig. 2. Age differentials in self-ratings in different health domains among 5,475 adults aged 50 and over in Vadu, India.

As age increased, self-ratings of difficulty with mobility older people without any formal education were signifi- increased significantly (Fig. 2). Increasing difficulty in cantly more likely (70%) to self-rate their health status as ratings for self-care, pain, cognition, sleep and vision were poor compared to their more educated contemporaries; also seen as age increased. However, there was no and older people without a spouse were marginally more statistically significant change in participation in inter- likely to rate poor health status compared to those living personal activities or affect with increasing age. with their spouse (Table 3). Socio-economic status did Older males rated higher levels of difficulty in perform- not appear to influence self-reports of health. ing functions and tasks in all health domains compared to Self-reported quality of life was not significantly older females. Similar statistically significant trends were influenced by age, sex or education. The elderly popula- seen for older men or women who had lost their spouse tion belonging to the lowest SES quintiles, as well those compared to their married contemporaries (Fig. 3), with without a living spouse, rate poorer quality of life than the single elderly female widow rating the most difficulty their better off counterparts and those with a living in performing tasks in any of the health domains. spouse Education was directly related to function in all health domains. At lower educational levels, the self-ratings for Discussion difficulty in performing functions in all health domains The 20th century challenged us with population growth Á were higher (Fig. 4). Self-ratings of function and dis- the 21st century challenge is to cope with ageing. India is ability were similar across all quintiles of socio-economic home to one of the world’s largest populations which is status. ageing rapidly.It is projected that by 2030 about 45% of the Table 2 shows that males self-reported significantly less health burden in India, largely due to non-communicable disability, and significantly better overall health than diseases, will be borne by the older adults (19). To cope their female contemporaries. However, there was no with an ageing India, policy makers need to be informed significant difference in self-reported quality of life across with evidence on interrelated domains including work and age groups and sex. retirement benefits, private wealth and income security, the Multivariate analysis showed that males self-reported implications of family and societal level transfer systems, better health status compared to females; self-reports of health and well-being of the ageing population. As poorer health status increased as people became older; populations age, the social and economic demands on

92 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2128 Social gradients in self-reported health status

Fig. 3. Differentials in self-ratings of health domains by marital status among 5,475 adults aged 50 and over in Vadu, India.

Fig. 4. Differentials in self-ratings of health domains by education level among 5,475 adults aged 50 and over in Vadu, India.

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2128 93 Siddhivinayak Hirve et al.

Table 2. Age and sex differentials in health, disability and Table 3. Factors associated with self-rated poor health and quality of life outcomes for 5,475 adults aged 50 and over in quality of lifea for 5,475 adults aged 50 and over in Vadu, Vadu, India India

Males Females Poor quality of life Poor health (n2,850) (n2,625) p-value OR (95% CI) OR (95% CI)

Mean WHODASi score (SD) Sex 50Á59 years 80.0 (13.1) 77.4 (13.4) B0.001 Males 1.07 (0.93Á1.22) 0.73 (0.64Á0.83) 60Á69 years 78.3 (13.8) 75.4 (13.5) B0.001 Females 1 1 70Á79 years 75.4 (14.0) 72.9 (14.1) 0.006 Age 80 years and over 74.9 (15.2) 70.0 (17.7) 0.01 50Á59 years 1 1 Mean health status score (SD) 60Á69 years 1.01 (0.87Á1.17) 1.18 (1.03Á1.35) 50Á59 years 69.8 (11.2) 67.3 (9.7) B0.001 70Á79 years 1.13 (0.95Á1.36) 1.53 (1.29Á1.83) 60Á69 years 67.8 (9.8) 66.0 (8.7) B0.001 80 years and over 1.05 (0.78Á1.41) 1.78 (1.32Á2.39) 70Á79 years 65.9 (9.0) 64.6 (8.5) 0.025 Education 80 years and over 65.9 (9.8) 62.6 (8.9) 0.003 No formal education 1.04 (0.77Á1.44) 1.7 (1.27Á2.26) Mean WHOQoL score (SD) 56 years 1.22 (1.03Á1.44) 1.39 (1.19Á1.63) 50Á59 years 75.3 (4.5) 74.8 (4.5) 0.02 6 years 1 1 60Á69 years 74.8 (4.7) 74.5 (4.5) 0.09 Marital status 70Á79 years 74.1 (5.0) 74.1 (5.2) NS Now single 1.19 (1.01Á1.41) 1.05 (0.89Á1.24) 80 years and over 74.7 (5.4) 73.3 (6.1) 0.049 Currently in partnership 1 1

Socio-economic status First quintile 1.56 (1.25 1.95) 1.05 (0.85 1.31) families, communities and nations will grow with implica- Á Á Second quintile 1.41 (1.16Á1.71) 1.36 (1.12Á1.64) tions for the formal and informal social and medical care Third quintile 1.18 (0.99Á1.41 1.10 (0.93Á1.30) systems (20). Well-being, a person’s perceived level of Fourth quintile 1.07 (0.9Á1.28) 0.85 (0.72Á1.01) satisfaction with his work, his marriage, his health and life Fifth quintile 1 1 as a whole, though hard to measure, continues to be a good measure of success of governmental programmes and aLogistic model controlling for family size. policies. The mobility domain vignette example introduced the concept of vignettes to anchor self-ratings of health in the populations across different socio-demographic groups, mobility domain to a concrete level of function or or within populations over time. disability. Vignettes have been used in the social sciences This paper underlines the importance of socio-demo- since the 1950s (21) and more recently in health and graphic factors as predictors which influence SRH in medicine (22, 23). The difference is that we use vignettes various health domains. Despite lowered expectations of as scale anchors rather than as random variants of the function and performance, the self-reports of disability same vignette. This means that a vignette describes the significantly increased with age (biological influence) as same level of function or health status to all respondents. well as environment (socio-cultural influence). The older Also, the vignette is anchored to the self-rating through woman, though with a longer life expectancy compared the use of identical questions and response categories. to her male contemporary, is disadvantaged on multiple The underlying assumption for anchoring vignettes is fronts Á due to her advancing age; due to societal norms that of response consistency (i.e. a person evaluates a of being a woman which limit her mobility and function; hypothetical level of health in the same way s/he would due to her being less educated, less empowered. This self-assess his/her own health) and vignette equivalence inability to perform and function and the consequent (i.e. the level described by a vignette is understood deleterious effect on health, are compounded if the older similarly by individuals independent of age, sex, educa- woman loses her spouse at an early age. The presence or tion or any other characteristic). Hence, the primary absence of the spouse of an older person significantly purpose of anchoring vignettes linked to self-assessments altered self-reports on health and quality of life. The lack is to detect and adjust for differences in response category of significant associations between age, sex, education cut-points so as to make categorical self-reports more and quality of life, seen otherwise with health, needs comparable. This approach allows for studying differ- further study to understand the linkages between health ences in categorical cut-points between and within and quality of life in its various dimensions.

94 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2128 Social gradients in self-reported health status

Late life outcomes are wide ranging. Old age experi- 9. Ushasree SS. Perceptions of social supports among the aged ence is very different for those who are financially secure women. J Indian Acad Appl Psychol 2000; 26: 115Á8. 10. Bagchi K. Healthy ageing, health and population. Soc Gerontol and educated than for those who are poor and unedu- Res 2000; 23: 11Á6. cated; those who are healthy than those who are ill; and 11. Subramanian T, Uenkatarao T, Ramakrishnan R, Suresh those who find themselves alone than those who are Kumar SK, Gupte MD. Study on psycho-social aspects of embedded in strong social networks. Understanding geriatric people in a rural area. Indian J Prev Soc Med 1999; 30: health, disability and well-being in later life has wide 66Á73. 12. Yadava KNS, Yadava SS, Roberts RE. Ageing and health implications for informing policy as India matures hazards in rural northern India. Health and Popul 1996; 19: demographically. 1Á18. 13. Alam M, Mukherjee M. Ageing, activities of daily living, disabilities and the need for public health initiatives: some Acknowledgements evidence from a household survey in Delhi. Asia Pac Popul J 2005; 20: 47Á77. The study has been supported by the INDEPTH Network, through 14. Fillenbaum GG. Social context and self-assessments of health a supplemental grant to the World Health Organization, Geneva, by among the Elderly Duke University Medical Center. J Health the National Institute on Aging, USA. The authors acknowledge the Soc Behav 1979; 20: 45Á51. role of Stephen Tollman, Somnath Chatterjee and Paul Kowal in 15. Salomon JA, Mathers CD, Chatterji S, Sadana R, .U¨ stu¨nTB, leading this INDEPTH WHO-SAGE initiative and to Nawi Ng and Murray CJL. Quantifying individual levels of health: defini- Kathy Kahn for coordinating efforts for a concerted publication. tions, concepts and measurement issues. In: Murray CJL, Evans Thanks are due to Nirmala Naidoo for statistical support in DB, eds. Health systems performance assessment: debates, estimating IRT health scores. Finally, the authors thank the Vadu methods and empiricism. Geneva: World Health Organization; DSS field-based staff for their quality work and the older subjects of 2003, pp. 301Á18. Vadu who willingly consented to the study. 16. Ustun TB, Chatterji S, Villaneuva M, Bendib L, Celik C, Sadana R, et al. WHO Multi-country survey study on health Conflict of interest and funding and responsiveness 2000Á2001. Geneva: World Health Organi- The authors have not received any funding or benefits zation; 2001. (GPE discussion paper 37) 17. Salomon JA, Tandon A, Murray CJL. Comparability of self from industry to conduct this study. rated health: cross sectional multi-country survey using anchor- ing vignettes. BMJ 2004; 328: 258. References 18. Reeve BB, Hays RD, Chang CH, Perfetto EM. Applying item response theory to enhance health outcomes assessment. Qual Life Res 2007; 16: 1Á3. 1. Rajan SI. Population ageing and health in India. Mumbai: 19. Chatterji S, Kowal P, Mathers C, Naidoo N, Verdes E, Smith JP, Centre for Enquiry into Health and Allied Themes [CEHAT]; et al. The health of aging populations in China and India. 2006. p. 42. Available from: http://www.cehat.org/humanrights/ Health Aff 2008; 27: 1052Á63. rajan.pdf [cited 1 May 2010]. 20. Panel on a Research Agenda and New Data for an Aging 2. Swain P, Sherin Raj TP. Demography of ageing in India: state World, Committee on Population and Committee on National and district level analysis. [Unpublished]. Presented at the Statistics, Division of Behavioral and Social Sciences and international seminar on demographic changes and implica- Education, National Research Council. Preparing for an aging tions, Department of Demography, University of Kerala, world: The case for cross-national research. Washington, DC: Trivandrum, India, 7Á9 December 2004, p. 12. National Academy Press. pp. 1Á14. 3. Bose A. Beyond demography of ageing. [Unpublished]. Pre- 21. Herskovits MJ. The hypothetical situation: a technique of field sented at the international seminar on demographic changes and research. Southwest J Anthropol 1950; 6: 32Á40. implications, Department of Demography, University of Kerala, 22. Koedoot CG, De Haes JC, Heisterkamp SH, Bakker PJ, De Trivandrum, India, 7Á9 December 2004, p. 9. Graeff A, De Haan RJ. Palliative chemotherapy or watchful 4. Kinsella K. Global aging: the challenge of success. Popul Bull waiting? A vignettes study among oncologists. J Clin Oncol 2005; 60: 34Á6. 2002; 20: 3658Á64. 5. Wilmoth JM. Living arrangement transitions among America’s 23. Hughes R, Huby M. The application of vignettes in social and older adults. Gerontologist 1998; 38: 434Á44. nursing research. J Adv Nurs 2002; 37: 382Á6. 6. Elango S. A study of health and health related social problems in the geriatric population in a rural area in Tamil Nadu. Indian J Pub Health 1998; 42: 7Á8. *Siddhivinayak Hirve 7. Goel PK, Garg SK, Singh JV, Bhatnagar M, Chopra H, Bajpai KEM Hospital Research Centre SK. Unmet needs of the elderly in a rural population of Meerut. Rasta Peth, Pune 411011 Indian J Comm Med 2003; 28: 165Á6. Maharashtra, India 8. Kishore S, Garg BS. Socio-medical problems of aged population Tel: 91 20 66037336 in a rural area of Wardha District. Indian J Pub Health 1997; Fax: 91 20 26125603 41: 43Á8. Email: [email protected]

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.2128 95 æINDEPTH WHO-SAGE Supplement Health inequalities among older men and women in Africa and Asia: evidence from eight Health and Demographic Surveillance System sites in the INDEPTH WHO-SAGE study Nawi Ng1,2,3*#, Paul Kowal4,5, Kathleen Kahn1,2,6#, Nirmala Naidoo4, Salim Abdullah2,7, Ayaga Bawah2, Fred Binka2, Nguyen T.K. Chuc2,8, Cornelius Debpuur2,9, Thaddeus Egondi2,10, F. Xavier Go´ mez-Olive´ 2,6, Mohammad Hakimi2,3, Siddhivinayak Hirve2,11, Abraham Hodgson2,9, Sanjay Juvekar2,11, Catherine Kyobutungi2,10, Hoang Van Minh2,8, Mathew A. Mwanyangala2,7, Rose Nathan2,7, Abdur Razzaque2,12, Osman Sankoh2, P. Kim Streatfield2,12, Margaret Thorogood2,13, Stig Wall1#, Siswanto Wilopo2,3, Peter Byass1#, Stephen M. Tollman1,2,6# and Somnath Chatterji4

1Department of Public Health and Clinical Medicine, Centre for Global Health Research, Epidemiology and Global Health, Umea˚ University, Umea˚ , Sweden; 2INDEPTH Network, Accra, Ghana; 3Purworejo HDSS, Faculty of Medicine, Gadjah Mada University, Yogyakarta, Indonesia; 4World Health Organization, Multi-Country Studies Unit, Geneva, Switzerland; 5University of Newcastle Research Centre on Gender, Health and Ageing, Newcastle, NSW, Australia; 6MRC/Wits Rural Public Health and Health Transitions Research Unit (Agincourt), School of Public Health, University of the Witwatersrand, Johannesburg, South Africa; 7Ifakara Health Institute, Ifakara, Morogoro, Tanzania; 8FilaBavi HDSS, Faculty of Public Health, Hanoi Medical University, Hanoi, Viet Nam; 9Navrongo HDSS, Navrongo, Ghana; 10African Population & Health Research Center, Nairobi, Kenya; 11Vadu Rural Health Programme, KEM Hospital Research Centre, Pune, Maharashtra, India; 12Matlab HDSS, ICDDR,B, Dhaka, Bangladesh; 13Warwick Medical School, University of Warwick, Coventry, UK

Background: Declining rates of fertility and mortality are driving demographic transition in all regions of the world, leading to global population ageing and consequently changing patterns of global morbidity and mortality. Understanding sex-related health differences, recognising groups at risk of poor health and identifying determinants of poor health are therefore very important for both improving health trajectories and planning for the health needs of ageing populations. Objectives: To determine the extent to which demographic and socio-economic factors impact upon measures of health in older populations in Africa and Asia; to examine sex differences in health and further explain how these differences can be attributed to demographic and socio-economic determinants. Methods: A total of 46,269 individuals aged 50 years and over in eight Health and Demographic Surveillance System (HDSS) sites within the INDEPTH Network were studied during 2006Á2007 using an abbreviated version of the WHO Study on global AGEing and adult health (SAGE) Wave I instrument.

#Editor, Nawi Ng, Supplement Editor, Kathleen Kahn, Chief Editor, Stig Wall, Deputy Editor, Peter Byass, Supplement Editor, Stephen M. Tollman, have not participated in the review and decision process for this paper.

Global Health Action 2010. # 2010 Nawi Ng et al. This is an Open Access article distributed under the terms of the Creative Commons Attribution- 96 Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 Health inequalities among older men and women in Africa and Asia

The survey data were then linked to longitudinal HDSS background information. A health score was calculated based on self-reported health derived from eight health domains. Multivariable regression and post-regression decomposition provide ways of measuring and explaining the health score gap between men and women. Results: Older men have better self-reported health than older women. Differences in household socio- economic levels, age, education levels, marital status and living arrangements explained from about 82% and 71% of the gaps in health score observed between men and women in South Africa and Kenya, respectively, to almost nothing in Bangladesh. Different health domains contributed differently to the overall health scores for men and women in each country. Conclusion: This study confirmed the existence of sex differences in self-reported health in low- and middle- income countries even after adjustments for differences in demographic and socio-economic factors. A decomposition analysis suggested that sex differences in health differed across the HDSS sites, with the greatest level of inequality found in Bangladesh. The analysis showed considerable variation in how differences in socio-demographic and economic characteristics explained the gaps in self-reported health observed between older men and women in African and Asian settings. The overall health score was a robust indicator of health, with two domains, pain and sleep/energy, contributing consistently across the HDSS sites. Further studies are warranted to understand other significant individual and contextual determinants to which these sex differences in health can be attributed. This will lay a foundation for a more evidence-based approach to resource allocation, and to developing health promotion programmes for older men and women in these settings. Keywords: ageing; survey methods; public health; burden of disease; demographic transition; disability; well-being; health status; INDEPTH WHO-SAGE

Access the supplementary material to this article: INDEPTH WHO-SAGE questionnaire (including variants of vignettes), a data dictionary and a password-protected dataset (see Supplementary files under Reading Tools online). To obtain a password for the dataset, please send a request with ‘SAGE data’ as its subject, detailing how you propose to use the data, to [email protected]

Received: 28 June 2010; Revised: 8 July 2010; Accepted: 8 July 2010; Published: 27 September 2010

eclining rates of fertility and mortality are especially women. Widespread availability of antiretro- driving demographic transitions in all regions virals is improving the quality and length of life lived with Dof the world, leading to global population HIV, but the overall effects on mortality patterns, life ageing. This includes substantial growth in the numbers expectancy, population structure and social roles will be and proportions of older adults in low- and middle- considerable for years to come. All this furthers the idea income countries, estimated at an annual growth rate of that multiple transitions are underway in contrasting 2.6%. In 2010, about 9.9% of the total Asian and 5.4% of settings. the total African populations are aged 60 years and over. Estimates of life expectancies at birth and at 60 years of By 2050, these population proportions of older people are age provide an objective way of measuring and comparing projected to increase to 23.6% and 10.7%, respectively. the health status of populations over time. In most Along with population ageing, the burden of morbidity countries, the life expectancies of women exceed those of and mortality in the population will also undergo change men and these differences are expected to widen in low- from burden profiles dominated by infectious diseases to income countries over the next 30Á40 years. Despite living those affected by chronic non-communicable diseases longer, there are indications that, compared with men, (NCD) (1). The chronic NCD burden is predicted to women in low-income countries report poorer health increase over the next 20 years from 60% to 79% in Asia (4Á6). Understanding sex-related health differences along and from 28% to 51% in Africa (2). The impact of HIV/ with gendered aspects of health, recognising groups at risk AIDS in eastern and southern Africa has been extreme, of poor health and identifying determinants of poor health leading to major reversals in mortality and different are all critical for planning the health needs of ageing patterns of demographic transition. The dominant sce- populations and improving health trajectories. nario in many sub-Saharan African countries will be This article discusses this pattern in eight Health and co-existing chronic infectious and non-communicable Demographic Surveillance System (HDSS) sites within the disease (3). The consequences for population ageing are INDEPTH Network (International Network for the considerable and impact the roles played by older people, Demographic Evaluation of Populations and Their

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 97 Nawi Ng et al.

Health, http://www.indepth-network.org) across Africa Study instruments and variables and Asia. The study used data collected using a modified, This study used a modified and shortened INDEPTH summary version of the WHO Study on global AGEing WHO-SAGE instrument consisting of health status and adult health (SAGE) which was linked with long- description, subjective well-being and quality of life itudinal HDSS background variables. This collaboration modules (see information at the end of the abstract). between WHO-SAGE and the INDEPTHÁHDSS sites The study questionnaire was developed through a con- links the SAGE survey tools with longitudinal HDSS data sultative process between INDEPTH and WHO-SAGE collection platform in order to improve understanding of with the goal of integrating a feasible number of useful the determinants of adult health and ageing in low- and SAGE modules into routine surveillance update activities middle-income countries in Africa and Asia. with minimum impact on existing HDSS procedures and The work underscores the importance of using interna- maximum return on measuring health and well-being. The tional survey data on self-reported health and function- survey instrument consisted of questions in eight health ality of older adults to complement statistics on life domains (affect, cognition, interpersonal relationships, expectancy and burden of illness. Our goal is to determine mobility, pain, self-care, sleep/energy and vision) with the extent to which various factors impact upon measures related anchoring vignettes. In each domain, two ques- of health, and how this occurs differentially for men and tions were asked to assess how much difficulty the women. We measure differences in self-reported health by respondent had in performing activities during the last sex, and explain how these differences can be attributed to 30 days. The summary instrument also assessed functional demographic and socio-economic determinants measured status using Activities of Daily Living (ADL) or Instru- in this study. These analyses inform an understanding of mental Activities of Daily Living (IADL) type of ques- the distribution and the socio-demographic and economic tions, and covered subjective well-being and quality of life determinants of self-reported health, which can contribute issues. This instrument was translated and back-translated to the development of health-promotion programmes and in eight local languages. Standardised training, interview more general support and development initiatives for older protocols and quality assurance procedures were used men and women. across all participating sites. Centralised training was provided to principal investigators from each site, who in Methods turn trained their respective survey teams: site-based training averaged 4.5 days in duration across the sites. Study population Mean interview time was 20 min. Three sites integrated This multi-centre INDEPTH WHO-SAGE study was the INDEPTH WHO-SAGE module into their routine conducted during 2006Á2007 in eight HDSS sites in Africa HDSS surveillance, while the remaining five sites con- and Asia: Agincourt (South Africa), Ifakara (Tanzania), ducted the INDEPTH WHO-SAGE study as a separate Nairobi (Kenya), Navrongo (Ghana), Filabavi (Viet data collection activity. Detailed descriptions of instru- Nam), Matlab (Bangladesh), Purworejo (Indonesia) and ments, survey protocols and quality control measures are Vadu (India) (7). The HDSS sites were selected to include described in a companion article in this volume (9). different geographic and socio-economic contexts. A total The INDEPTH WHO-SAGE questionnaire also col- of 93,347 individuals aged 50 years and over were lected information on overall self-reported health using identified from the surveillance databases across all eight the question ‘In general, how would you rate your health field sites. In six sites, all adults 50 years and over were today?’, using a 5-point response scale. However, the targeted for face-to-face interview; in the other two main outcome of interest in this article is the health score. sites (Navrongo and Matlab) a random sample of house- In brief, the composite health score was calculated based holds with at least one member aged 50 years and over was on self-reported health derived from the eight health selected. Respondents within these households were domain items. Each item response was based on a 5-point selected using Kish tables (8). In both cases, older ordered categorical scale. Due to its multidimensionality, individuals had a known non-zero probability of selection. the health score provided a more robust assessment of A total of 58,004 respondents aged 50 years and over were individual health levels than a single overall self-rated invited to participate, and the response rate was 80%, general health question and was subsequently used as the resulting in a final total sample of 46,269, ranging from health outcome variable in the planned analyses (10, 11). 2,072 in Nairobi to 12,395 in Purworejo. A total of 2,334 The composite health scores were calculated using item respondents (5.0%) were later excluded from the analysis response theory with a partial credit model (12). Each because of incomplete socio-demographic information item was calibrated using chi-squared fit statistics to [item non-response: age (n11); education (n450); assess its contribution to the composite health score. socio-economic status (n1,627); marital status (n121); The raw scores were transformed through Rasch model- living arrangements (n125)], giving a total sample of ling into a continuous cardinal scale, with 0 representing 43,935. worst health and a maximum score of 100 representing

98 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 Health inequalities among older men and women in Africa and Asia

best health (9). The psychometric properties of the health All the analyses were weighted by the 2007 population score have been assessed and reported elsewhere (13). age and sex distribution at each HDSS site. The Background information for each respondent was descriptive results were standardised to the WHO world obtained by linking the SAGE results to selected, standard population distribution to account for the standardised variables from the HDSS site databases, different population distributions across HDSS sites which contain extensive data on individual demographic (17). All statistical analyses were conducted in STATA characteristics as well as household-level information. Version 10.0 (18). The variables were harmonised across sites to ensure comparability. The socio-economic index for households Ethical considerations in each site was based on a locally derived wealth index; The research was approved by the Ethical Committee or all households in a site were allocated to wealth quintiles Board in each HDSS site and/or their host institutions, which were developed using principal component factor and the Ethics Review Committee at WHO, Geneva. analysis (14) on a range of asset variables including Informed consent was obtained from each individual dwelling characteristics and household possessions (such prior to the study. as livestock and durable goods). The wealth index was derived by each HDSS independently. Since these are Results relative measures, it was not possible to make direct A total of 43,935 respondents aged 50 years and over comparisons of quintiles across sites, but it is possible to (24,434 women and 19,501 men) were included in the compare health outcomes across wealth quintiles within analyses. Table 1 provides demographic and socio- each site/country, and time-trends in outcomes by wealth economic characteristics of the respondents. The smaller quintile across all sites. number of women in Nairobi and men in Agincourt reflects the dynamics of labour and social migration Data analyses occurring in these two settings. Overall, more women Descriptive results are presented for demographic and participated than men (55.6% and 44.4%, respectively) socio-economic variables at each site. Means and 95% with substantial variation across the sites. In Agincourt, confidence intervals (CI) for the health scores are women constituted three-quarters of respondents, com- presented to describe variations in different population pared to only 35% in Nairobi. The majority of respon- subgroups across the eight HDSS sites. dents were aged between 50 and 59 years (42%), along The health score was used as the dependent variable in with a substantial proportion of the oldest old (6.8% aged regression analyses. A mean score for each domain was 80 years and over). Nairobi had 72% of respondents aged obtained by taking the average of responses in the two 50Á59 years and only 2.3% aged 80 years and over. In domain-specific questions. The contribution of each contrast, Filabavi had 7.4% men and 14.3% women health domain (affect, cognition, interpersonal relation- respondents aged 80 years and over. In general, women ships, mobility, pain, self-care, sleep/energy and vision) to respondents and those from African sites had lower the health score was determined using its regression education levels than men and those from Asian sites. coefficient, and the analyses were adjusted by household Almost two-thirds of male respondents in Filabavi wealth quintiles and living arrangements, and respon- reported more than six years of education, in contrast dents’ age, education levels and marital status. Differ- to only 6% in Ifakara and 13% in Navrongo. The ences in health score by sex were then analysed to corresponding figures for women ranged from 2.3% in ascertain how much demographic and socio-economic Ifakara to 33% in Filabavi. Over 88% of male respon- factors contributed to the observed differences. dents in Asian sites were in current partnerships; while in Multivariable linear regression was used to assess the African settings, the corresponding proportion ran- statistical associations between socio-economic and de- ged from 76% in Agincourt to 87% in Nairobi. There mographic characteristics as independent variables, and were more older women in African sites who were not the health score as the dependent variable, for all currently in a relationship compared to women in Asian respondents, separately by sex and HDSS sites. A post- sites. Notably, 74% of women respondents in Nairobi regression decomposition based on Blinder-Oaxaca were either widowed, divorced or never married. Overall, methods (15, 16) was performed in order to show the more than 90% of respondents lived with other family extent to which sex-based differences in outcomes were members, except in Nairobi where up to 29% of men and attributable to differences in sex distributions of socio- 21% of women lived alone. economic and demographic characteristics, and how Tables 2 and 3 show the distributions of the health much to other factors. Together, multivariable regression score for men and women by different demographic and and decomposition provided a way of measuring and socio-economic characteristics across the HDSS sites. In explaining an outcome gap, which in this case was the all sites, both men and women aged 80 years and over mean difference in health score between men and women. consistently had lower health scores compared to respon-

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 99 Nawi Ng et al. India Vadu, Indonesia Purworejo, 2007 Á Matlab, Bangladesh Filabavi, Viet Nam Ghana Navrongo, Kenya Nairobi, Ifakara, Tanzania Agincourt, South Africa Distribution of study populations in eight Health and Demographic Surveillance System (HDSS) sites in Africa and Asia, 2006 . 596979 40.6 33.0 39.2 19.4 27.8 45.5 24.4 33.9 45.7 16.5 29.6 72.2 18.1 19.9 58.0 25.2 5.7 41.9 10.1 33.5 44.7 18.8 37.8 41.7 14.3 29.5 35.4 21.5 24.4 44.0 25.8 32.1 43.1 18.9 35.7 38.0 17.7 31.5 35.9 22.8 34.8 46.0 22.5 35.5 45.3 14.7 37.2 13.3 Á Á Á 50 60 70 80 and overNo formalAt most 6 yearsMore than 6 yearsIn partnership 7.1Single 8.6 23.7 26.4 49.9Living together in 20.4 household 4.1 15.8Living 63.8 alone 72.9 89.0 20.9 76.4 6.5 6.2First quintile 41.6 96.4 (lowest) 56.1 41.1Second quintile 2.3 2.3Third 59.5 97.5 quintile 23.6 25.5 84.5 6.7Fourth 15.0 42.6 quintile 52.7 98.2 58.9Fifth 50.1 quintile 11.0 (highest) 4.6 16.5 87.3 5.7 NA 70.7 15.5 86.8 12.7 15.4 95.0 3.6 79.4 3.2 26.5 17.8 NA 49.9 21.6 34.2 96.4 5.0 28.4 2.5 81.9 19.1 17.8 7.4 13.2 19.5 16.8 2.0 55.7 94.7 63.8 24.6 73.5 35.4 23.2 19.7 1.8 14.3 21.1 27.6 10.6 33.2 98.7 33.7 18.1 92.8 16.5 NA 21.9 29.3 15.9 4.9 33.4 41.3 23.5 91.1 25.6 20.6 64.6 60.5 13.1 20.1 30.8 NA 46.6 3.5 72.2 22.0 62.5 99.6 96.4 4.3 3.6 18.8 7.2 26.2 20.0 20.5 26.7 14.2 51.4 95.0 23.9 7.7 53.4 23.3 13.7 24.6 39.5 5.3 8.1 23.7 36.7 56.8 21.7 96.3 88.0 11.9 16.2 6.7 4.6 3.6 16.2 1.3 17.0 84.5 22.5 90.2 4.6 60.4 38.6 20.5 3.8 46.6 13.8 7.1 18.8 8.9 22.1 99.0 91.3 26.5 7.2 8.3 12.0 16.5 26.3 4.1 16.7 20.7 96.5 0.4 66.8 22.4 39.6 18.5 21.8 16.4 17.9 22.3 5.0 20.5 29.2 8.7 19.0 16.8 24.5 3.7 10.2 25.9 33.2 19.8 20.3 21.5 12.6 20.8 15.5 9.8 20.0 19.9 19.8 14.7 21.3 1.0 22.8 30.2 22.8 3.5 20.2 29.6

Table 1 CharacteristicsTotal subjectsAge group (years) M 949Education F levels 2,890 2,388 M 2,636 1,298Marital status F 693 1,634 MLiving 2,660 arrangements 3,462 F 5,054Household M socio-economic 1,999 status 2,005 F 5,420 6,333 M 2,351 2,163 F M F M F M F

100 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 Health inequalities among older men and women in Africa and Asia 70.2) 70.4) 68.3) 68.4) 68.4) 77.2) 68.8) 68.3) 69.0) 70.5) 68.4) 67.6) 67.6) 66.5) 68.6) 70.9) Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á India Vadu, 2007 Á 74.8) 69.3 (68.5 74.5) 69.3 (68.3 74.8) 67.4 (66.5 73.4) 67.2 (66.0 73.8) 67.1 (65.9 74.6) 69.2 (61.2 73.9) 68.3 (67.9 73.2) 66.3 (64.3 74.1) 68.5 (68.0 75.0) 69.7 (68.8 73.8) 67.8 (67.1 73.8) 66.0 (64.3 65.0) 65.8 (64.0 69.0) 65.6 (64.8 73.7) 67.9 (67.2 77.7) 70.1 (69.3 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Indonesia Purworejo, 64.4) 74.2 (73.6 63.3) 73.9 (73.4 63.6) 74.1 (73.4 62.8) 72.7 (72.1 63.5) 73.0 (72.3 68.5) 72.4 (70.2 63.2) 73.6 (73.3 64.8) 72.2 (71.2 63.2) 73.8 (73.5 64.5) 74.4 (73.8 63.3) 73.5 (73.1 63.0) 72.7 (71.7 56.5) 64.0 (63.0 60.2) 68.4 (67.9 62.8) 73.2 (72.7 66.3) 77.3 (76.8 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Matlab, Bangladesh 71.3) 63.7 (63.1 70.2) 62.6 (61.8 69.9) 62.6 (61.6 69.0) 61.8 (60.9 68.0) 62.6 (61.6 69.8) 64.5 (60.4 69.6) 62.8 (62.4 68.7) 62.5 (60.2 69.6) 62.8 (62.4 70.6) 63.8 (63.1 68.9) 62.7 (62.0 68.0) 62.5 (61.9 60.8) 54.9 (53.4 65.9) 59.3 (58.4 69.4) 62.2 (61.6 73.1) 65.7 (65.1 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Filabavi, Viet Nam 70.5) 70.6 (69.9 67.3) 69.5 (68.8 66.3) 69.2 (68.5 67.0) 68.2 (67.3 66.9) 66.7 (65.4 68.6) 67.2 (64.7 66.5) 69.3 (68.9 65.5) 66.9 (65.1 66.9) 69.3 (68.9 70.4) 70.2 (69.7 66.3) 68.2 (67.5 62.9) 59.7 (58.7 63.1) 65.3 (64.6 66.6) 68.8 (68.2 69.1) 72.5 (71.9 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Ghana Navrongo, Mean health score and 95% confidence interval 74.3) 67.8 (65.1 71.1) 66.2 (65.1 76.2) 65.4 (64.5 74.1) 66.1 (65.2 72.6) 66.1 (65.4 73.5) 66.3 (64.0 72.2) 66.0 (65.6 71.6) 64.4 (63.3 72.3) 66.4 (65.9 77.5) 67.3 (64.3 73.1) 65.8 (65.4 71.6) NA 65.8 (63.7 63.9) 61.0 (59.1 72.1) 62.1 (61.2 72.0) 65.9 (65.2 75.3) 68.4 (67.7 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Kenya Nairobi, 72.1) 69.6 (68.2 72.6) 73.9 (71.7 73.5) 71.8 (69.5 71.6) 71.1 (69.6 77.1) 72.0 (70.4 71.9) 71.2 (70.1 71.6) 69.1 (66.6 72.2) 71.4 (70.5 73.7) 75.0 (72.5 72.1) 71.6 (70.1 72.9) 69.8 (68.0 63.0) 60.1 (56.4 68.1) 69.1 (66.1 72.4) 70.5 (68.9 75.4) 74.4 (73.5 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Ifakara, Tanzania 69.6) NA 71.8 (69.3 66.8) 71.2 (70.3 68.4) 71.5 (70.4 68.0) 72.5 (71.5 67.3) 70.7 (69.7 70.3) 74.2 (71.4 67.3) 71.4 (70.9 67.0) 70.3 (68.9 68.0) 71.6 (71.1 70.2) 71.8 (69.9 67.9) 71.5 (71.0 66.9) 71.5 (70.1 64.3) 61.4 (59.9 66.7) 67.0 (65.9 67.9) 71.5 (70.7 69.1) 74.6 (73.8 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Agincourt, South Africa Distribution of health score across subgroups of men in eight Health and Demographic Surveillance System (HDSS) sites in Africa and Asia, 2006 . 79 65.5 (64.3 69 66.6 (65.3 59 67.8 (66.5 Á Á Á First quintile (lowest) 65.6 (64.0 Fifth quintile (highest) 68.0 (66.5 Fourth quintile 65.6 (64.5 Third quintile 66.6 (64.9 Second quintile 66.3 (64.5 Living alone 68.0 (65.7 Living together in household 66.5 (65.7 Single 65.5 (64.0 In partnership 67.1 (66.2 More than 6 years 68.8 (67.3 At most 6 years 66.4 (64.9 No formal 65.9 (64.8 80 and over 62.6 (60.8 70 60 50 Household socio-economic status Living arrangements Marital status Education levels Age group (years) Table 2 Characteristics

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 101 Nawi Ng et al. 67.0) 67.4) 66.2) 66.1) 66.6) 68.8) 66.2) 66.6) 66.3) 69.0) 66.1) 66.8) 64.0) 64.7) 66.6) 67.7) Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á India 2007 Vadu, Á 71.9) 66.3 (65.6 71.3) 66.5 (65.6 71.8) 65.4 (64.6 71.7) 65.0 (64.0 70.8) 65.5 (64.4 71.0) 66.9 (65.0 71.3) 65.8 (65.4 70.6) 65.8 (65.0 71.9) 65.8 (65.3 73.7) 67.5 (66.0 71.4) 65.7 (65.3 71.1) 65.3 (63.8 63.7) 62.5 (60.9 66.5) 63.9 (63.1 70.4) 66.0 (65.4 75.1) 67.1 (66.4 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Indonesia Purworejo, 56.9) 71.4 (70.8 56.7) 70.8 (70.2 55.8) 71.2 (70.6 55.7) 71.1 (70.6 55.7) 70.2 (69.7 59.3) 70.0 (69.0 55.8) 71.0 (70.7 55.9) 70.2 (69.7 56.5) 71.6 (71.2 59.3) 72.9 (72.1 57.0) 71.1 (70.7 55.5) 70.5 (69.9 53.0) 62.7 (61.7 52.3) 66.0 (65.5 56.0) 70.0 (69.6 58.2) 74.7 (74.3 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Matlab, Bangladesh 67.4) 56.2 (55.6 66.3) 56.0 (55.4 66.4) 55.0 (54.2 65.7) 54.8 (53.9 64.7) 54.9 (54.1 65.8) 57.7 (56.0 66.0) 55.4 (55.1 65.2) 55.3 (54.8 66.5) 56.0 (55.5 68.1) 58.0 (56.6 65.6) 56.3 (55.5 64.7) 55.1 (54.7 58.3) 51.1 (49.1 62.3) 51.4 (50.5 65.3) 55.4 (54.9 69.2) 57.8 (57.3 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Filabavi, Viet Nam 63.4) 66.8 (66.3 63.1) 65.8 (65.2 62.9) 65.8 (65.3 63.0) 65.1 (64.6 63.4) 64.0 (63.3 64.3) 64.6 (63.4 62.8) 65.7 (65.4 62.2) 64.7 (64.2 64.6) 66.2 (65.9 64.2) 67.4 (66.7 62.8) 65.2 (64.9 57.4) 57.7 (57.1 59.7) 61.9 (61.4 62.5) 64.9 (64.4 65.6) 68.8 (68.4 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Ghana Navrongo, Mean health score and 95% confidence interval 68.7) 62.1 (60.7 67.9) 62.5 (61.8 66.4) 62.3 (61.7 66.4) 62.5 (61.9 69.2) 62.9 (62.3 66.6) 62.8 (61.2 66.8) 62.5 (62.2 66.0) 61.9 (61.5 71.6) 64.1 (63.6 67.4) 62.7 (61.2 68.2) 62.5 (62.2 65.6) NA 63.3 (61.9 59.0) 55.7 (53.9 63.5) 59.1 (58.4 65.7) 62.1 (61.6 70.8) 65.2 (64.7 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Kenya Nairobi, 69.3) 66.0 (64.0 70.4) 64.7 (62.9 70.4) 64.8 (63.1 68.3) 66.8 (64.5 70.7) 65.1 (63.5 69.1) 65.7 (64.6 68.7) 65.0 (63.9 70.2) 69.0 (66.4 74.0) 64.5 (61.7 68.5) 67.0 (65.7 69.9) 64.4 (63.1 60.2) 56.4 (53.8 65.4) 60.7 (57.9 69.0) 64.1 (62.5 72.8) 69.6 (68.4 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Ifakara, Tanzania 66.5) NA 66.7 (64.8 65.1) 68.6 (68.0 66.2) 69.4 (68.5 65.0) 69.4 (68.3 66.5) 67.3 (66.3 65.4) 67.8 (64.8 65.5) 68.7 (68.3 65.1) 68.1 (67.4 66.3) 69.5 (68.9 67.9) 71.0 (68.1 65.7) 67.9 (67.2 65.4) 69.2 (68.6 61.4) 58.6 (57.0 63.4) 64.4 (63.4 66.3) 68.3 (67.6 66.8) 72.1 (71.5 Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Á Agincourt, South Africa Distribution of health score across subgroups of women in eight Health and Demographic Surveillance System (HDSS) sites in Africa and Asia, 2006 . 79 62.7 (62.1 69 65.7 (65.0 59 66.2 (65.6 Á Á Á First quintile (lowest) 65.6 (64.6 Fifth quintile (highest) 65.8 (65.0 Fourth quintile 64.3 (63.6 Third quintile 65.3 (64.5 Second quintile 64.2 (63.4 Living alone 63.7 (62.0 Living together in household 65.1 (64.7 Single 64.6 (64.1 In partnership 65.8 (65.2 More than six years 66.7 (65.5 At most six years 65.0 (64.2 No formal 65.0 (64.5 80 and over 60.3 (59.2 70 60 50 Household socio-economic status Living arrangements Marital status Education levels Age group (years) Table 3 Characteristics

102 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 Health inequalities among older men and women in Africa and Asia

dents in younger age groups. The discrepancies in health men and women in Navrongo and Filabavi, respectively. score between the lowest and the highest age groups were Inclusion of additional determinants (level of education, less in Agincourt and Vadu than in other HDSS sites. marital status, living arrangements and household wealth Both men and women with higher levels of education also quintiles) showed that up to 82% of the sex difference in consistently had higher health scores compared to the mean health score in Agincourt was attributable to respondents with lower levels of education, except for the distribution of the determinants between the two women in Nairobi and Navrongo where the patterns were groups, with a remaining 18% attributable to other not entirely clear. In all sites, both men and women who factors not included in the model. In contrast, almost were not in current partnerships also had marginally none of the health score disparity between men and lower health scores than those with partners. There were women in Matlab was attributable to this set of determi- no statistically significant within-site differences in health nants. The results of the fully adjusted model, therefore, scores observed between those who lived alone and those provide a better understanding of the way in which who lived together with other family members, nor across known factors contributed to sex differences in health different household socio-economic quintiles. There was scores across the fieldsites. a clear gradient in health score across different levels of self-reported health categories. The average health scores ranged from 52.0 (95% CI: 50.4Á53.6) in men who Discussion reported their health as ‘very bad’ to 76.7 (75.9Á77.5) in This article presents novel findings on how the differences men who reported their health as ‘very good’. The in health between men and women can be partially corresponding figures were 48.0 (46.5Á49.5) and 74.5 explained by socio-demographic and social factors, by (73.5Á75.5) for women (data not shown). unexplained inequality, and by the differences in unex- Each of the eight health domains contributed differ- plained inequality between settings. The aim of the ently to the overall health score in each site. Table 4 shows decomposition analysis was to move beyond a basic the commonalities and differences in contributions from comparison of sex differences in self-reported health, each domain across the sites. Matlab had the least and instead begin to unravel the determinants of the dispersion across the domains, whereas Purworejo had differences and variations across contrasting African and the most. Four health domains were identified as Asian settings. By statistically regressing available (and contributing the most to the overall health score: pain/ commonly used) independent variables, such as age, discomfort (in Ifakara and Purworejo men and women, education, marital status, socio-economic status and and in Matlab and Filabavi women), vision (in Nairobi living arrangements, the decomposition technique char- and Vadu), mobility (in Matlab and Filabavi men) and acterised the association of other factors Á potentially sleep/energy (in Navrongo and Agincourt). Interpersonal gender-related issues Á on health scores. Referring to Table relations contributed relatively less to the overall health 5, model 5, a possible interpretation is that gendered score than the other domains, except in Vadu. Self-care aspects of society in the Matlab area of rural Bangladesh contributed the least with the regression coefficients contribute more to the differences in reported health ranging from 0.14 among women from Ifakara (com- between men and women than in the Agincourt area of pared to a pain domain coefficient of 3.01 in the same rural South Africa. This suggests that the influence of site) to 0.94 in men from Purworejo. gendered aspects of health warrants closer examination A decomposition of the health score by sex was when investigating sex-based differences in health. How- conducted using a separate regression model adjusted ever, caution should be taken with this hypothesis until the for the effects of socio-economic and demographic limitations outlined below are taken into account. characteristics. Table 5 shows that in all sites, men had Three key results emerge from this cross-site study on higher health scores than women across all age-groups health and ageing in eight low- and middle-income (pB0.001). The gaps in the health score between men and countries. Firstly, despite women having higher life women were significantly larger in Matlab and Nairobi expectancy than men, older men reported better health compared to the other HDSS sites. There were large than older women in these settings. These results are in discrepancies in the proportion of the health score line with findings from Europe and North America difference between men and women attributable to group showing that women reported poorer health than men differences in socio-economic and demographic charac- (19, 20). The INDEPTH WHO-SAGE results also teristics; and similarly in the proportion of the gap that indicated significantly larger sex differences in health in was attributed to other influences not adjusted for in the Nairobi and Matlab than in the other HDSS sites. A model, such as gender discrimination. Within the propor- previous study from Matlab also reported poorer self- tion of the inequality attributed to individual character- reported health in women than in men, independent of istics, sex differences in age contributed from 13.4% to age. However, the contribution of sex to self-reported 24.8% of the disparity observed in health score between health disappeared after controlling for objective physical

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 103 Nawi Ng et al. tem (HDSS) 3.17 Vision 2.61 Mobility 2.59 Pain 2.52 Affect 2.43 Interpersonal 2.26 Cognition 1.39 Self-care 1.39 Sleep/energy 2.38 Vision 2.35 Pain 2.19 Interpersonal 2.11 Mobility 2.02 Affect 1.91 Cognition 1.51 Self-care 1.50 Sleep/energy                 ed as the outcome variable, and the 4.06 Pain 3.41 Cognition 3.18 Vision 2.81 Sleep/energy 2.43 Affect 2.19 Mobility 0.94 Interpersonal 3.40 Pain 3.04 Cognition 2.68 Vision 2.30 Mobility 2.30 Sleep/energy 2.16 Affect 1.31 Interpersonal               2.09 Mobility 1.85 Pain 1.82 Affect 1.66 Sleep/energy 1.55 Cognition 1.48 Vision 0.96 Self-care 0.94 Interpersonal 0.94 Self-care 1.51 Pain 1.51 Interpersonal 1.49 Affect 1.46 Vision 1.46 Mobility 1.39 Self-care 1.37 Sleep/energy 1.34 Cognition 0.68 Self-care                 2.81 Mobility 2.70 Sleep/energy 2.38 Pain 2.14 Cognition 1.88 Affect 1.80 Vision 1.10 Interpersonal 2.10 Pain 2.04 Sleep/energy 2.03 Mobility 1.88 Cognition 1.59 Affect 1.48 Vision 1.21 Interpersonal 0.59 Self-care                2.66 Sleep/energy 2.20 Affect 2.15 Mobility 2.04 Pain 2.00 Cognition 1.54 Interpersonal 1.51 Vision 0.19 Self-care 0.20 Self-care 1.90 Sleep/energy 1.78 Mobility 1.74 Pain 1.71 Affect 1.69 Cognition 1.49 Interpersonal 1.34 Vision 0.81 Self-care                 3.88 Vision 3.57 Sleep/energy 3.53 Pain 3.38 Affect 2.56 Mobility 2.38 Cognition 1.87 Interpersonal 2.53 Pain 2.41 Mobility 2.30 Vision 2.14 Cognition 2.02 Affect 1.94 Sleep/energy 1.74 Interpersonal 0.54 Self-care                2007 Á 3.51 Pain 3.19 Mobility 2.60 Vision 2.42 Sleep/energy 2.30 Cognition 1.89 Affect 0.56 Interpersonal 0.46 Self-care3.01 Pain 0.16 Self-care 2.99 Mobility 2.16 Vision 2.15 Cognition 2.07 Sleep/energy 1.92 Affect 0.71 Interpersonal 0.14 Self-care                 Regression coefficients for each domain (ranked from highest to lowest) with health score as outcome in eight Health and Demographic Surveillance Sys . 2.48 Sleep/energy 2.25 Cognition 2.24 Affect 2.23 Pain 1.80 Vision 1.70 Mobility 1.50 Interpersonal 0.50 Self-care 2.29 Sleep/energy 2.22 Pain 2.12 Cognition 2.01 Affect 1.58 Mobility 1.53 Vision 1.31 Interpersonal 0.80 Self-care                 sites in Africa and Asia, 2006 Agincourt, South AfricaMen Ifakara, Tanzania Nairobi, Kenya Navrongo, Ghana Filabavi, Viet Nam Matlab, Bangladesh Purworejo, Indonesia Vadu, India Women Table 4 Note: Numbers represent regression coefficientsregression for analyses each were health adjusted domain derived for from age separate (as regression continuous analyses variable), for education each level, site. marital Health status, score was living us arrangements, and wealth quintiles.

104 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 Health inequalities among older men and women in Africa and Asia

performance, limitations in activities of daily living, and

1.0 acute and chronic morbidity (21). India Vadu,  Secondly, individual and household socio-economic determinants contributed differently across settings in explaining the sex differences in reported health. Differences in household socio-economic levels and 0.9

 living arrangements, and respondent’s age, education and Indonesia Purworejo, marital status, provided virtually no explanation in Bangladesh while accounting for 71% and 82% of the 2007

Á sex difference in health score observed in Nairobi, Kenya and Agincourt, South Africa, respectively. Importantly, 1.9 0.4 22.5 35.6 Matlab,   inequalities observed in the health score, and the sex

Bangladesh differences between sites, may also be explained by individual and contextual factors not assessed in this study, such as occupational status, history of chronic morbidity, presence of physical disabilities and other Filabavi, Viet Nam environmental and socio-demographic risk factors at household and village levels. Thirdly, different health domains contributed differ- ently to the overall health score for men and women in 9.6 46.1 1.9 8.0 24.8 13.4 24.8 Ghana  each setting. Questions on self-care, which assess respon-  Navrongo, dents’ difficulties in washing/dressing or bathing and maintaining general appearance, have been used exten- sively in different health measurement tools (22, 23) but Kenya

Nairobi, consistently contributed least to overall health scores, in both men and women and in almost all the study sites. This might be due to the help given by members of extended families in many of these field settings. This study provides deeper understanding on how various Ifakara, Tanzania functional domains affect people’s perception of their health. Despite its usefulness in predicting future mor- bidity and mortality in both developed and developing countries (24Á26), a single question on self-rated health 48.2 21.3 67.5 21.8 44.6 1.6 24.6 36.2 79.7 28.7 70.5 23.5 45.7 0.8 24.8 35.1 81.5 30.3 69.1 22.1 44.6

Agincourt, provides little indepth understanding of something as South Africa complex and multifaceted as health. This study, however, showed a consistent trend towards better health scores in people who rated their health as ‘very good’ compared to those who rated their health as ‘very bad’. This domain- specific knowledge is vital in laying the foundation for rational resource allocation and for developing appro- priate evidence-based health promotion programmes for older adults. The study attempts to measure and compare the health of older adults in low- and middle-income countries, information largely lacking in resource-constrained set- tings. Increasing longevity will have substantial health, economic and social impacts in all countries, and will particularly affect under-resourced and under-performing health systems in low-income countries, which are gen- erally poorly prepared to provide the chronic care needed to manage non-communicable conditions in older people

Decomposition analysis of predictors of inequality in health scores between men and women in eight INDEPTH HDSS sites, 2006 (3, 27, 28). This study has highlighted prominent sex . differences in the health of older adults and raises the need to further study the factors contributing to these MenWomenDifference between men and women 2.05 3.04 6.48 66.67 64.62 3.07 71.78 68.74 4.55 73.13 66.65 65.87 62.81 7.30 68.93 64.39 2.64 62.86 55.56 2.45 72.92 70.28 68.45 66.00 Characteristics marital status status, and living arrangements status, living arrangements, and household wealth quintiles Mean health score Model 1: % explained byModel inclusion 2: of % age explained by inclusion of age and education level 20.8 6.6 5.1 30.3 6.2 21.1 Model 3: % explained by inclusion of age, education level, and Model 4: % explained by inclusion of age, education level, marital Table 5 Model 5: % explained by inclusion of age, education level, marital disparities. This will be important for developing targeted

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 105 Nawi Ng et al.

interventions to address differences in health between ens the benefit of embedding the INDEPTH WHO- men and women. SAGE study in the HDSS operation. The study was designed as an add-on to established This comparative study may therefore benefit from HDSS sites in Africa and Asia. Embedding this study analyses incorporating vignette-based adjustments (data within HDSS sites allowed for data linkages between this for which have been collected) that map self-reported cross-sectional study and the rich demographic and health to a common comparable scale in each domain socio-economic information available in HDSS data- (32, 33). These adjustments might improve the cross-site bases. The infrastructure established for this research comparability of the results. Similarly, subsequent ana- provides the unique opportunity to follow these popula- lyses correlating health outcomes by sex with observed tions longitudinally in a scientifically reliable manner. mortality Á a robust potential with HDSS longitudinal Linking the health and function indices with future data collection Á will probably be enlightening. morbidity and mortality data, collected routinely as Despite these limitations, the study provides a robust part of regular HDSS update rounds, will allow deeper data set, baseline and data collection platform that can be understanding of the dynamics of health transition and used to inform future interventions Á and their evaluation population ageing in low- and middle-income countries Á for older people’s health across contrasting geographic (29Á31). The results of this study may also serve as a and socio-cultural settings. baseline for observing trends and changes in older people’s health in the future, whether occurring naturally Conclusion or following policy shifts. This INDEPTH WHO-SAGE study examined sex differ- There are some limitations to this study. Firstly, the ences in health among older adults within low- and study subjects may not have been representative of older middle-income countries and found that men reported people in their respective countries Á although, in all significantly better health than women. It also unveiled cases, they reflect poorer, often rural, populations. In wide variation in how individual and household socio- some HDSS sites, a random sample of the older adults economic characteristics explain the gaps in self-reported under surveillance was recruited into the study, whereas health observed between men and women in Africa and others surveyed the entire surveillance population aged Asia. Further studies are needed to examine individual 50 years and over. Due to the differing population and contextual determinants to which the health gaps structure within each HDSS and differences in sampling between older men and women can be attributed, strategies, all prevalence data were standardised to the including gender roles, thus addressing the health in- WHO standard population (17). Secondly, the compar- equalities observed. We expect such analyses to inform ability of this cross-national study on self-reported health our understanding of the distribution and determinants may be compromised by the dynamics of ageing and the of health and well-being by sex and age, and to provide cultural influences on health in the different settings. stronger evidence on which to base national and global The instrument used to assess self-reported health in the policies on population health and ageing. While the different domains might not be able to fully capture gender paradox between health and life expectancy exists people’s experiences and expectations for their health. in all these settings, our results affirm that old age will However, this method for measuring health has been used bring particular problems for women in low-resource as part of the World Health Survey in some 70 countries societies. There will be clear need for gender-sensitive with robust results (32). Future research should compare health interventions to address the higher level of poor how these self-reported health items are correlated with health reported in older women and the documented more objective measures, such as blood pressure and health differences between the sexes. other findings from medical examination. Thirdly, since the wealth quintiles, serving as a proxy for socio- economic status, were constructed by each HDSS, they Acknowledgements are relative rather than absolute measures and were not The authors would like to acknowledge the help of Dr. Richard harmonised across sites. The expected patterns of health Gibson and Dr. Jenny Stewart Williams from the Research Center by wealth were not clearly demonstrated within or across on Gender, Health, and Ageing, University of Newcastle, Australia HDSS sites and did not contribute significantly to the for their statistical advice. decomposition results. This may need to be addressed in future analyses of the dataset using longitudinal Conflict of interest and funding approaches. Fourthly, the cross-sectional nature of the Financial support was provided by the US National data limits the possibility of drawing causal associations Institute on Aging through an interagency agreement on how health influences socio-economic status or vice- with the World Health Organization, supplemented by versa. The potential to use these cross-sectional data as a support from Umea˚ University for the Filabavi and baseline for further longitudinal data analyses strength- Purworejo sites. Both WHO and INDEPTH contributed

106 Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 Health inequalities among older men and women in Africa and Asia

financial and human resources to the collaboration. The 15. World Bank. Unravelling causes II: the Oaxaca approach. Umea˚ Centre for Global Health Research (supported by Quantitative techniques for health equity analysis-technical note #15. Washington, DC: World Bank; 2004. FAS, the Swedish Council for Working Life and Social 16. Jann B. The Blinder-Oaxaca decomposition for linear regression Research, Grant No. 2006-1512) provided technical models. Stata J 2008; 8: 453Á79. support and advice to the sites and co-hosted with 17. Ahmad O, Boschi-Pinto C, Lopez A, Murray C, Lozano R, INDEPTH an analytic and writing workshop in 2008. Inoue M. Age standardization of rates: a new WHO standard. The Health and Population Division, School of Public Geneva: World Health Organization; 2001. 18. StataCorp LP. Stata 10.0. TX, USA; 2007. Health, University of the Witwatersrand, South Africa 19. Bardage C, Pluijm SMF, Pedersen NL, Deeg DJH, Jylha¨ M, serves as the satellite secretariat providing scientific Noale M, et al. Self-rated health among older adults: a cross- leadership, technical and administrative support for the national comparison. Eur J Ageing 2005; 2: 149Á58. INDEPTH Adult Health and Ageing initiative. 20. Case A, Paxson C. Sex differences in morbidity and mortality. Demography 2005; 42: 189Á214. 21. Rahman MO, Barsky AJ. Self-reported health among older Bangladeshis: how good a health indicator is it? Gerontologist References 2003; 43: 856Á63. 22. Pluijm SM, Bardage C, Nikula S, Blumstein T, Jylha M, Minicuci N, et al. A harmonized measure of activities of daily 1. Omran AR. The epidemiologic transition. A theory of the living was a reliable and valid instrument for comparing epidemiology of population change. Milbank Mem Fund Q disability in older people across countries. J Clin Epidemiol 1971; 49: 509Á38. 2005; 58: 1015Á23. 2. Online statistics for the Global Burden of Disease [Update 2004. 23. Katz, S. Assessing self-maintenance: activities of daily living, Database on the Internet]. World Health Organization; 2008. mobility and instrumental activities of daily living. JAGS 1983; Available from: http://www.who.int/healthinfo/global_burden_ 31: 721Á6. disease/2004_report_update/en/index.html [cited 29 June 2009]. 24. Idler EL, Benyamini Y. Self-rated health and mortality: a review 3. Mayosi BM, Flisher AJ, Lalloo UG, Sitas F, Tollman SM, of twenty-seven community studies. J Health Soc Behav 1997; Bradshaw D. The burden of non-communicable diseases in 38: 21Á37. South Africa. Lancet 2009; 374: 934Á47. 25. Frankenberg E, Jones NR. Self-rated health and mortality: does 4. Whitson H, Landerman L, Newman A, Fried L, Pieper C, the relationship extend to a low income setting? J Health Soc Cohen H. Comorbidities and the gender difference in disability: Behav 2004; 45: 441Á52. analysis of the cardiovascular health study. J Amer Geriatr Soc 26. Yu ES, Kean YM, Slymen DJ, Liu WT, Zhang M, Katzman R. 2009; 57: S1ÁS14. Self-perceived health and 5-year mortality risks among the 5. Murtagh K, Hubert H. Gender differences in physical disability elderly in Shanghai, China. Am J Epidemiol 1998; 147: 880Á90. among an elderly cohort. Am J Public Health 2004; 94: 1406Á11. 27. United Nations Department of Economic and Social Affairs 6. Olsen KM, Dahl SA. Health differences between European PD. World population ageing 2007. New York: United Nations; countries. Soc Sci Med 2007; 64: 1665Á78. 2007. 7. INDEPTH Network. Population and health in developing 28. Strong K, Mathers C, Epping-Jordan J, Beaglehole R. Prevent- countries. Ottawa: INDEPTH Network, IDRC; 2002. ing chronic disease: a priority for global health. Int J Epidemiol 8. Kish L. Statistical design for research. New York: Wiley; 1987. 2006; 35: 492Á4. 9. Kowal P, Kahn K, Ng N, Naidoo N, Abdullah S, Bawah A, 29. Jagger C. Compression or expansion of morbidity Á what does et al. Ageing and adult health status in eight lower-income the future hold? Age Ageing 2000; 29: 93Á4. countries: the INDEPTH WHO-SAGE collaboration. 30. Fries JF. Aging, natural death, and the compression of Global Health Action Supplement 2, 2010. DOI: 10.3402/ morbidity. N Engl J Med 1980; 303: 130Á5. gha.v3i0.5302. 31. Cohen B, Menken J, editors. National Research Council. Ageing 10. Salomon J, Mathers C, Chatterji S, Sadana R, .U¨ stu¨n T, Murray in Sub-Saharan Africa: recommendation for furthering re- C. Quantifying individual levels of health: definitions, concepts, search. Washington, DC: The National Academies Press; 2006. and measurement issues. In: Murray C, Evans D, eds. Health 32. Salomon JA, Tandon A, Murray CJ. Comparability of self rated systems performance assessment: debates, methods and empiri- health: cross sectional multi-country survey using anchoring cism. Geneva: World Health Organization; 2003. p. 301. vignettes. BMJ 2004; 328: 258. 11. Ju¨rges H, Avendano M, Mackenbach J. How comparable are 33. King G, Murray CJ, Salomon JA, Tandon A. Enhancing the different measures of self-rated health? Evidence from five validity and cross-cultural comparability of measurement in European countries. Mannheim: Mannheim Research Institute survey research. Am Political Sci Rev 2004; 98: 191Á207. for the Economics of Ageing; 2007. 12. Zheng X, Rabe-Hesketh S. Estimating parameters of dichot- omous and ordinal item response models with gllamm. Stata J *Nawi Ng 2007; 7: 313Á33. Department of Public Health and Clinical Medicine 13. Moussavi S, Chatterji S, Verdes E, Tandon A, Patel V, Ustun B. Centre for Global Health Research, Epidemiology and Global Health Depression, chronic diseases, and decrements in health: results Umea˚ University from the World Health Surveys. Lancet 2007; 370: 851Á8. SE-901 85 Umea˚ , Sweden 14. Vyas S, Kumaranayake L. Constructing socio-economic status Tel: 46 90 7851391 indices: how to use principal component analysis. Health Policy Fax: 46 90 138977 2006; 21: 459Á68. Email: [email protected]

Citation: Global Health Action Supplement 2, 2010. DOI: 10.3402/gha.v3i0.5420 107