Access to care

For children under‐five across high pneumonia mortality countries in sub‐Saharan Africa

Camielle Noordam

Promotor Prof. dr. G.J. Dinant

Copromotor Dr. J.W.L. Cals

Beoordelingscommissie Prof. dr. C.J.P.A. Hoebe (voorzitter) Dr. P. van den Hombergh Dr. J.S.M. Krumeich Prof. dr. J.F.M. Metsemakers Prof. dr. S.S. Peterson Contents

Part I Introduction 5

Chapter 1 General introduction 7

Part II The three phases of delay in care 17

Chapter 2 Associations between caregivers’ kowledge and care seeking 19 behaviour for children with suspected pneumonia in six sub‐Saharan African countries Submitted

Chapter 3 Care seeking behaviour for children with suspected pneumonia 33 in countries in sub‐Saharan Africa with high pneumonia mortality PLoS One 2015;10:e0117919

Chapter 4 The use of counting beads to improve the classification of fast 53 breathing in low resource settings: A multi‐country review Health Policy and Planning 2015;30:696–704

Part III A potential solution to decrease delays 71

Chapter 5 Improvement of maternal health services through the use of 73 mobile phones Tropical Medicine & International Health 2011;16:622–626

Chapter 6 Improving care‐seeking for facility‐based health services in a rural, 83 resource‐limited setting: Effects and potential of an mHealth project African Population Studies 2015;28:1643‐1662

Chapter 7 Assessing scale‐up of mHealth innovations based on intervention 103 complexity: Two case studies on child health programmes in and Zambia Journal of Health Communication 2015; 0: 1–11

Part IV Discussion 123

Chapter 8 General discussion 125 Summary 135 List of publications 139

PART I

Introduction

5

6 CHAPTER 1

General introduction

7 Chapter 1

8 General introduction

CHILD MORTALITY

Over the past decades, child mortality has reduced significantly. Estimates from the United Nations illustrate that there has been a global decline in the under‐five mortality of 53 percent; from 91 deaths per 1,000 live births in 1990 to 43 in 2015.1‐2 Despite these changes, 5.9 million children died before their fifth birthday in 2015 (i.e., more than 16,000 deaths a day), mostly from preventable diseases.2 Infectious diseases, also known as transmissible or communicable diseases, accounted globally for more than half of the under‐five deaths, followed by deaths during or shortly after birth. Of the infectious diseases, pneumonia is the leading cause of the under‐five mortality attributing to 16% of all child deaths, followed by diarrhoea (9%) and malaria (5%). Nutritional status influences these outcomes; about 45 percent of the under‐five mortality is attributable to under‐nutrition.1‐5 Of the children under the age of five, the incidence of infectious diseases is the highest for those under the age of 2; more specifically 81% of the deaths due to pneumonia occur within the first two years of a child’s life.6 Figure 1.1 shows the differences in causes of under‐five mortality between high‐ and low‐income countries, illustrating that as income levels within countries decrease, the proportion of deaths due to infectious diseases increases.

Malaria Diarrhoea 10% Other 10% 22% Infectious AIDS diseases 2% 51% Other Pneumonia neonatal 17% Pertussis, 27% tetanus, Sepsis measles, 5% meningitis b 7%

Figure 1.1 Causes of under‐five mortality for high‐ and low income countries; a) high‐income countries, 1.4% of global under‐five mortality, and b) low‐income countries, 33% of global under‐five mortality. Data from Committing to Child Survival: A Promise Renewed. Progress Report 2013. © United Nations Children’s Fund (UNICEF) September 2013

9 Chapter 1

Of all children, a child living in sub‐Saharan Africa is most likely to die before the age of five, where on average 1 out of every 8 children dies before their fifth birthday.2 Huge differences are seen in the chance of survival within and between these countries, where Angola has the highest mortality rate (157 per 1,000 live births) and Seychelles the lowest (14 per 1,000).1‐2 Figure 1.2 shows the differences in mortality by country.

Figure 1.2 The differences in under‐five mortality by country, with the highest rates found in sub‐Saharan Africa. Printed with permission from Committing to Child Survival: A Promise Renewed. Progress Report 2015. © United Nations Children’s Fund (UNICEF) September 2015

One of the reasons for child mortality to decline over the past decades is due to an increase in coverage of effective health interventions. These interventions focus not only on increasing access to care upon the onset of an illness (i.e., more effective and affordable treatments), but also on measures which prevent children from becoming ill in the first place (e.g., clean water, sanitation, education, improved nutrition and vaccinations).2‐4 While improved access to these interventions has saved a lot of lives, especially children living in isolated and marginalized settings still fail to reach them.7 Not only do these children fail to access preventive measures, but upon illness they also often fail to reach acceptable, affordable and appropriate health care, in time.

10 General introduction

ACCESS TO CARE

To know how to increase coverage of these interventions for children in sub‐Saharan Africa, especially amongst those who currently fail to reach them in time, is important. To do so, donors and policy makers need to understand the underlying determinants which prevent children from accessing these interventions in the first place. The model of ‘three delays’ has been used to help untangle challenges associated with care seeking. The model focuses on delays in accessing care, as the understanding is that the chance to survive is linked to the timelines in which care is received.8 In this thesis, the model will be applied to assess the challenges associated with delays in accessing care for children, more specifically those with symptoms of pneumonia (also referred to as ‘suspected pneumonia cases’). While the model was initially designed to categorize factors affecting the onset of obstetric complications and its outcomes, it is not the first time it is used to assess child health outcomes.9‐10 Analyses based on such modelling help create a more comprehensive understanding of why care is – or is not sought in time. This as it looks at challenges at household, as well as facility level. This is needed, as the children accessing care represent only a subset of all the children actually requiring health services. Hence delays which occur at home need to be examined too – even more so in sub‐Saharan African settings, where sick children often fail to reach the formal health system.10

How challenges in accessing care can be captured by three stages of delays Thaddeus and Maine designed the model of three phases of delay, which was first presented in 1994.8 The three phases are as followed; 1. The delay in deciding to seek care on the part of the individual, the family or both, which can be influenced – amongst others – by the status of women; distance from the health facility; costs; poor recognition and/or understanding of the illness (to assess the complications and/ or risks); previous experiences; and perceived quality of care. 2. The delay in reaching an adequate health care facility, which is mostly determined by geographical aspects, such as the distribution of facilities and the conditions of the road. 3. And, the delay in receiving adequate health care at the health facility, which can be influenced by poor quality and lack of resources as a result of inadequately trained and/ or motivated staff; out‐of‐stocks; inadequate referral systems; etc.

As the chance to survive is linked to the timelines in which care is received, it is evident that in areas where mortality is high, coverage of effective interventions is insufficient.

11 Chapter 1

Therefore, the analyses of care seeking behaviour, based on the model of three delays, can help improve programming to ensure more effective coverage of life‐saving health interventions.11 To date, little is known on how these delays affect care seeking behaviour across high mortality countries in sub‐Saharan African countries, to which extent care seeking patterns are similar, and what lessons learned and potential best practices are which should be shared across resource limited settings.

Leveraging mobile technology to reduce delays in accessing care To improve coverage of effective programs, it is not only important to understand what the existing challenges are, but also to identify ways in which these can be overcome. With mobile phone users increasing almost a three‐fold between 2005 and 2010 and reaching 367 million subscribers in mid‐2015 in sub‐Saharan Africa,12 there are high expectations that this technology can help connect isolated communities and healthcare services, thereby reducing delays in accessing care. The use of mobile phones to improve access to health is referred to as mHealth. Over the past decade, mHealth initiatives have focussed on addressing various aspects of the three phases of delay; for example by focussing on increasing knowledge, providing financial support, strengthening provider‐to‐provider communication systems, data collection methods, and ‐ amongst others – strengthening the supply chain management.13 Nevertheless, there is limited evidence on how mobile phones can most affectively address these delays, with only few evaluating the effectiveness of these initiatives in resource limited settings.14‐19 Finally, while the expectations are high, little is known on why these initiatives fail to go to scale in rural settings in sub‐Saharan Africa.20

Problem statement In summary, in sub‐Saharan African countries most children die of preventable and treatable illnesses because they fail to reach acceptable, affordable and appropriate health care in time. Pneumonia is responsible for most of these deaths. The extent to which the three phases of delay ‐ (deciding to seek, as well as reaching and receiving care) ‐ affect care seeking for children in sub‐Saharan African settings is yet unknown. While mobile technology appears to present great opportunities to address these delays, there is still little evidence on how this technology can best be used to improve access to health care. A better understanding of these delays and the potential of mHealth initiatives can help to identify programmatic coverage with their successes and challenges, as well as to identify opportunities.21‐22

12 General introduction

AIM, SCOPE AND OUTLINE OF THIS THESIS

Aim The aim of this thesis is to expand the knowledge on the three phases of delays in accessing health care, and to assess if and how the rapidly expanding field of mHealth can decrease these delays in order to improve programmatic coverage of effective health interventions across various settings in sub‐Saharan Africa. The specific objectives, which this thesis aims to answer, are: 1. How do the delays in care (deciding to seek, as well as reaching and receiving) affect care seeking behaviour across sub‐Saharan Africa? And, how can better knowledge of these delays lead to improved programming? This thesis seeks to improve the knowledge on these questions by focussing on the exemplary condition of pneumonia. 2. What is the potential of mobile technology to improve health outcomes by addressing the delays in accessing care, in resource limited settings? And, what are programmatic challenges in implementing mHealth strategies?

Scope To investigate the three phases of delays in all its complexity is far beyond the scope of a single thesis. Nevertheless, all three delays are considered in this thesis as these can adversely affect child health outcomes in sub‐Saharan Africa. A better understanding of de magnitude of each of these delays is therefore relevant for policy makers and programmers; especially those working in resource limited settings. To enable an assessment of all three phases of delays, the complexity of each delay is reduced to manageable proportions. In other words, in order not to state generalities, the strategy of this thesis is to give exemplary case studies for each delay, to illustrate the complexities at hand. For example, the first delay – deciding to seek care – is influenced by more than just knowledge; nevertheless, this thesis only assesses the association between knowledge and care seeking. In general, the focus of this thesis is on children under the age of five, living in sub‐ Saharan Africa settings with high mortality rates. More specifically, the first three chapters will focus on pneumonia; globally the main cause of childhood mortality.

13 Chapter 1

Outline This thesis is built around two main parts, namely a section which addresses the three phases of delay (Part II) and a section which assesses the potential of mobile phones as a solution (Part III). Part II, the section on the three phases of delay in care, is built around three chapters. Each chapter focusses on one of the phases of delay. The first chapter analyses causes of the delay in deciding to seek care (Chapter 2). This assessment is based on national survey data from countries in sub‐Saharan Africa, and assesses the correlation between knowledge of pneumonia symptoms and care seeking behaviour. The second chapter analyses the causes of the delay in reaching care (Chapter 3). As there are many reasons why care is not reached in time, this chapter starts with assessing to which care providers caregivers take their child and what influences this. This assessment is also based on national survey data, focussing on care seeking behaviour for children under five with symptoms of pneumonia. The third, and last chapter of this section, focusses on the delay in receiving adequate health care (Chapter 4). This delay can be caused by a myriad of reasons often linked to a poor quality and lack of resources. As mentioned, one of the reasons is inadequately trained staff. For this chapter, data were analysed from five different international organizations working in resource limited settings in sub‐Saharan Africa. The focus of this chapter is on the accuracy of community health workers in classifying breathing rates and the effect of counting beads in helping them to overcome key challenges related to classifying them. Part III, the section a potential solution, is built around three chapters all three focusing on the use of mobile phones. The first chapter is an assessment on the potential of mobile phones to improve access to health services in resource‐limited settings (Chapter 5). This assessment is based on a literature review and focusses on the evidence regarding the impact of mobile phones to improved maternal and new‐born health focussing on each phase of delay. The second chapter of part III investigates the impact of a mHealth initiative in improving care seeking behaviour in Malawi. The initiative is a toll‐free hotline and mobile messaging service, which connects caregivers to health workers and health information (Chapter 6). The last chapter of this section assesses what the key challenges are in relation to scaling‐up mHealth interventions (Chapter 7). This assessment is based on two case studies, nutrition surveillance and early infant diagnosis of HIV, in Malawi and Zambia. The last part of this thesis is the general discussion, which will focus on how improved knowledge of the delays and mHealth can help identify programmatic coverage with its successes and challenges.

14 General introduction

REFERENCES

1. United Nations Children's Fund (UNICEF) global databases 2015. Available: http://data.unicef.org/child‐ mortality/under‐five.html. Accessed 2016, Jan 12. 2. UNICEF (2015) Levels & trends in child mortality. Report 2015. New York: UNICEF. 3. World Health Organization (WHO) and UNICEF (2013) Ending preventable child deaths for pneumonia and diarrhoea by 2025: The integrated Global Action Plan for Pneumonia and Diarrhoea (GAPPD). Geneva: WHO. 4. UNICEF (2015) Committing to child survival: A Promise Renewed. Progress report: 2015.New York: UNICEF. 5. UNICEF (2014) Committing to child survival: A Promise Renewed. Progress report: 2014. New York: UNICEF. 6. Fischer Walker CL, Rudan I, Liu L, Nair H, Theodoratou E, et al. Global burden of childhood pneumonia and diarrhea. Lancet 2013;381:1405–1416. 7. UNICEF (2010) Narrowing the gaps to meet the goals. Available: http://www.unicef.org/nutrition/index_55927.html. Accessed 2016, Jan 12. 8. Thaddeus S, and Maine D. Too far to walk: Maternal mortality in context. Soc Sci Med 1994;38: 1091‐1110. 9. Waiswa P, Kallander K, Peterson S, Tomson G, Pariyo GW. Using the three delays model to understand why newborn babies die in eastern Uganda. Trop Med Int Health 2010;15:964–972. 10. Källander K, Hildenwall H, Waiswa P, Galiwango E, Peterson S, et al. Delayed care seeking for fatal pneumonia in children aged under five years in Uganda: a case‐series study. WHO Bulletin volume 2008;86:321‐416. 11. Gulliford M, Figueroa‐Munoz J, Morgan M, Hughes D, Gibson B, et al. What does ‘access to health care’ mean? J Health Serv Res Policy 2002;7:186‐188. 12. GSMA website. Available: http://www.gsma.com/. Accessed 2016, Jan 12. 13. Labrique AB, Vasudevan L, Kochi E, Fabricant R, Mehl G. mHealth innovations as health system strengthening tools: 12 common applications and a visual framework. Glob Health Sci Pract 2013;1: 160‐171. 14. Tomlinson M, Rotheram‐Borus MJ, Swartz L, Tsai AC. Scaling up mHealth: Where is the evidence? PLoS Med 2013;10:e1001382. 15. Free C, Phillips G, Galli L, Watson L, Felix L, et al. The effectiveness of mobile‐health technology‐based health behaviour change or disease management interventions for health care consumers: A systematic review. PLoS Med 2013;10:e1001362. 16. Free C, Phillips G, Watson L, Galli L, Felix L, et al. The Effectiveness of Mobile‐Health Technologies to Improve Health Care Service Delivery Processes: A Systematic Review and Meta‐Analysis. PLoS Med 2013;10:e1001363. 17. Fotso JC, Robinson AL, Noordam AC, Crawford J. Fostering the use of quasi‐experimental designs for evaluating public health interventions: Insights from an mHealth project in Malawi. African Population Studies 2015;29:1607‐1627. 18. Medhanyie AA, Moser A, Spigt M, Yebyoa H, Little A, et al. Mobile health data collection at primary health care in Ethiopia: a feasible challenge. J Clin Epidemiol 2015;68:80‐86. 19. Medhanyie AA, Little A, Yebyo H, Spigt M, Tadesse K, et al. Health workers’ experiences, barriers, preferences and motivating factors in using mHealth forms in Ethiopia. Hum Resour Health. 2015;13:2. 20. Asiimwe C, Gelvin D, Lee E, Ben Amor Y, Quinto E, et al. Use of an innovative, affordable, and open‐ source short message service–based tool to monitor malaria in remote areas of Uganda. Am J Trop Med Hyg 2011;85:26–33. 21. Chopra M, Mason E, Borrazzo J, Campbell H, Rudan I, et al. Ending of preventable deaths from pneumonia and diarrhoea: an achievable goal. Lancet 2013;381:1499‐1506. 22. Lavis JN, Posada FB, Haines A, Osei E. Use of research to inform public policymaking. 2004;364:1615‐21.

15 Chapter 1

16 PART II

The three phases of delay in care

17

18 CHAPTER 2

Associations between caregivers’ knowledge and care seeking behaviour for children with suspected pneumonia in six sub‐Saharan African countries

Noordam AC, Sharkey AB, Hinssen P, Dinant GJ, Cals JWL Submitted

19 Chapter 2

ABSTRACT

Pneumonia is the main cause of child mortality world‐wide and most of these deaths occur in sub‐Saharan Africa (SSA). Treatment with effective antibiotics is crucial to prevent these deaths; nevertheless only 2 out of 5 children with symptoms of pneumonia are taken to an appropriate care provider in SSA. While various factors associated with care seeking have been identified, the relationship between caregivers’ knowledge of danger signs and actual care seeking for their child with symptoms of pneumonia is not well researched. Based on data from Multiple Indicator Cluster Surveys, we assessed the association between caregivers’ knowledge of symptoms related to pneumonia – namely fast or difficulty breathing – and care seeking behaviour for these symptoms. We analysed data of 4,163 children with symptoms of pneumonia and their caregivers. Across all 6 countries only around 30% of caregivers were aware of at least one of the two symptoms. We found no association between caregivers’ knowledge of danger signs and actual care seeking for their child with symptoms of pneumonia in Central African Republic, , Malawi, and Sierra Leone. Our study shows that in the Democratic Republic of the Congo and the association between knowledge and care seeking was significant (P≤0.01), even after adjusting for key variables (including wealth, residence, education). These findings reveal an urgent need to increase community awareness of danger signs and symptoms of pneumonia, while simultaneously designing context specific strategies to address the fundamental challenges associated with timely care seeking.

20 Associations between knowledge and care seeking behaviour

INTRODUCTION

Pneumonia is responsible for more deaths among children under five years of age than any other infectious disease. In 2015, pneumonia killed an estimated 922,000 children under‐five globally; most of these deaths were in sub‐Saharan Africa.1 Timely treatment with effective antibiotics is critical to prevent pneumonia‐related deaths.2 Unfortunately, early care seeking for childhood illnesses remains a challenge in countries with high mortality rates.3‐6 Estimates from sub‐Saharan Africa indicate that only 2 out of the 5 children with pneumonia specific symptoms are taken to an appropriate provider for care.1 Key factors associated with care seeking include household income,3,6‐7 education levels of the primary caregiver,7 limited caregivers’ recognition of the disease,8 younger age and sex of the child,7 geographical areas (including rural locations)3 and .9 In addition, a systematic review of studies conducted across sub‐Saharan Africa identified cultural beliefs and illness perceptions, perceived severity of the illness, previous experience with health services, habit, treatment costs and efficacy, and gender as key factors for care seeking.3 As failure to recognize an illness is expected to leading to delays in care seeking, the first step to address pneumonia specific mortality is by ensuring that caregivers are aware of pneumonia specific symptoms.10 To date, the relationship between specific knowledge of these symptoms, recognition of them and related care seeking is not well researched.8 And, only a few studies specifically focus on the association between knowledge of symptoms related to pneumonia – namely fast or difficulty breathing – and care seeking behavior.11,12 We hypothesize that knowledge of these specific symptoms will enable caregivers to recognize an illness, consequently seeking timely and appropriate care. Such information is critical to plan effective strategies to reduce pneumonia mortality in high‐burden settings, particularly where rates of care seeking are inadequate.

METHODOLOGY

Data sources Our analyses are based on data from Multiple Indicator Cluster Surveys (MICS) conducted in sub‐Saharan Africa. These nationally representative household surveys are conducted by national implementing agencies with the support of the United Nations Children’s Fund (UNICEF). MICS surveys collect statistically sound and internationally comparable data on a variety of topics related to maternal and child

21 Chapter 2 health, including knowledge of symptoms (also referred to as danger signs) of childhood illnesses and care seeking behaviour for children under the age of five years.13.14 Sub‐Saharan African countries were selected for inclusion in this study if they had a MICS conducted during or after 2010, the data was national representative, and the datasets were available upon the start of our analyses (August 2015). Information on knowledge of symptoms of childhood illnesses is obtained during interviews using the ‘individual women’s’ questionnaire, which is administered to women age 15 through 49 years. Data on care seeking behaviour are obtained via the ‘children under‐ five’ questionnaires, which is administered primarily to mothers of children under the age of five years. When the mother is deceased or is living elsewhere, the questionnaire is administered to the child’s primary caregiver.

Survey methods and ethics Sampling frames are usually based on the most recent national census and therefore do not include non‐household populations; i.e. they exclude populations living in group quarters (e.g. hospitals, military barracks) and those living on the street. Usually, a two‐ stage cluster sampling approach is used; the first stage: select enumeration areas and do a listing of households, second stage: select households from list. While the surveys are conducted in different countries ‐ and may therefore vary due to limitations in costs and practical considerations (including security) ‐ all surveys ‘adhere to the fundamentals of scientific sampling, including complete coverage of the targeted population, use of suitable sample size, the need to conduct household listing and pre‐ selection of sample households’.13 Implementing agencies are required to obtain ethical approval as abide by the laws of the country. Survey tools, datasets and more detailed information on country specific survey methods can be found on the MICS website.14

Research questions For each country, we first assessed the proportion of caregivers (either mothers, or primary caregivers) of children under‐five that mentioned one or both symptoms of childhood illnesses linked to pneumonia, i.e. fast and/ or difficulty breathing, as a reason to seek care. The specific interview question asked to caregivers is the following open‐ended question: “Sometimes children have severe illnesses and should be taken immediately to a health facility. What types of symptoms would cause you to take a child under the age of 5 to a health facility right away?” A subsequent probe question asked: “Any other symptoms?” The responses were categorized as: child is not able to drink or breastfeed; becomes sicker; develops a fever; has fast breathing; has difficulty in breathing; has blood in stools; is drinking poorly; any other (unspecified); and some countries included categories such as diarrhoea and/ or vomiting. We classified caregivers who were able to identify either fast or difficulty breathing as having

22 Associations between knowledge and care seeking behaviour knowledge of pneumonia symptoms. As responses were not categorized as “cough” or “chest in‐drawing,” we were not able to include these symptoms in our analysis of pneumonia specific knowledge. Second, we calculated how many of these caregivers reported that their child had a cough and fast or difficulty breathing due to a problem in the chest in the past two weeks, as cases with symptoms of pneumonia. The interview questions related to these cases are: “Has (name) had an illness with cough at any time in the last 2 weeks?” “When (name) had an illness with cough, did he/she breathe faster than usual with short rapid breaths or have difficulty breathing?” “Was the fast or difficult breathing due to a problem in the chest or to a blocked or runny nose?” The children of whom the caregiver reported that they had a cough with fast or difficulty in breathing, due to a problem in the chest in the past two weeks, where considered as cases with symptoms of pneumonia. Third, we assessed which proportion of these caregivers brought their child to an appropriate health provider. The interview questions related to care seeking are: “Did you seek advice or treatment for the illness from any source?” “Where did you seek advice or treatment?” A subsequent probe asked, “Anywhere else?” We defined ‘appropriate’ health care provider as one working at either a private or public hospital, primary health care facility or any other government service, and who has undergone formal training and received accreditation, authorizing them to treat children with signs of acute respiratory tract infections.7 Once we knew which proportion sought appropriate care, we assessed if the caregivers who mentioned at least one pneumonia specific symptom as reasons to seek immediate care – i.e. fast or difficulty breathing – were indeed more likely to seek care from these providers, as supposed to those who did not mention one of these symptoms.

Data analysis We conducted our analyses using SPSS version 21. The data analysis was conducted by two independent researchers [CN and PH]. We first created a new dataset by merging the two (i.e. the ‘individual women’s’ and ‘children under‐five’) data files, matching eligible cases for both surveys based on cluster, household and line number. For caregivers with more than one child under the age of five, we used the files of the youngest child ‐ in the case of twins we kept the child that was mentioned last. We weighted the data using the sample weight variables. Cases with missing data (i.e. surveys which were party completed, as well as missing variables needed for these analyses) were excluded from the analyses. We calculated the percentage of caregivers included in the merged dataset who reported fast or difficulty breathing as reasons to

23 Chapter 2 seek immediate care from a health facility. We then calculated cross tabulations and performed chi‐square tests to assess the association between care seeking and knowledge of at least one symptom (i.e. fast or difficulty breathing). Based on the literature on care seeking, we included the following variables in the multivariate analyses: household wealth quintile (poorest, poorer, middle, richer and richest); residence (rural, urban); caregiver’s age (15‐19, 20‐24, etc.) and their level of education (none, primary, and secondary or higher); child age (< 2 years, 2‐5 years) and their sex (boy, girl); and the total number of children ever born (<2, 2‐3 and 4+). When possible we also adjusted for geographical location (regions) and religion. With the categories for the last two variables being country specific, not always included in the datasets, and – in some cases – sample size restrictions, we pre‐defined some parameters for these variables when available in the dataset: 1) the sample has to be large enough (≥10 cases per category); 2) geographical location has to refer to regions e.g. the North, East, South or West and not to any other groupings (i.e. provinces, or districts), and 3) we first assess if we can include geographical location (regions), after which we conduct the same assessment for religion. We performed a multivariate logistic regression for the dependent variable (care seeking from an ‘appropriate’ health provider) and the independent variables, in order to examine the association. We then calculated the adjusted odds ratios (ORs) with corresponding 95% confidence intervals (CIs).

RESULTS

Of countries in sub‐Saharan Africa, in which a MICS was conducted during or after 2010 and for which survey results were available before the start of our analyses, we selected the six countries with the largest sample sizes for analysis: Central African Republic (CAR), Chad, Democratic Republic of the Congo (DRC), Malawi, Nigeria, and Sierra Leone. Four surveys were from 2010 (CAR, Chad, DRC and Sierra Leone), Nigeria’s survey was from 2011 and Malawi’s was from 2013‐14.

Background characteristics Table 2.1 shows the samples per country, after merging the women’s and children’s datasets, with Nigeria as the largest sample (N=16,242) and Sierra Leone as the smallest (N=6,033). The vast majority of caregiver‐child combinations included in the analyses across the six countries live in rural settings, and most of the caregivers have at least four children.

24 Associations between knowledge and care seeking behaviour

of

n 963 813 925 556

cases 1349 1292 1245 1185 4349 1684 4295 1119 2090 2824 3058 2975 3031 2998 6033

calculation Leone ‐

re

reported

Sierra % 9.2

22.4 21.4 20.6 19.6 16.0 72.1 27.9 71.2 13.5 15.3 18.5 34.6 46.8 50.7 49.3 50.3 49.7 and manual

size

n 607 3697 3377 3028 3071 3070 5054 6828 3202 6212 2551 5189 8502 9858 6384 8341 7901

reasons,

11188 16242

sample

these

Nigeria

total

% 3.7 For

22.8 20.8 18.6 18.9 18.9 68.9 31.1 42.0 19.7 38.2 15.7 31.9 52.3 60.7 39.3 51.4 48.6

the

differ).

and data. will

(n),

n 3079 3014 2851 2522 2486 1814 1670 9702 2574 2754 4947 6250 7015 6936 7058 6892 1088

12137 13951 missing

category

for

numbers Malawi this

% 7.8

22.1 21.6 20.4 18.1 17.8 87.0 13.0 12.0 69.6 18.4 19.7 35.5 44.8 50.3 49.7 50.6 49.4 and

within

adjusted (%)

also

n 473

1600 1484 1459 1394 1185 5238 1884 1715 3044 2363 1205 2173 3744 4687 2436 3575 3548 7123 missings

are

the percentage

DRC in as

which

% 6.6

22.5 20.8 20.5 19.6 16.6 73.5 26.5 24.1 42.7 33.2 16.9 30.5 52.6 65.8 34.2 50.2 49.8

analysis, category

averages

this

n weeks. 763 947 each

in 1897 1992 2078 2123 1784 7755 2118 7289 1821 1403 2728 5742 5820 4053 4920 4954 9873

for

two

weighted Chad

vary

past

on included

% 7.7 9.6

27.6 41.1 21.0 21.5 18.1 50.2 20.2 58.2 14.2 21.5 18.4 19.2 78.5 73.8 49.8 58.9 will the

in

based size

are

individuals n

492

1093 2378 2580 1389 1328 1151 3372 1478 3182 1261 2450 2905 6821 1475 4371 2823 3449 4241 sample all

of pneumonia

CAR total of

the % percentages 7.2

16.0 34.9 37.8 20.4 19.5 16.9 49.4 21.7 46.6 18.5 35.9 42.6 21.6 64.1 41.4 50.6 62.2 and (i.e.,

symptoms

characteristics

with

numbers

differences

cases

size

children Background +

slight children

level child

of

index

the calculations,

sample reported show

age years 2.1

years

of 3 5 1

all ‐ ‐ ‐

Secondary Poor Middle Rich Richest Urban Primary 2 4+ 2 Female Poorest Rural None 0 <2 Male Residence Number Education Child Sex Total Total Wealth

might Table For

25 Chapter 2

The largest differences across the countries are found in levels of education, with most caregivers having had no education in Chad (73.8%), Sierra Leone (71.2%) and Nigeria (42.0%) and at least primary school in Malawi (69.6%), DRC (42.7%) and CAR (42.6%). The percentage of cases with symptoms of pneumonia (i.e. those for whom the mother or caregiver mentioned that their child had a cough and fast and/or difficulty breathing due to a problem in the chest in the past two weeks) ranged from 3.7% (n=607) in Nigeria to 9.6% (n=947) in Chad (Table 2.1). Hence, in the total population we had 4163 cases with symptoms of pneumonia.

Knowledge of fast and/or difficulty in breathing Of the two symptoms linked to pneumonia, caregivers were most aware of the symptom ‘difficulty in breathing’, ranging from 17.4% of the sample in Chad to 24.0% in CAR (Table 2.2). Across all 6 countries around 30% (ranging from 29.2‐32.9%) of caregivers were aware of at least one of the two symptoms. We characterised these caregivers as caregivers with appropriate knowledge of danger symptoms for pneumonia, see Table 2.2. When assessing the percentage of caregivers who mentioned both symptoms, this ranged from 4.5% in Chad to 11.2% in CAR.

Table 2.2 Percentage of caregivers surveyed who had knowledge of pneumonia specific danger signs. CAR Chad DRC Malawi Nigeria Sierra Leone Symptoms mentioned as reason to seek care Difficulty breathing 24.0% 17.4% 20.6% 19.9% 22.7% 21.7% Fast breathing 17.0% 16.3% 16.0% 14.8% 19.5% 19.5% Fast and difficulty in breathing 11.2% 4.5% 6.8% 5.1% 10.3% 8.3% Fast or difficulty in breathing (defined as 29.7% 29.2% 29.7% 29.6% 31.9% 32.9% “knowledge” in this paper) Total number of survey sample 6821 9873 7123 13951 16242 6033

Care seeking behaviour Table 2.3 shows the associations between knowledge of at least one symptom for pneumonia and care seeking from appropriate health care providers, also with adjustments for the predefined variables in a multivariate logistic regression model. Of children with symptoms of pneumonia in the previous two weeks, those living in Sierra Leone and Malawi were most likely (73.2% and 68.8%, respectively) to be brought to an appropriate provider. In Chad and CAR this was only 27.4% and 30.9%, respectively. Care seeking was only moderately better in Nigeria (41.9%) and DRC (44.2%).

26 Associations between knowledge and care seeking behaviour

Table 2.3 Associations between knowledge of at least one symptom for pneumonia and care seeking from appropriate health care providers.

Children with symptoms of pneumonia, taken to an appropriate provider Total taken to provider N (%) Unadjusted OR + (95% Adjusted1 OR + (95% CI) and reported CI) knowledge of danger signs CAR Total 151/ 489 (30.9) Knowledge ‐ No 104/ 357 (29.1) ref ref Knowledge ‐ Yes 47/ 132 (35.6) 1.3 (0.9‐2.1) 1.3 (0.9 – 2.2) Chad Total 254/ 928 (27.4) Knowledge ‐ No 166/ 630 (26.3) ref ref Knowledge ‐ Yes 88/ 298 (29.5) 1.2 (0.9 – 1.6) 1.1 (0.8 – 1.6) DRC Total 209/ 473 (44.2) Knowledge ‐ No 120/ 309 (38.8) ref ref Knowledge ‐ Yes 89/ 164 (54.3) 1.9 (1.3 – 2.7)** 2.0 (1.3 – 3.0)** Malawi Total 740/ 1076 (68.8) Knowledge ‐ No 519/ 743 (69.9) ref ref Knowledge ‐ Yes 221/ 333 (66.4) 0.9 (0.6 – 1.1) 0.8 (0.6 – 1.1) Nigeria Total 253/ 604 (41.9) Knowledge ‐ No 155/ 405 (38.3) ref ref Knowledge ‐ Yes 98/ 199 (49.2) 1.6 (1.1‐2.2)** 1.9 (1.3 – 2.9)** Sierra Total 402/ 549 (73.2) Leone Knowledge ‐ No 267/ 369 (72.4) ref ref Knowledge ‐ Yes 135/ 180 (75.0) 1.1 (0.8 – 1.7) 1.1 (0.8 – 1.9) Knowledge is defined as those caregivers aware of at least one pneumonia symptom (i.e. fast or difficulty breathing). All calculations: numbers (n), percentages (%), odds ratios (OR) and 95% confidence intervals (CI) are based on weighted averages, adjusted for missing data. Numbers presented in this table are rounded. Manual re‐calculation might therefore show slight differences. 1Adjusted for pre‐defined variables, namely; wealth, residence, maternal education, maternal age, child’s age, child’s sex, and total number of children ever born. Based on the pre‐defined criteria geographical location (defined as regions) were included in Malawi, Nigeria and Sierra Leone and religion in Chad, Malawi, and Nigeria. Statistical significance: *p≤0.05, **p≤0.01

In DRC, a child whose caregiver mentioned at least one symptom related to pneumonia as reasons to seek immediate care was 1.9 times more likely to have been brought to an ‘appropriate’ health provider during his or her last illness than a child whose caregiver did not mention one of the symptoms (95% CI = 1.3‐2.7, P≤0.01). The association remained largely the same after adjusting for wealth, residence, maternal age and education, the age and sex of the child, and number of children ever born

27 Chapter 2

(OR=2.0, 95% CI = 1.3‐3.0, P≤0.01). A significant association was also found in Nigeria, where caregivers with knowledge of symptoms were 1.6 times more likely to seek care from an ‘appropriate’ health provider (95% CI = 1.1‐2.2, p≤0.01). After adjusting for the pre‐defined variables as mentioned for DRC, and for religion (Muslim, Christian) and geographical location (6 regions; north‐east, north‐west, south‐south, south‐east, south‐west, north‐centre) the OR increased to 1.9 (95% CI = 1.3‐2.9, p≤0.01). For the other four countries, the analyses reveal no significant association between knowledge and care seeking.

DISCUSSION

In this study, we found low levels of caregivers knowledge of symptoms related to childhood pneumonia in sub‐Saharan Africa; on average only around 30% of caregivers were aware of either fast or difficulty in breathing as a reason to seek care. These caregivers, who were aware of at least one of these symptoms, were not necessarily more likely to seek care from an appropriate health care provider. We found a significant association between knowledge of at least one symptom and care seeking in DRC and Nigeria. This association was evident even after adjusting for key background characteristics. For the other four countries (CAR, Chad, Malawi and Sierra Leone) the associations were not significant. Our analyses also reveal large differences in care seeking for children with symptoms of pneumonia, ranging from 27% in Chad to 73% in Sierra Leone.

Low levels of knowledge of pneumonia specific symptoms are confirmed by other studies, even compared to knowledge of other illnesses such as malaria and diarrhoea.15,16 Of the two symptoms, we found that caregivers more frequently mentioned difficulty in breathing as compared to fast breathing. A study conducted in Nigeria found that mothers were more likely to recognise pneumonia based on fever and cough, than on either fast or difficulty breathing.17 This same study also found that mothers were less likely to recognize pneumonia based on more severe symptoms such as chest in‐drawing and central cyanosis.17 Studies conducted in Sierra Leone18 and Nigeria11 also report a lack of knowledge on symptoms and that caregivers are often not aware of ways to prevent or treat common childhood illnesses, including pneumonia.

Other studies also conclude that care seeking patterns vary between and within countries, as they are linked to a dynamic process influenced by a range of socio‐ economic, cultural and geographic access factors.3,7,19‐21 More specifically, a study from

28 Associations between knowledge and care seeking behaviour

Malawi confirms that wealth, urban‐rural residence, maternal education ‐ amongst others ‐ were associated with care seeking.22 Studies also report that care seeking is linked not only to the ability of caregivers to recognize an illness, but also to caregivers’ perception of the severity of the illness: the more severe caregivers’ perceived the illness, the more likely they were to seek care.8,23 The complexity of factors that influence care seeking behaviour could explain the fact that we found no significant association between knowledge of symptoms and care seeking behaviour in most of the countries included in these analyses. While this is concerning, other studies have shown that knowledge alone is insufficient to change behaviour; e.g. knowledge of the importance of using a condom to reduce the risk of HIV transmission, does not necessarily translate to higher use of condoms amongst those most at risk.24 Other studies also found that community based approaches such as provision of integrated case management of childhood illnesses ‐ including pneumonia ‐ can be an important strategy.25,26 A study from Nigeria27 showed that community information activities for malaria lead to improved knowledge, home management and referral practices.

Limitations There are some limitations related to these analyses. The children with symptoms of pneumonia are based on caregivers’ perceptions of symptoms and their ability to recall events, which could lead to incorrect estimates.28 In relation to this, we were not able to assess if knowledge of pneumonia specific symptoms lead to more accuracy in the recognition of pneumonia specific symptoms, subsequently leading to better identification of suspected cases; or, as the cases are not clinically validated as those requiring medical care, not seeking care might in some cases be justified. Also, we were not able to disaggregate the data by severity of the illness, or by the severity of symptoms (e.g. chest in‐drawing). And, as caregivers were not prompted about specific symptoms, it is possible that pneumonia specific symptoms just did not come to mind when interviewed, this could have led to under reporting of these specific symptoms. Finally, we were not able to discern where caregivers learned about symptoms of childhood illnesses and whether or not they learned this after care seeking for the child’s last illness. A study from Nigeria29 showed that only 23% of the caregivers received health information from the healthcare providers after being hospitalized.

Conclusion We hypothesized that knowledge of disease specific symptoms improves the ability of caregivers to recognize an illness and seek appropriate care and found low levels of knowledge on pneumonia among caregivers in six high burden countries in sub‐Saharan Africa. However, we found limited associations between knowledge and care seeking in

29 Chapter 2 this study, which suggests that knowledge alone is not sufficient to take action. With low levels of knowledge of symptoms of pneumonia across these high pneumonia mortality settings, emphasis should be put on education programs, which not only focus on the primary caregivers, but on all those involved in decision‐making processes and care seeking. In addition, factors other than knowledge (e.g. empowerment, costs) should also be addressed to improve care seeking behaviour. In other words, there is a need to increase community awareness of pneumonia, while simultaneously designing context specific strategies to address the fundamental challenges associated with timely care seeking.

Acknowledgement The authors would like to express their gratitude to colleagues in UNICEF for reviewing the manuscript.

30 Associations between knowledge and care seeking behaviour

REFERENCES

1. United Nations Children's Fund (UNICEF) global databases 2015. Available: http://data.unicef.org/child‐ health/pneumonia.html. Accessed 2016, Jan 12. 2. Kallander K, Young M, Qazi S. Comment. Universal access to pneumonia prevention and care: a call for action. Lancet Respir Med 2014;2:950‐952. 3. Colvin CJ, Smith HJ, Swartz A, Ash JW, de Heer J, et al. Understanding careseeking for child illness in sub‐Saharan Africa: A systematic review and conceptual framework based on qualitative research of household recognition and response to child diarrhoea, pneumonia and malaria. Soc Sci Med 2013;86:66e78. 4. Mosites EM, Matheson AI, Kern E, Manhart LE, Morris SS, et al. Care‐seeking and appropriate treatment for childhood acute respiratory illness: an analysis of Demographic and Health Survey and Multiple Indicators Cluster Survey datasets for high mortality countries. BMC Public Health 2014;14:446. 5. Herbert HK, Lee ACC, Chandran A, Rudan I, Baqui AH. Care seeking for neonatal illness in low‐ and middle‐income countries: a systematic review. PLoS Med 2012;9:e1001183. 6. Barros AJD, Ronsmans C, Axelson H, Loaiza E, Bertoldi AD, et al. (2012) Equity in maternal, newborn, and child health interventions in Countdown to 2015: a retrospective review of survey data from 54 countries. Lancet 2012;379:1225‐1233. 7. Noordam AC, Carvajal‐Velez L, Sharkey AB, Young M, Cals JWL. Care Seeking Behaviour for Children with Suspected Pneumonia in Countries in Sub‐Saharan Africa with High Pneumonia Mortality. PLoS One 2015;10:e0117919. 8. Geldsetzer P, Williams TC, Kirolos A, Mitchell S, Ratcliffe LA, et al. The Recognition of and Care Seeking Behaviour for Childhood Illness in Developing Countries: A Systematic Review. PLoS One 2014;9:e93427. 9. Mebratie AD, Van de Poel E, Yilma Z, Abebaw D, Alemu G, et al. Healthcare‐seeking behaviour in rural Ethiopia: evidence from clinical vignettes. BMJ Open 2014;4:e004020. 10. UNICEF and World Health Organization (WHO) (2006) Pneumonia the forgotten killer of children. Available: http://www.childinfo.org/files/Pneumonia_The_Forgotten_Killer_of_Children.pdf. Accessed 2016 Feb 22. 11. Ndu IK, Ekwochi U, Osuorah CDI, Onah KS, Obuoha E, et al. Danger signs of childhood pneumonia: Caregiver awareness and care seeking behavior in a developing country. Int J Pediatr. 2015;2015:167261. 12. Tuhebwe D, Tumushabe E, Leontsini E, Wanyenze RK. Pneumonia among children under five in Uganda: symptom recognition and actions taken by caretakers. Afr Health Sci. 2014;14:993‐1000. 13. Hancioglu A, Arnold F. Measuring coverage in MNCH: tracking progress in health for women and children using DHS and MICS household surveys. PLoS Med 2013;10:e1001391. 14. UNICEF Statistics and Monitoring 2015. Multiple Indicator Cluster Survey (MICS) Available at: http://mics.unicef.org. Accessed 2016, Feb 2. 15. Bedford KJA, Sharkey AB. Local Barriers and Solutions to Improve Care‐Seeking for Childhood Pneumonia, Diarrhoea and Malaria in Kenya, Nigeria and Niger: A Qualitative Study. PLOS One 2014;9: e100038. 16. Ukwaja KN, Talabi AA, Aina OB. Pre‐hospital care seeking behaviour for childhood acute respiratory infections in south‐western Nigeria. International Health 2012;4:289‐294. 17. Uwaezuoke SN, Emodi IJ, Ibe BC. Maternal perception of pneumonia in children: a health facility survey in Enugu, eastern Nigeria. Ann Trop Paediatr 2002;22:281‐285. 18. Kanu JS, Tang Y, Liu Y. Assessment on the Knowledge and Reported Practices of Women on Maternal and Child Health in Rural Sierra Leone: A Cross‐Sectional Survey. PLoS One 2014;9: e105936. 19. Diaz T, George AS, Roa SR, Bangura PS, Baimba JB, et al. Healthcare seeking for diarhoea, malaria and pneumonia among children in four poor rural districts in Sierra Leone in the context of free health care: results of a cross‐sectional survey. BMC Public Health 2013;13:157 20. Hodgins S, Pullum T, Dougherty L. Understanding where parents take their sick children and why it matters: a multi‐country analysis. Glob Health Sci Pract 2013;1:328‐356.

31 Chapter 2

21. Prach LM, Treleven E, Isiguzo C, Liu, J. Care‐seeking at patent and proprietary medicine vendors in Nigeria. BMC Health Serv Res 2015;15:231. 22. Oyekale AS. Assessment of Malawian mothers’ malaria knowledge, healthcare preferences and timeliness of seeking fever treatments for children under five. Int J Environ Res Public Health 2015;12:521‐540. 23. Burton DC, Flannery B, Onyango B, Larson C, Alaii J, et al. Healthcare‐seeking behaviour for common infectious disease‐related illnesses in rural Kenya: a community‐based house‐to‐house survey. J Health Popul Nutr 2011;29:61‐70. 24. Olowookere SA, Adeleke NA, Fatiregun AA, Abioye‐Kuteyi EA. Pattern of condom use among clients at a Nigerian HIV Counseling and Testing Centre. BMC Research Notes 2013;6:289. 25. Das JK, Lassi ZS, Salam RA, Bhutta ZA. Effect of community based interventions on childhood diarrhea and pneumonia: uptake of treatment modalities and impact on mortality. BMC Public Health 2013;13:S29. 26. Bhutta ZA, Das JK, Walker N, Rizvi A, Campbell H, et al. Interventions to address deaths from childhood pneumonia and diarrhea equitably: what works and at what cost? Lancet 2013;381: 1417‐1429. 27. Okeke, TA. Improving malaria recognition, treatment and referral practices by training caretakers in rural Nigeria. J Biosoc Sci 2010;42:325‐339. 28. Hazir T, Begum K, el Arifeen S, Khan AM, Huque MH, et al. Measuring coverage in MNCH: A prospective validation study in Pakistan and Bangladesh on measuring correct treatment of childhood pneumonia. PLoS Med 2013;10:e1001422. 29. Nwaneri UD, Oviawe OO, Oviawe, O. Do caregivers receive health information on their children’s illnesses from healthcare providers while hospitalized? Niger Postgrad Med J 2014;21: 279‐284.

32 CHAPTER 3

Care seeking behaviour for children with suspected pneumonia in countries in sub‐Saharan Africa with high pneumonia mortality

Noordam AC, Carvajal‐Velez L, Sharkey AB, Young M, Cals JWL PLoS One 2015;10:e0117919

33 Chapter 3

ABSTRACT

Pneumonia is the leading cause of childhood mortality in sub‐Saharan Africa (SSA). Because effective antibiotic treatment exists, timely recognition of pneumonia and subsequent care seeking for treatment can prevent deaths. For six high pneumonia mortality countries in SSA we examined if children with suspected pneumonia were taken for care, and if so, from which type of care providers, using national survey data of 76530 children. We also assessed factors independently associated with care seeking from health providers, also known as ‘appropriate’ providers. We report important differences in care seeking patterns across these countries. In Tanzania 85% of children with suspected pneumonia were taken for care, whereas this was only 30% in Ethiopia. Most of the children living in these six countries were taken to a primary health care facility; 86, 68 and 59% in Ethiopia, Tanzania and Burkina Faso respectively. In Uganda, hospital care was sought for 60% of children. 16‐18% of children were taken to a private pharmacy in Democratic Republic of Congo (DRC), Tanzania and Nigeria. In Tanzania, children from the richest households were 9.5 times (CI 2.3‐39.3) more likely to be brought for care than children from the poorest households, after controlling for the child’s age, sex, caregiver’s education and urban‐rural residence. The influence of the age of a child, when controlling for sex, urban‐rural residence, education and wealth, shows that the youngest children (<2 years) were more likely to be brought to a care provider in Nigeria, Ethiopia and DRC. Urban‐rural residence was not significantly associated with care seeking, after controlling for the age and sex of the child, caregivers education and wealth. The study suggests that it is crucial to understand country‐specific care seeking patterns for children with suspected pneumonia and related determinants using available data prior to planning programmatic responses.

34 Care seeking behaviour for children with suspected pneumonia

INTRODUCTION

Acute respiratory infections (ARIs) are the most common illnesses in childhood, of which lower respiratory tract infections (LRTIs) are the most severe in developing countries.1 Pneumonia, a common and severe LRTI, was responsible for 15% of all deaths among children under‐five in sub‐Saharan Africa (SSA) in 2013 and most of these deaths were concentrated in a few countries.2‐4 Cough and fast and/ or difficult breathing (i.e. tachypnea and/or dyspnea) due to a problem in the chest are clinically recognized as signs of childhood pneumonia.5 Effective antibiotic treatment for pneumonia exists, and therefore timely recognition of these signs and symptoms by primary caregivers and subsequent care seeking for treatment from ‘appropriate’ providers can prevent many of these deaths.6 Nevertheless, only 50% of children in SSA with suspected pneumonia were taken for care in 2010.7 Caregivers may not seek care for myriad reasons: both financial (e.g., the cost of services or treatment, transportation costs, loss of wages) and non‐financial (e.g., gender and social norms, insufficient knowledge of danger signs and illness severity, and previous experiences with health services).8‐11 Further analysis of care seeking behaviours by primary caregivers, and on child, caregiver and household characteristics associated with care seeking is needed to further optimise future strategies within integrated approaches to prevent and treat childhood pneumonia.12 We examined care seeking behaviour by caregivers of children under‐five years of age with suspected pneumonia in sub‐Saharan countries with high rates of childhood pneumonia mortality, and examined to what extend caregivers and household characteristics influenced care seeking.

METHODOLOGY

Data sources We analysed data from the Demographic and Health Surveys (DHS) or Multiple Indicator Cluster Surveys (MICS) of countries in sub‐Saharan Africa identified by the Global Action Plan for Pneumonia and Diarrhoea (GAPPD) as being among those with the highest burden for pneumonia.12 DHS and MICS surveys are typically conducted by government statistics agencies every 3 to 5 years, with the support and technical assistance of the United States Agency for International Development (USAID) and the United Nations Children's Fund (UNICEF) respectively. Both surveys are relatively similar in content and scope and have comparable results. DHS and MICS survey programs enable low‐and middle‐income countries (LMICs) to produce estimates of a range of indicators in the areas of health, education, child protection and HIV & AIDS.

35 Chapter 3

DHS and MICS work together to harmonize tools and methods to enable comparisons of key indicators across countries and over time. Both surveys adhere to the fundamentals of scientific sampling, including complete coverage, suitable sample sizes, pre‐selection of sample households, and sample documentation. However, limitations due to cost or other practical considerations, such as security, might result in some inconsistencies.13 These survey data are available publically at www.dhsprogram.com and www.data.unicef.org respectively. Countries were considered eligible for inclusion in this analysis if: (1) a population‐ based survey was conducted during or after 2010, (2) the data was available at the start of our analysis (June 2013) and (3) the survey included the standard question on care seeking for children under‐five with suspected pneumonia (cough and rapid or difficulty breathing due to a problem in the chest) and whether or not the caregiver sought care and from where they sought care during the past two weeks. Questions relating to child’s health are included in the women’s questionnaire in DHS and in the questionnaire for under‐fives in MICS. In DHS, only mothers were interviewed, while in MICS the under‐five questionnaire is administered to either mothers or primary caregivers of children under‐five. The following questions (similar in DHS and MICS) were used for the analysis: 1. Has (NAME) had an illness with a cough at any time in the last 2 weeks? (Yes/ No/ Don’t know) 2. When (NAME) had an illness with a cough, did he/she breathe faster than usual with short, rapid breaths or have difficulty breathing? (Yes/ No/ Don’t know) 3. Was the fast or difficult breathing due to a problem in the chest or to a blocked or runny nose? (Problem in chest only/ Blocked or runny nose only/ Both/ Other (specify)/ Don’t know) 4. Did you seek advice or treatment for the illness from any source? (Yes/ No/ Don’t know) 5. A. Where did you seek advice or treatment? B. Anywhere else? (Various services which fall under the following categories: Public sector, Private sector, Other sources and Other (specify) ) In order to be included in this analysis, the respondent had to answer “yes” to questions 1, 2 and 4, and ‘a problem in the chest’ to question 3. Multiple answers were possible for question 5 (sections A and B).

Method of analysis We used recode manuals and guides from both DHS and MICS prior to analysis and recoding, which was conducted using STATA 12.1. We weighted the data according to sample size by using the samples weight variables (V005 in DHS and chweight in MICS)

36 Care seeking behaviour for children with suspected pneumonia provided in the dataset and explained in dataset guides. In case of missing data, the cases were excluded from the analysis. Data analysis was cross‐checked by an independent researcher. To analyse the rate of care seeking for pneumonia, we first calculated how many children under‐five had suspected pneumonia (cough and rapid or difficulty breathing due to a problem in the chest), and then the percentage of those taken for care. For those taken for care, we then categorised them into health providers, also known as ‘appropriate’ providers (i.e., accredited by that country’s government authorities) or other provider also known as ‘non‐appropriate’ providers (i.e., those not accredited to provide antibiotics). These other providers, for this analysis, include private pharmacies, shops, and traditional healers amongst others, for a more detail see the subsection below ‘categorisation of health care providers and facilities’. In both surveys caregivers can report more than one source of care to which they brought their child with suspected pneumonia during the past two weeks. We calculated frequencies and cross tabulations and performed chi‐square (χ2) tests to identify variables associated with the dependent variable ‘care seeking behaviour from an ‘appropriate’ provider’. To determine the adjusted associations between child and caregiver characteristics and care seeking behaviour, we predefined the following groupings of independent variables: child’s age (grouped as <2 years and 2‐5 years), child’s sex (male, female), residential setting (urban, rural), primary caregiver’s or mother’s education (none, primary, secondary and higher) and household wealth quintile (poorest, poorer, middle, richer and richest), which is a composite measure of a household’s cumulative living standard used by DHS and other surveys in countries that lack reliable data on income and expenditures.14 We selected ‘appropriate’ providers as the dependent variable as they have undertaken formal training and are accredited to provide antibiotics and for that reason, we wanted to assess which factors are associated with seeking care from these providers. We performed a multivariate logistic regression model for the dependent variable in order to examine independence of associations (p≤0.05), with all predefined variables in the model. In addition, we calculated odds ratios (ORs) with corresponding 95% confidence intervals (CIs).

Categorization of health care providers and facilities When we refer to ‘any provider’, this includes all possible care providers (i.e. both ‘appropriate’ and ‘non‐appropriate’ providers). The definition of health providers, also known as ‘appropriate’ providers (and therefore referred to as such throughout this paper), varies among countries. However, this category includes both private and public providers who have undergone formal training and facilities that have received accreditation and are therefore authorized to treat children with signs of ARI, for example: ‐ Hospitals: For all countries this includes private and government hospitals.

37 Chapter 3

‐ Primary health care (PHC) facilities: This category includes both private and public health centres, clinics, dispensaries and posts and in some cases a private doctor or other medical personal, village/ community health workers, mobile (outreach) services as well as health facilities supported by non‐governmental organizations (NGOs). ‐ Unidentified other services: These are other government led facilities; however, they are not specified in the country specific databases. Other providers, also known as ‘non‐appropriate’ providers (and therefore referred to as such throughout this paper): include providers that are neither accredited nor authorized to prescribe antibiotics for children with signs of ARI. This varies by country, for this analysis, it includes private pharmacies, shops, and markets as well as relatives and those practicing traditional medicine or faith healing. It also includes private services which are unidentified in the country specific databases.

RESULTS

Of the ten highest burden countries initially considered for the analysis; six met the predefined inclusion criteria of having a DHS or MICS survey in 2010 or later (with data publically available for analysis) and including the standard indicator on suspected pneumonia and care seeking: Burkina Faso, Democratic Republic of Congo (DRC), Ethiopia, Nigeria, Tanzania and Uganda, see Table 3.1. Together, these six countries account for 297,000 deaths or 53% of childhood pneumonia mortality among children under‐five in sub‐Saharan Africa, and 26% of global childhood pneumonia mortality.15 The remaining four countries were excluded for the following reasons: Angola did not have a survey assessing the indicators of interest available, Kenya did not have a survey in 2010 or later and relevant survey data for Niger and Mali were not available at the time we started our analysis. Table 3.1 shows the under‐five mortality per 1,000 live births based on the latest estimates developed by the UN Inter‐agency Group for Child Mortality Estimates.4 The table also presents characteristics of the surveys in the included countries. Data of 76530 children were available for analysis. Sample sizes per country ranged from 7535 (Uganda) to 25192 (Nigeria).

Care seeking for suspected pneumonia The number of children under‐five included in the survey with suspected pneumonia, as reported by caregivers in the previous 2 weeks preceding the surveys, ranged from 267 in Burkina Faso to 1118 in Uganda in the adjusted country samples. In all countries, except Ethiopia, care was sought for the majority of children with suspected pneumonia.

38 Care seeking behaviour for children with suspected pneumonia

size

whom

Mortality

sought

suspected sample

from

for Child was

34.7% 29.9% 16.4% 70.5% 37.7% 15.0% with

for

care

no

adjusted

pneumonia Group Children

an

agency rounded, whom once. ‐

from

pneumonia.

provider

suspected are

Inter least from

at

40.3% 57.1% 78.9% sought 27.2% 39.7% 72.0% with UN

suspected care the

numbers was

by all

‘appropriate’ with pneumonia Children care

sought

surveys, who

children

developed

the

and

suspected and

6.3% 1.9% 7.0% 3.5% 4.3% 14.8% from

with Percentage

estimates

pneumonia

survey 2013

numbers respondents

in

of

the the pneumonia

on size

For (3) (0) (6) (6) (0) (6)

suspected

4 + + + + + +

(missing)

included

had

based 700 261 767 890 326 data.

sample 1115

who and

and Number

Children

year, births,

underlying

the

included

in

and the live

are

of 7667 7535

11093 14001 11042 25192 survey Children 1000

source

included per children quality

data

all

the per

five live ‐

on deaths

98 64 52 66 119 117 births analysis, are rely

seeking,

1000

Under mortality the

rate care

in

For

estimates

Year 2010 2010 2011 2011 2010 2011 Survey mortality data. included

five ‐ modelled

DHS MICS Data source DHS MICS DHS DHS missing

Countries under

these and

Faso

the

3.1

on

Country Burkina DRC Ethiopia Nigeria Tanzania Uganda Table Data Estimation; (weighted)

39 Chapter 3

Figure 3.1 shows the large variation in care seeking patterns for suspected pneumonia found across the six countries, juxtaposed with overall levels of pneumonia mortality. In Tanzania 85% of children with suspected pneumonia were taken to a provider by their caregiver, as opposed to only 30% in Ethiopia.

Figure 3.1 Care seeking for signs of childhood ARI by provider type, and overall level of pneumonia specific child mortality by country Sources: Care seeking data comes from DHS and MICS, mortality data comes from the United Nations Inter‐agency Group for Child Mortality Estimation (2013)

Differences are also seen across the countries in overall levels of care seeking from ‘appropriate’ providers (i.e. health care providers), ranging from 27% in Ethiopia to 79% in Uganda. In both these countries, when people do seek care they most often use ‘appropriate’ providers. Uganda and Tanzania were the only two countries where more than 70% of children were brought to an ‘appropriate’ provider. In Nigeria and DRC, overall care seeking is just over 60%, but care seeking from ‘appropriate’ providers stands at or just below 40%. All six countries present relatively high levels of pneumonia specific mortality. Among only those children for whom care was sought, Table 3.2 shows to which specific type of health care facilities and providers these children were brought. In Ethiopia, Tanzania and Burkina Faso children were most often taken to a primary health care (PHC) facility, which includes health centres, clinics and posts as well as community based and outreach services. Hospitals were also frequently accessed for care in Nigeria and Uganda; in fact, in Uganda children were more likely to be taken to a hospital than a PHC facility. With respect to ‘non‐appropriate’ or other providers, private pharmacies are frequently accessed in DRC, Tanzania and Nigeria (18, 18 and 16% respectively). Traditional practitioners and other providers (e.g. vendors, shops, churches, relatives or friends) are also often consulted in DRC and Nigeria in particular.

40 Care seeking behaviour for children with suspected pneumonia

for

1 4 1 1 0 6 2 2 ‘non

94 60 34 30 %

100 83.6 hospital,

adjusted

is

Uganda

7 0 identified ‐ n 35 12 11 61 19 18 999 938 339 596 301 932 un Data

government

‐ ‐ ‐ ‐

other and

48 20 82 68 15 18 18

% 100 rounded. 85.0

any

are

private

and a

Tanzania

to n 59 43 52 52

142 296 244 201 277 friend Numbers

taken

or

‐ ‐

‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐

2 7 7 3 4 6 care.

16 11 62 29 26 38 16 was %

100 62.3 for

relative child

a

Nigeria

if

n 93 11 43 66 40 19 23 94 33 brought

596 369 174 152 555 227 church, (e.g.

‐ ‐ ‐ ‐

‐ ‐ ‐ ‐ ‐ ‐

children

shop, 2 0 6 1 8 8 3 6 0

46 29 92 86 %

100 of 29.5

(%).

separately

vendor,

Ethiopia

4 0 7 0 2 number a

n 71 14 20 20 14

245 113 225 211 226 countries as

included the

is six

as

such

‐ ‐ ‐ ‐

the

1 0 3 6 visit

20 11 20 57 12 34 10 43 18 than %

100 65.3

each within

DRC provider higher

4 1 n 57 62 52 15 93 31 102 520 103 294 175 457 226 provider,

providers

therefore

care

6 7 8 appropriate’ are ‐ 1 59 79 20 59 21 %

100 and/or 70.1

one Faso

‘non

than

numbers

facilities

1 n 11 38 39 13 15 other Burkina

188 149 110 111 183

the

more

each any

to

to

to

care

‐ ‐ (i.e. (i.e.

taken

hospital),

visits

for visits

of worker

clinic care

‐ ‐

all were

under

‐ ‐

‐ ‐ ‐ ‐ other brought

health provider

number ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐

appropriate’

‐ Practitioner

children Includes

(outreach) medic

any a centre

included

a

Total ‐ ‐ ‐ ‐ provider. services Pharmacy

to ‘appropriate’ ‘non

‐ ‐ the

children

are

facility other

of

of provider

size.

providers Private Dispensary Mobile Unidentified Health Post Clinic Community NGO

‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐ visits visits visits %

3.2

visits Hospital PHC Any Private Traditional

some

Total Total

health) Total Total

other) both Table As sample appropriate’

41 Chapter 3

Table 3.3 shows the results of the multivariate logistic regression to examine potential associations between child or caregiver characteristics and care seeking behaviour from ‘appropriate’ providers. In Nigeria, the influence of the age of a child, when controlling for the other variables (i.e., sex, urban‐rural residence, education and wealth) shows that younger children (<2 years) were 1.7 times more likely to be brought to a care provider than children between 2 and 5 years of age (95% CI 1.1‐2.5, p<0.01). This significant association between younger age of the child and care seeking from ‘appropriate’ providers was also found for Ethiopia and DRC. In general, the influence of a caregiver’s education, when controlling for other variables shown in Table 3.3 shows that higher educated caregivers and higher wealth quintiles were also associated with higher levels of care seeking from ‘appropriate’ providers in this group of countries. In Uganda having at least a primary education, when controlling for age and sex of the child, urban‐rural residence and wealth, was associated with higher likelihood to seek care. In Tanzania, Ethiopia, Nigeria and Burkina Faso, caregivers from the richest quintile were much more likely as those from the poorest quintile to seek care for their child with suspected pneumonia, after controlling for age and sex of the child, urban‐rural residence and education, with odds ratios ranging from 4.7 (95% CI 1.5‐15.1) to 9.4 (95% CI 2.3‐39.3). Similar patterns did not emerge in the DRC and Uganda. With regard to the sex of a child, controlling for the child’s age, urban‐rural residence, education and wealth, girls were 1.7 times more likely to be brought to seek care in Uganda than boys (95% CI 1.2‐2.4, p<0.05). We found no statistically significant differences in care seeking patterns between rural and urban settings after controlling for the age and sex of a child, education and wealth.

42 Care seeking behaviour for children with suspected pneumonia

we **

CI)

1.6) 1.4) 1.5) 2.8) 1.5) 1.9) 2.8)* 3.7)* ‐ ‐ ‐ ‐ ‐ ‐

‐ ‐ 0.8) ‐

ref ref ref ref ref (95%

(0.6 (0.7 (0.5 (0.6 (0.5 (0.4

(1.1 (1.0 +

(0.4 rounded,

**p<0.01.

1.0 1.0 0.8 1.2 0.9 0.9 (79.0) 1.8 1.9 OR

0.6 are

Uganda table *p<0.05,

(79.2) (80.2) (78.7) (77.2) (83.3) (78.7) (82.3) (79.1) (74.9) (78.1) (80.8) (81.6) (78.3) (69.7)

% 880/1115

this

N in

207/261 603/752 435/553 124/161 447/537 766/974 140/170 445/562 433/578 148/190 114/141 164/201 261/334 113/162

significance:

CI)

presented 2.7) 3.0) 7.0) 2.6) 2.5) 3.2) 6.0)

8.5)* ‐ ‐ ‐ ‐ ‐ ‐ ‐

‐ 39.3)** ‐ ref ref ref ref ref (95%

(0.6 (0.8 (0.8 (0.7 (0.4 (0.4 (0.2

(1.2 + Statistical

(2.3

1.2 1.5 2.3 1.4 1.0 1.2 1.0 3.2 OR Numbers

9.5

cases.

data. Tanzania

(72.8) (72.1) (69.0) (67.1) (77.1) (74.9) (67.6)

(81.6) (93.4) (57.0) (57.8) (86.1) (83.4) (77.6) (64.7)

%

N 238

weighted ‐ missing

40/48 54/58 35/61 48/83 74/85 22/26 59/76 40/62 un

107/155 161/240 118/153 128/170 117/172 235/326 for 173/

49

provider.

25

CI)

2.3) 2.0) 1.4) 2.0) 2.7) 3.5) on ‐ ‐ ‐ ‐ ‐ ‐

2.5)** 3.8)** adjusted 15.7)** ‐ ‐

‐ ref ref ref ref ref (95%

(0.8 (0.8 (0.6 (0.6 (0.9 (0.8

+ (1.1 (1.2

also

(2.3

based

1.4 1.2 0.9 1.1 1.6 1.7 OR 1.7 2.2 6.0 are

(%)

N

Nigeria ‘appropriate’

which

(39.7) (35.4) (44.6) (36.1) (38.8) (40.8) (53.0) (57.9) (32.3)

(33.9) (40.8) (27.8) (47.2) (45.0)

(74.5) % cases.

any

N 890

67/90 averages 79/233 57/140 82/294 68/144 58/129

186/418 252/699 184/474 170/416 101/191 124/214 173/535 167/473 354/ from

weighted

un CI)

3.9) 1.6) 3.3) 1.2) 2.0) 2.8)* ‐ ‐ ‐ ‐ ‐

25 5.6)** weighted ‐

27.0)**

‐ seeking ref ref ref ref ref (95%

on (0.9 (0.5 (0.8 (0.5 (0.3

(1.1 +

(1.3

than

(3.1

1.9 0.9 1.7 0.8 0.8 1.8 OR 2.7 care 9.2 less

based )

‐ (

are Ethiopia )

and N

(27.2) ( (32.5) (25.4) (29.1) (25.2) (24.8)

(25.2) (27.9) (22.7) (15.5) (33.2) (22.5) (CI)

(61.8) (46.9) %

N 18/22 42/67 32/69 37/148 55/198 43/191 29/188 57/173 91/406 117/361 100/393 109/375 176/698 136/547 209/767 interval

differences.

CI)

4.5) 3.3) 1.9) 2.4) 1.2) 1.6) 2.2) 2.8) slight 2.3)* ‐ ‐ ‐ ‐ ‐ ‐ ‐ ‐

‐ characteristics

ref ref ref ref ref (95%

confidence (0.7 (0.7 (0.7 (0.5 (0.4 (0.6 (0.6 (1.0

(1.0 +

show

1.8 1.3 0.8 0.8 1.2 1.3 1.8 1.1

1.5 OR 95%

DRC and

caregiver

therefore

(40.3) (45.3) (36.9) (44.6) (34.5) (40.0) (40.8)

(44.7) (38.5) (32.4) (41.4) (42.1) (40.0) (48.1) (37.4)

(OR) %

and

N might,

ratio’s 47/105 63/164 53/164 66/160 87/206 48/115 74/153 65/174

172/379 143/386 140/314 111/321 216/540 131/320 282/700 child

odds

CI) key

2.4) 1.8) 5.9) 1.4) 4.0) 3.3) calculation 7.9)* ‐ ‐ ‐ ‐ ‐ ‐

‐ (%), ‐ 15.1)**

‐ re ref ref ref ref ref (95%

(0.8 (0.3 (0.9 (0.5 (0.6 (0.5

(1.2 +

(1.5

1.4 0.8 2.4 0.8 1.5 1.3 among Faso (57.1) 3.0

OR

4.7

Manually

percentages )

‐ Burkina ( (54.0) (54.2)

(61.2) (53.3) (62.0) (52.7) 149/261 (n), (36.8) (67.0)

(64.7) (65.5) (56.7) (47.2) (70.3)

. .

(%)

N analysis.

15/22 30/46 46/71 37/65 23/48 46/66 13/35 25/37 Associations 83/135 78/146 71/114 66/126 103/190 109/202 numbers these

quintile:

+

for

years

index

3.3

child:

<5 years

education

STATA years

of 5

‐ calculations:

<2 Secondary 2 Urban Richer Female Rural Richest No Male Middle Poorer Primary Poorest Country Variable

Residence:

Education:

Sex

Age:

Total

Wealth used Table All

43 Chapter 3

DISCUSSION

We found considerable variation in care seeking behaviour for suspected pneumonia across six high pneumonia mortality countries in sub‐Saharan Africa. The three countries found to have the lowest levels of care seeking from ‘appropriate’ providers (i.e. health facility/ provider) were Ethiopia, Nigeria and DRC.

Overall care seeking Country variations in care seeking for childhood infections, including pneumonia, have been reported previously.16‐17 We concur with the conclusions of the Hodgins study16 that programs should be adjusted to country specific needs based on identified barriers. Our data clearly show that both care seeking from ‘appropriate’ providers and childhood pneumonia mortality is high in Uganda, suggesting that the quality of care available to these children may be sub‐standard or that children present for care late, as has been reported in other studies.18‐20 In Ethiopia, where we found care seeking to be the lowest among the six countries, strategies that investigate what the key challenges are related to accessing care need to be prioritized. Previous studies conducted in Ethiopia indicate that lack of knowledge and delay in recognition of the severity of an illness are important factors that predict care seeking,21‐22 as are religion23 and household wealth.24 Further, vast distances to facilities within the country were the impetus for the health extension worker program, which has made important strides in increasing coverage of treatment for childhood illnesses, although some challenges remain.25 Our analysis confirms that there is a strong association between wealth and care seeking in Ethiopia, which we also found in Tanzania, Nigeria and Burkina Faso. Other studies report that the associations between wealth and care seeking are not only linked to whether care is sought or not, but also from which facility. Studies from Nigeria and Tanzania reported that poorer women are more likely to utilize facilities which provide poor quality services.26‐27 In Tanzania, women living in rural areas tend to visit primary health care (PHC) facilities more often, whereas richer and higher educated women visit hospitals or better equipped health facilities. The main reasons for bypassing PHC facilities are related to the lack of diagnostic aids and drugs.28 In each setting, once solutions to locally specific barriers in care seeking are identified, it will be critical to ensure that demand generation efforts are not jeopardized by sub‐ standard service and treatment availability.29‐30 An earlier study that reported higher treatment rates in countries with well‐established private sector services suggested that the appropriateness of treatments provided may be a challenge.17 Our study also found high use of ‘non‐appropriate’ providers (e.g. private pharmacies, traditional practitioners and other services, such as shops, churches, relatives, etc.) in

44 Care seeking behaviour for children with suspected pneumonia both DRC and Nigeria, 43 and 38% of the total care seeking respectively (see Table 3.2). Another recent study from DRC indicates that while ‘formal care’ (in this paper referred to as health providers or ‘appropriate’ providers) is valued, the cost of services creates barriers and results in families either self‐medicating or using traditional provider options.31 Another study from DRC also concluded that costs were a barrier, but suggested that distrust of government health services is also a problem.32 Similarly, studies of care seeking for childhood illnesses in Nigeria have reported high use of home care, drug vendors and private clinics due to financial constraints, wishing to try home management first, and poor recognition of the severity of the illness or waiting for the child to improve.33‐35 As Quinley & Govindasamy have reported, drug shops do often function as de facto clinics.36 Strategies for improving the quality of care in these service delivery points may be an important program strategy to consider.16 The only association between sex of the child and care seeking we found was in Uganda, where girls were actually more likely to be brought for care than boys. Earlier published findings from Asia indicate preferential care seeking for male children,37‐38 and previous African studies of the association between the sex of a child and care seeking in Tanzania, where no association was found.39,40 In addition, while studies have reported associations between better health outcomes and higher levels of caregiver education,34‐41 for associations between caregiver education and care seeking ‐ hence increasing the likelihood of receiving correct treatment ‐ we only observed an association in Uganda. We found differences between care seeking in rural versus urban areas in Burkina Faso, Ethiopia, Nigeria and Tanzania, as has been reported in previous studies.21,42 For example, in Ethiopia the percentage seeking care from an ‘appropriate’ provider in rural areas is 25, in contrast to 47% in urban areas. However, when we controlled for wealth, education, sex and the age of a child, there was no independent association of urban‐rural residence. This may be (partly) explained by lower education and wealth status of those living in rural areas. In relation to this, our analysis shows that lower wealth quintiles were associated with lower levels of care seeking from ‘appropriate’ providers in Burkina, Ethiopia, Tanzania and Nigeria, independent of the actual residency of caregivers. In other words, the poorest are the least likely to seek care, independent of where they live (i.e. in urban or rural settings). This implies that those living in rural areas are more disadvantaged due to poverty and access to health services. These associations were not found in DRC and Uganda. Our study indicates that care seeking for children under the age of 2 with suspected pneumonia was more frequent than for children between 2‐5 years of age, particularly in Nigeria, Ethiopia and DRC, this relationship was significant. Although similar findings were reported in previous studies from Nairobi43 and Nigeria,44 studies from Uganda19 and Ethiopia24 reported no association between care seeking and age of the child. Seeking treatment for younger children with suspected pneumonia is especially critical

45 Chapter 3 to decrease childhood mortality due to pneumonia, not only because the incidence of pneumonia is highest amongst these children, but also because 81% of child deaths due to pneumonia occur within this age group.45 Demand generation efforts should take this into account and should focus on preventive measures as well, e.g. increasing coverage of immunization, and improving breastfeeding practices and nutritional status.46‐47 One of the strategies to improve the quality of care is through Integrated Management of Childhood Illnesses (IMCI) – a strategy designed to reduce child mortality and morbidity due to common illnesses. IMCI has been implemented in all 6 countries included in these analyses. The strategy aims to improve family and community health practices, case management of health staff and the overall health system,48 although the extent to which this is achieved in any particular setting will vary. The level of implementation, quality of training and supervision will have an impact on care seeking behaviour and the quality of care.49,50 Moreover, the quality of care is not merely affected by the capacity of health workers, but also on the availability of essential resources, inkling appropriate medicines.49 The IMCI protocol guides health workers to classify a child as having pneumonia when s/he presents with a cough and fast and/ or difficult breathing due to a problem in the chest. Despite the protocol, health workers are often challenged to classify and prescribe treatment for these suspected pneumonia cases.19,51

Limitations There are limitations related to these analyses. Pneumonia prevalence, which is collected in household surveys primarily for use as a denominator for indicators relating to pneumonia, should be interpreted with caution as it depends on caregivers’ perception of the signs and symptoms (which may or may not be accurate) and their capacity to recall the events (which may be prone to recall bias), leading to incorrect estimates.52 Moreover, the prevalence of suspected pneumonia varies seasonally, which also influences care seeking (i.e., it may be more difficult to take a child for care during harvest and rainy seasons). In relation to this, caregivers may identify signs and symptoms such as cough and difficulty or rapid breathing due to a problem in the chest, while, clinically these may refer to another acute respiratory tract (ARI) infection, rather than to pneumonia specifically. Secondly, survey data do not allow us to determine the specific pathways of care taken (i.e. which care provider – either ‘appropriate’ or ‘inappropriate’ ‐ was visited first), if the same provider was visited multiple times or if the same health worker assessed a sick child in a health centre and again later in his or her capacity as a private pharmacist. Having this additional information would allow for more nuanced analyses of care seeking behaviours in these settings.

46 Care seeking behaviour for children with suspected pneumonia

Finally, survey data also do not provide information on severity of illness, and it is therefore not possible to distinguish whether or not seeking care from PHC facility was appropriate, or if the child should have gone directly to hospital. In relation to this, it would be interesting to assess if a child received treatment, however, as Hazir et al (2013) concluded ‘…data in its current format from DHS/MICS surveys should not be used for the purpose of monitoring antibiotic treatment rates in children with pneumonia at the present time as their quality is jeopardized’.52 This is because the identified cases depend on the caregiver’s interpretation of signs and symptoms; the validity of receiving antibiotics is therefore dependent on the accuracy of this interpretation.

Further research areas In this analysis we included (when applicable) multiple providers/facilities per child, which could include both ‘appropriate’ and ‘inappropriate’ provider categories. Further research could assess the percentage of care sought from more than one type of provider and why. A better understanding of the complexity of care seeking and associated delays including the timing of care‐seeking could reveal additional information about quality of care and user preferences. Further research should aim to understand the correlation between mortality and care seeking, and assess if there are differences in this association between any care seeking and care seeking from ‘appropriate’ providers. These analyses should focus not merely on pneumonia, but also on the other main causes of illnesses (e.g. diarrhoea and fever).53‐54 Finally, we need to know why overall care seeking is unacceptably low in some countries, even for the youngest children (who benefitted from slightly higher levels of care seeking in this study but who are also the most likely to die from pneumonia). It is critical to identify strategies to improve the quality of services visited most often by caregivers, including the ‘informal’ sector, e.g. private pharmacies.

Conclusion In conclusion, this study illustrates that prior to planning strategies to decrease pneumonia mortality, it is crucial to understand care seeking patterns (and the related determinants) between countries and within countries using available data, such as national survey data. Further research is needed to better understand the reason behind these findings by conducting more systematic analyses at national and sub‐ national level, including the assessment of socioeconomic, knowledge and information, cultural and health system factors that influence care seeking. Locally specific research is also needed to understand why families choose the providers and facilities they choose, and then programmatic strategies should be developed that engage local community members to identify relevant, feasible and acceptable solutions.

47 Chapter 3

Acknowledgement Special thanks to Paddy Hinssen for cross‐checking our analysis.

48 Care seeking behaviour for children with suspected pneumonia

REFERENCES

1. Simoes EAF, Cherian T, Chow J, Shahid‐Salles SA, Laxminarayan R, et al. Acute Respiratory Infections in Children. In: Jamison DT, Breman JG, Measham AR, et al., editors. Disease Control Priorities in Developing Countries. 2nd edition. Washington (DC): World Bank; 2006. Chapter 25. Available:http://www.ncbi.nlm.nih.gov/books/NBK11786/. Accessed 2014 Sep 22. 2. United Nations Children’s Fund (UNICEF) (2014) Levels & trends in child mortality. Report 2014. New York: UNICEF. 3. Liu L, Oza S, Hogan D, Perin J, Rudan I, et al. Global, regional, and national causes of child mortality in 2000–13, with projections to inform post‐2015 priorities: an updated systematic analysis. Lancet. 2015; 385:430‐440. 4. UNICEF (2014) Committing to child survival: A Promise Renewed. Progress report 2014. New York: UNICEF. 5. Pio A. Standard case management of pneumonia in children in developing countries: the cornerstone of the acute respiratory infection programme. Bull World Health Organ 2003;81:298‐300. 6. UNICEF and World Health Organization (WHO) (2006) Pneumonia the forgotten killer of children. Available: http://www.childinfo.org/files/Pneumonia_The_Forgotten_Killer_of_Children.pdf. Accessed 2014 Oct 6. 7. WHO and UNICEF (2012) Pneumonia and diarrhea: Tackling the deadliest diseases for the world’s poorest children. Geneva: WHO. 8. Colvin CJ, Smith HJ, Swartz A, Ahs JW, de Heer J, et al. Understanding careseeking for child illness in sub‐ Saharan Africa: a systematic review and conceptual framework based on qualitative research of household recognition and response to child diarrhoea, pneumonia and malaria. Soc Sci Med 2013;86: 66‐78. 9. Ellis AA, Doumbia S, Traoré S, Dalglish SL, Winch PJ. Household roles and care‐seeking behaviours in response to severe childhood illness in Mali. J Biosoc Sci 2013;45:743–759. 10. Ferdous F, Das SK, Ahmed S, Farzana FD, Kaur G, et al. The impact of socio‐economic conditions and clinical characteristics on improving childhood care seeking behaviors for families living far from the health facility. Sci J Public Health 2013;1:69–76. 11. Scott K, McMahon S, Yumkella F, Diaz T, George A. Navigating multiple options and social relationships in plural health systems: a qualitative study exploring healthcare seeking for sick children in Sierra Leone. Health Policy Plan 2014;29:292‐301. 12. UNICEF and WHO (2013) Ending preventable child deaths from pneumonia and diarrhoea by 2025. The integrated Global Action Plan for Pneumonia and Diarrhoea (GAPPD). Geneva: WHO. 13. Hancioglu A, Arnold F. Measuring coverage in MNCH: tracking progress in health for women and children using DHS and MICS household surveys. PLoS Med 2013;10:e1001391. 14. Rutstein SO, Johnson K. (2004) The DHS Wealth Index. DHS Comparatie Reports No. 6. Calverton, Maryland USA: ORC Macro. Available: http://dhsprogram.com/pubs/pdf/CR6/CR6.pdf. Accessed 2014 Nov 24. 15. UNICEF (2013) Committing to child survival: A Promise Renewed. Progress report 2013. New York: UNICEF. 16. Hodgins S, Pullum T, Dougherty L. Understanding where parents take their sick children and why it matters: a multi‐country analysis. Glob Health Sci Pract 2013;1:328–356. 17. Mosites EM, Matheson AI, Kern E, Manhart LE, Morris SS, et al. Care‐seeking and appropriate treatment for childhood acute respiratory illness: an analysis of Demographic and Health Survey and Multiple Indicators Cluster Survey datasets for high mortality countries. BMC Public Health 2014;14:446 18. Källander K, Nsungwa‐Sabiiti J, Balyeku A, Pariyo G, Tomson G, et al. Home and community management of acute respiratory infections in children in eight Ugandan districts. Ann Trop Paediatr. 2005;25:283–291. 19. Källander K, Hildenwall H, Waiswa P, Galiwango E, Peterson S, et al. Delayed care seeking for fatal pneumonia in children age under five years in Uganda: a case‐series study. Bull World Health Organ 2008;86:332–338.

49 Chapter 3

20. Hildenwall H, Tomson G, Kaija J, Pariyo G, Peterson S. “I never had the money for blood testing”— Caretakers’ experiences of care‐seeking for fatal childhood fevers in rural Uganda—a mixed methods study. BMC Int Health Hum Rights 2008;8:12. 21. Tsion A, Tefera B, Ayalew T, Amare D. Mothers’ health care seeking behaviour for childhood illnesses in Derra district, Northshoa zone, Oromia Regional Stata, Ethiopia. Ethiopian J Health Sci 2008;18(3): 87– 94. 22. Awoke W. Prevalence of childhood illness and mothers’/caregivers’ care seeking behavior in Bahir Dar, Ethiopia: A descriptive community based cross sectional study. Open Journal of Preventive Medicine 2013;3:155–159. 23. Mebratie AD, Van de Poel E, Yilma Z, Abebaw D, Alemu G, et al. Healthcare‐seeking behaviour in rural Ethiopia: evidence from clinical vignettes. BMJ Open 2014;4:e004020. 24. Deressa W, Ali A, Berhane Y. Household and socioeconomic factors associated with childhood febrile illnesses and treatment seeking behaviour in an area of epidemic malaria in rural Ethiopia. Transactions of the Royal Society of Tropical Medicine and Hygiene 2007;101:939–947. 25. Miller NP, Amouzou A, Tafesse M, Hazel E, Legesse H, et al. Integrated Community Case Management of Childhood Illness in Ethiopia: Implementation Strength and Quality of Care. Am J Trop Med Hygiene 2014;91:424–434. 26. Kahabuka C, Kvåle G, Hinderaker SG. Care‐Seeking and Management of Common Childhood Illnesses in Tanzania—Results from the 2010 Demographic and Health Survey. PLoS One 2013;8: e58789. 27. Onwujekwe O, Hanson K, Uzochukwu B. Do poor people use poor quality providers? Evidence from the treatment of presumptive malaria in Nigeria. Trop Med Int Health 2011;16:1087–1098. 28. Kahabuka C, Kvåle G, Moland KM, Hinderaker SG. Why caretakers bypass Primary Health Care Facilities for child care—a case from rural Tanzania. BMC Health Services Research 2011;11:315. 29. Bhutta ZA, Darmstadt GL, Haws RA, Yakoob MY, Lawn JE. Delivering interventions to reduce the global burden of stillbirths: improving service supply and community demand. BMC Pregnancy and Childbirth 2009;9(Suppl 1):S7. 30. Ridde V, Morestin F. A scoping review of the literature on the abolition of user fees in health care services in Africa. Health Policy Plan 2010;26:1–11. 31. Chenge MF, Van der Vennet J, Luboya NO, Vanlerberghe V, Mapatano MA, et al. Health‐seeking behaviour in the city of Lubumbashi, Democratic Republic of the Congo: results from a cross‐sectional household survey. BMC Health Serv Res 2014;14:173. 32. Maketa V, Vuna M, Baloji S, Lubanza S, Hendrickx D, et al. Perceptions of health, health care and community‐oriented health interventions in poor urban communities of Kinshasa, Democratic Republic of Congo PLoS One 2013;8:e84314. 33. Okeke TA, Okeibunor JC. Rural–urban differences in health‐seeking for the treatment of childhood malaria in south‐east Nigeria. Health Policy 2010;95:62–68. 34. Ukwaja KN, Talabi AA, Aina OB. Pre‐hospital care seeking behaviour for childhood acute respiratory infections in south‐western Nigeria. Int Health 2012;4:289–294. 35. Tinuade O, Iyabo R‐A, Durotoye O. Health‐care‐seeking behaviour for childhood illnesses in a resource‐ poor setting. J Paediatr Child Health 2010;46:238–242. 36. Quinley J, Govindasamy P. The Treatment of Childhood Illness in Nepal: Further Analysis of the 2006 Nepal Demographic and Health Survey. Calverton, Maryland, USA: Macro International 2007. 37. Pandey A, Sengupta PG, Mondal SK, Gupta DN, Manna B, et al. Gender differences in healthcare‐seeking during common illness in a rural community of West Bengal, India. Journal of Health Population Nutrition 2002;20:306–311. 38. Shaikh BT, Hatcher J. Health seeking behaviour and health service utilization in Pakistan: challenging the policy makers. J Public Health 2005;27:49–54. 39. de Savigny D, Mayombana C, Mwageni E, Masanja H, Minhaj A, et al. Care‐seeking patterns for fatal malaria in Tanzania. Malaria Journal 2004;3:27. 40. Schellenberg JA, Victora CG, Mushi A, de Savigny D, Schellenberg D, et al. Inequities among the very poor: health care for children in rural southern Tanzania. Lancet 2003;361:561–566. 41. Bhutta ZA, Chopra M, Axelson H, Berman P, Boerma T. Countdown to 2015 decade report (2000–10): taking stock of maternal, newborn, and child survival. Lancet 2010;375:2032–2044.

50 Care seeking behaviour for children with suspected pneumonia

42. Holtz TH, Kachur SP, Marum LH, Mkandala C, Chizani N, et al. Care seeking behaviour and treatment of febrile illness in children aged less than five years: a household survey in Blantyre District, Malawi. Transactions of the Royal Society of Tropical Medicine and Hygiene 2003;97:491–7. 43. Taffa N, Chepngeno G. Determinants of health care seeking for childhood illnesses in Nairobi slums. Trans R Soc Trop Med Hyg 2003;97:491‐497. 44. Ukwaja K, Olufemi O. Home management of acute respiratory infections in a Nigeria district. African J Respir Med 2010;6:18–22. 45. Fischer Walker CL, Rudan I, Liu L, Nair H, Theodoratou E, et al. Global burden of childhood pneumonia and diarrhea. Lancet 2013;381:1405–1416. 46. Campbell H, el Arifeen S, Hazir T, O’Kelly J, Bryce J, et al. Measuring coverage in MNCH: Challenges in monitoring the proportion of young children with pneumonia who receive antibiotic treatment. PLoS Medicine 2013;10:e1001421. 47. Theodoratou E, Johnson S, Jhass A, Madhi SA, Clark A, et al. The effect of Haemophilus influenzae type b and pneumococcal conjugate vaccines on childhood pneumonia incidence, severe morbidity and mortality. International Journal of Epidemiology 2010;39 (Suppl 1) i172–i185. 48. Ketsela T, Habimana P, Martines J, Mbewe A, Williams A, et al. Opportunities for African’s Newborns, Chapter 5 Integrated Management of Childhood Illness (IMCI). Available: http://www.who.int/pmnch/media/publications/africanewborns/en/index1.html. Accessed 2014 Sep 22. 49. Pariyo GW, Gouws E, Bryce J, Burnham G, Uganda IMCI impact study team. Improving facility based care for sick children in Uganda: training is not enough. The London School of Hygiene and Tropical Medicine 2005;20 (Suppl 1):i58‐i68. 50. Armstrong Schellenberg JRM, Adam T, Mshinda H, Masanja H, Kabadi G, et al. Effectiveness and cost of facility‐based Integrated Management of Childhood Illnesses (IMCI) in Tanzania. Lancet 2004;364:1583– 1594. 51. Noordam AC, Barberá Laínez Y, Sadruddin S, van Heck PM, Chono AO, et al. The use of counting beads to improve the classification of fast breathing in low‐resource settings: a multi‐country review. Health Policy Plan. 2015;30:696‐704. 52. Hazir T, Begum K, el Arifeen S, Khan AM, Huque MH, et al. Measuring coverage in MNCH: A prospective validation study in pakistan and Bangladesh on measuring correct treatment of childhood pneumonia. PLoS Med. 2013;10:e1001422. 53. Sreeramareddy CT, Sathyanarayana TN, Kumar HNH. Utilization of health care services for childhood morbidity and associated factors in India: A national cross‐sectional household survey. PLoS One 2012;7:e51904. 54. Aremu O, Lawoko S, Moradi T, Dalal K. Socio‐economic determinants in selecting childhood diarrhea treatment options in Sub‐Saharan Africa: A multilevel model. Ital J Pediatr 2011;37:13.

51 Chapter 3

52 CHAPTER 4

The use of counting beads to improve the classification of fast breathing in low resource settings: A multi‐country review

Noordam AC, Barberá Laínez Y, Sadruddin S, van Heck PM, Chono AO, Acaye GL, Lara V, Nanyonjo A, Ocan C, Källander K Health Policy and Planning 2015;30:696–704

53 Chapter 4

ABSTRACT

To decrease child mortality due to common but life threatening illnesses, community health workers (CHWs) are trained to assess, classify and treat sick children. For pneumonia, CHWs are trained to count the respiratory rate of a child with cough and/ or difficulty breathing, and determine whether the child has fast breathing or not based on how the child’s breath count relates to age‐specific respiratory rate cut‐off points. International organisations training CHWs to classify fast breathing realized that many of them faced challenges counting and determining how the respiratory rate relates to age specific cut‐off points. Counting beads were designed to overcome these challenges. This paper presents findings from different studies on the utility of these beads, in conjunction with a timer, as a tool to improve classification of fast breathing. Studies conducted by the International Rescue Committee and Save the Children among illiterate CHWs assessed the effectiveness of counting beads to improve both counting and classifying respiratory rate against age specific cut‐off points. These studies found that the use of counting beads enabled and improved the assessment and classification of fast breathing. However, a Malaria Consortium study found that the use of counting beads decreased the accuracy of counting breaths among literate CHWs. Qualitative findings from these studies and two additional studies by UNICEF suggest that design of the beads is crucial: beads should move comfortably and a separate bead string, with color‐coding, is required for the age‐groups with different cut‐off thresholds – eliminating more complicated calculations. Further research, using standardised protocols and gold standard comparisons, is needed to understand the accuracy of beads in comparison to other tools used for classifying pneumonia, which CHWs benefit most from each different tool (i.e. disaggregating data by levels of literacy and numeracy) and what the impact is on improving appropriate treatment for pneumonia.

54 The use of counting beads to improve the classification of fast breathing

INTRODUCTION

Because of the overall complexity of diagnosis, the still staggering mortality, lack of diagnostic aids and the growing problem of antibiotic resistance for pneumonia, there is an urgent need for more robust data on tools for pneumonia diagnosis. Pneumonia is the leading cause of childhood mortality, accounting for 17% of global deaths of children under five.1 Timely recognition of pneumonia signs and symptoms, appropriate care seeking and access to antibiotic treatment can prevent many of these deaths. The Integrated Management for Childhood Illness (IMCI) protocol for ‘Caring for Newborns and Children in the Community’ guides community health workers (CHWs) to assess fast breathing as an indicator of non‐severe pneumonia in children with cough and/ or difficulty in breathing.2 There are two steps in detecting if a child has fast breathing or not: first, a CHW needs to visually count a child’s breath for one minute, second, the CHW has to determine how the child’s breath count relates to age specific respiratory cut off points.2 International organisations training CHWs with various degrees of literacy and numeracy realized that many of them faced challenges in counting and relating the breath count to age‐specific cut‐off points. However, accurate assessment of fast breathing is crucial as selected children (children 2‐11 months of age with a respiratory rate (RR) of 50 or more breaths per minute and children 12‐59 months of age with a RR of 40 or more) are classified as having pneumonia based on their breathing rate and require immediate treatment with antibiotics.3‐4 Watches and timers have been used as timing aids to facilitate one minute respiratory rate counting. Box 4.1 shows an example of an acute respiratory infection (ARI) timer that is distributed by UNICEF. Until now, there is limited evidence on counting devices and other affordable tools to help CHWs in resource poor settings improve classification of fast breathing. One study evaluating the effectiveness of an abacus with a built‐in sandglass concluded that CHWs were better able to correctly classify fast breathing with the breath counter.5 The potential of the use of counting devices, such as beads, is unknown due to lack of information regarding their effectiveness and utility. However, several organizations (including International Rescue Committee (IRC), Save the Children, Malaria Consortium, UNICEF and Population Services International (PSI)) have conducted small scale studies among CHWs with differing levels of literacy and numeracy, using counting beads, which contribute to the knowledge base. In this review we compile the current evidence‐base of the effectiveness of counting beads to assess and classify breathing rates to guide pneumonia diagnosis. These findings are essential to further guide integrated Community Case Management (iCCM) programing aiming to decrease pneumonia deaths in young children.

55 Chapter 4

The ARI timer makes a ticking sound every second and has an alarm after 30 seconds as well as a final alarm to inform the user that 1 minute has passed. The user must press the button to start the 1 minute timing, during which a child’s breath is counted.

Box 4.1 The ARI timer.

METHODS

The findings presented are based on studies conducted by IRC, Save the Children, Malaria Consortium and UNICEF to improve iCCM programming in South Sudan, Uganda and Ghana. All CHWs in these studies were trained in iCCM using the WHO/UNICEF IMCI ‘Caring for newborns and children in the community’ protocol. CHWs are named differently in the various countries, for example in South Sudan they are called community based distributors; however, in this paper, we refer to them as CHWs.

We compiled all research findings on the use of counting beads within these programs. While most findings presented are part of larger research initiatives, in this review we focused on the following questions: 1. What is known regarding primarily illiterate CHWs’ ability to assess and classify fast breathing without the use of counting beads? 2. Does the use of beads improve the ability of CHWs, particularly those with limited or no literacy and numeracy, to correctly classify fast breathing (hence, including tracking the breaths and classifying the breath count based on IMCI age‐specific respiratory rate cut‐off points)? 3. Does the use of beads improve the ability of literate CHWs to correctly assess the breath count? 4. What are CHWs’ perceptions of these tools, and does this differ by literacy level?

A brief description of the studies is provided below and a summary of the key methodological elements of the different studies can be found in Table 4.1. All studies used the ARI timer explained in Box 4.1 and the design of beads used by the various organizations can be found in Box 4.2. The results are separated by qualitative and quantitative research methods, as well as by organization. In addition, anecdotal evidence from two rapid assessments, from PSI in Democratic Republic of Congo (DRC) and from IRC in Sierra Leone, is not included in the methods or findings, but is used to strengthen the discussion.

56 The use of counting beads to improve the classification of fast breathing

+ and

and

district 2009 2012 2012

users

‐ ‐ ‐ and

coded and technique specific

sampling

Gonja

2007 2010 illiterate 2010

age CHWs

‐ color (NA)

45

Central Tolon/Kumbungu During not During Non 2012 ± snowball mainly During Purpose Ghana potential improvements Perceptions, experience interviews

study

this +

depth ideas

Mpigi ‐

for in

color aids

+ and and

prior prior

technique

sampling

research

groups,

of specific

applicable

months illiterate months diagnostic CHWs

3 3

Kyankwanzi district Age > coded 2011 0% > 87 Part Purpose snowball Focus observations UNICEF Uganda Not Perceptions for

CHWs

vs

33

(2

coded and

of

beads breaths breaths

opinions

county

2 2 ‐ & CHWs

color

± ± user

assessing

discussions +

94 sub and scenarios

/ included included

breathing

timer

Consortium

after standard standard application research

within within vs case group

2010 2010

fast

of specific

without) cases

illiterate communication CHWs CHWs CHWs

gold gold

Bukomero, Kibonga Age Since 2012 mobile Timer 0% Since Part All Count Focus Malaria Uganda 94 282 Video All of with one Count of Perceptions, experience on results stopped

gold

team coded the

US as

hospital children the

same County,

a

color

CHWs area

in

+

in

sick

27 beads

patient of attending

Children

‐ /

North

beads

Equatoria

research research research

finger Sudan & illiterate

out

the

of of of specific

‐ CHWs

cases

an

counting studies.

Kapoeta Eastern Part Age 2013 Timer 100% Part Part All Save South NA NA 69 training standard CHWs classification Selection at department

of

±

of

not coded were

knew

within different beads breaths +

counties potential

‐ was

standard CHWs CHWs sick

3 home &

advantage color

±

was

the + sub at 33 33

and

children point / / gold 2

district

CHW

prior prior of

timer sampling sampling

of

visit off

standard

the research ‐

finger within the (IRC) vs evident

cases cases

illiterate of specific

necessarily cut ‐ 1 1

gold Moroto the

mock

months months beads

in 33* Karamoja: 2011 33* Timer 66% 6 Age Interviews Uganda CHWs 6 Part Random Random 3 Perceptions A Count improvements which aware, of not children the elements

and

1 Committee 3 of

not East coded were

±

knew

beads breaths + potential

was

sick

3 home &

color Aweil Ghazal: standard Rescue

constraints

CHWs CHWs

±

within

+

at

and

in el children

point

32

32 CHW prior prior

gold timer / / sampling sampling time

visit off standard methodological

the research ‐

within finger of the Bahr Sudan vs illiterate

to

of specific county

necessarily cut ‐ ‐

cases cases key gold

mock months months due

Age Count 32 Interviews 2011 North sub 100% Part Timer of CHWs beads 6 32 Random A Random Perceptions International South 6 the which not improvements children aware,

the

but

of of

timer timer

focus

timer:

the place

iCCM: selected, level:

using using size: method: size method

of:

ARI beads:

beads: and

of of Summary research

location beads:

initially

Literacy Training Use Use Type Sample Sample Assessing impact Type assessment: Correct alone: Correct & Sample Sample Key Method

Characteristics component

4.1

component research

Quantitative

research were

research Qualitative Background 46

1 Country Geographic Year Table *

57 Chapter 4

Picture Description of the device Used by:

1 set of 2 ‘age‐specific and color‐coded’ strands of beads, Save the with 1 bead counted per breath. The two strands can be Children distinguished because the beads of 1 age group have different colours and sizes than the beads of the other strand. Several tiny beads are used to create space between the beads and the strand is tightly tied to hold the beads in place. At the end of age cut‐offs for each strand, there are 10 red/ pink beads which, if counted, indicate fast breathing. Strands have a clasp so they can be used open (straight) or closed (like a necklace). Beads are eclipse shaped and made from newspapers and glue. 1 strand of beads, non‐specific for children ages 0 to 5 years) Ministry of and color‐coded per 10 beads to ease counting. The strand is Health in necklace shaped and has a protruding start/ end bead. 1 bead Ghana and is counted per breath. Beads are made from plastic. CHWs UNICEF are using these beads in conjunction with the ARI Timer.

1 set of 2 ‘age‐specific and color‐coded’ strains of beads, with IRC and 1 bead counted per breath. The two strands can be Malaria distinguished because the 10 fast breathing beads of one age Consortium group are red and other green (matching the amoxicillin packaging in Uganda). There are no separator beads and they

are tied so that there is space for moving the beads across the string. The beads are made from plastic and round shaped. Strands have a clasp so they can be used open (straight) or closed (like a necklace). The white beads are 16mm in diameter, whereas the coloured beads are 12mm in diameter. 1 set of 2 ‘age‐specific and color‐coded’ strains of beads, with PSI 1 bead per breath. The two strands can be distinguished because the fast breathing beads of one age group are pink and other green. There are no separator beads. The beads are made from plastic and round shaped. The strand is straight, with a small orange bead at the beginning and end.

Box 4.2 Overview of devices used by the various organizations.

Quantitative research

International Rescue Committee study in Uganda and South Sudan As part of a larger assessment evaluating the quality of care provided by CHWs, IRC compared the CHWs’ ability to assess fast breathing using the ARI timer alone versus using the timer and beads. The geographical areas (payams) were selected on the basis of their geographical accessibility to the evaluation team and recent start‐up of the program.

58 The use of counting beads to improve the classification of fast breathing

First, CHWs were asked to describe the cut‐off points for fast breathing for the two age groups (2‐11 and 12‐59 months) and to count the child’s respiratory rate (RR) using the ARI timer. A trained clinician who was part of the evaluation team counted simultaneously with the CHW. The RR from the clinician was written down and the CHW was asked to say his/her count. Afterwards, CHWs were given the right counting bead string for the age of the child being assessed and were shown how to move their fingers along the beads and to stop when the timer beeped. Then CHWs were asked to repeat the assessment with the same child using the timer and the beads. The clinician used the counting beads simultaneously with, but out of sight of, the CHW. The beads were counted back and recorded for both the CHW and the trained clinician. Microsoft Excel was used to analyse the data.

Save the Children study in South Sudan The use of beads by CHWs with limited numeracy was part of an operational research lead by Save the Children. The research was conducted to assess the effectiveness of simulation based training of CHWs using video technology.

During the training, CHWs were shown video clips of cases with danger signs (including case scenarios of malaria, pneumonia and diarrhoea) and the interactions between CHWs and caretakers. The video also showed the assessment, classification, treatment and advice on home care. CHWs were given 1 set (two strings) of beads to count and classify the breath count along with the ARI timer acting as a stop watch. CHWs were taken to a hospital outpatient department for post training skills assessment, where they assessed sick children (40 out of the 69 cases had a history of cough). Each CHW managed (assessing, classifying and deciding to treat) 2‐3 sick children aged 2‐59 months. Each CHW also completed a knowledge questionnaire. As these particular CHWs were illiterate, one clinician trained in IMCI observed the CHWs’ management of sick children and recorded the assessment, classification and treatment findings on a study form. A senior evaluator who was also an IMCI trainer independently assessed the children at the same time and recorded his findings on a similar form. The RR was measured for all 69 children by the CHWs and the evaluator. To assess fast breathing the CHW had to choose the correct bead string for the age group. The observer marked the case as having fast breathing if the CHW reached the red beads (fast breathing for 2‐11 month and 12‐59 month age groups) within the minute. The evaluator used the ARI timer to count the child breaths and noted the actual breath count in the form. The data was entered in CSPro, and imported and analysed in Microsoft Excel.

Malaria Consortium study in Uganda Malaria Consortium assessed if there was a difference in accuracy in counting RR among CHWs using 1) the ARI timer alone versus 2) using the timer & beads and 3) using a mobile phone application, where the centre button on the phone was

59 Chapter 4 pressed for every breath observed, and which beeped after one minute, after which the count was displayed. The sub‐county was selected on the basis of its geographical location and mixture of CHWs with different age, sex and literacy levels (varying from being able to read and write well in any of the local languages (78%) to fairly well (22%)).

First, each CHW received a detailed explanation of how to count the RR, while simultaneously using the ARI timer, the same timer with beads as well as the mobile phone application. The CHWs then had the opportunity to practise the three different methods of counting the RR on videos of children with different breath counts. After familiarizing themselves with the three methods, each CHW was observed counting using the three options respectively, on children with different breath counts. The video case scenarios were displayed on laptop screens, enabling one CHW to assess a child at a time. The final RR was recorded for the three tools as follows: 1) when using the timer alone, the CHW was asked for the final RR count, 2) when using both the timer and beads, the beads were counted back by a research assistant who provided the final RR count, and 3) for the mobile phone application, the result was read from the screen. The breathing rate of the children in the video was known and was used as the gold standard. STATA 12 was used to analyse the data. The research only assessed the impact on counting the RR and did not assess the impact of these devices on classifying the breath count against the age‐specific cut‐off points.

Qualitative research Along with their quantitative research efforts, IRC and Malaria Consortium assessed CHWs’ perceptions about the beads, focussing on their opinions and proposed improvements. Data were collected through interviews. Malaria Consortium also assessed CHWs’ perceptions about how the use of counting beads helped them communicate the test results to caregivers.

Other qualitative research was initiated by UNICEF, as part of a broader effort to identify CHWs’ unmet needs regarding tools to support the assessment and classification of RRs. In Uganda the focus of the research was on CHWs’ experiences assessing RR using the ARI timer and their ideas regarding tools that might be useful to help them improve their assessment and the classification of RRs. The CHWs were not familiar with the use of beads. Building on these findings, a subsequent study was conducted in Northern Ghana, where the CHWs had been trained to use both the ARI timer and beads. The main objective of this study was to help UNICEF improve the design of diagnostic aids based on the challenges CHWs indicated they face while assessing, classifying and identifying treatment needs for children under‐five with rapid breathing.

60 The use of counting beads to improve the classification of fast breathing

FINDINGS

Table 4.2 summarizes the key findings of the various studies. In addition, an overview of the two different types of counting beads (a non‐age specific type and an age‐specific type with color‐coded beads) can be found in Box 4.3.

Box 4.3 Overview of the two types of counting beads intended for use in conjunction with an ARI timer. Non‐age specific counting beads These counting beads are designed to help the CHW keep track of the amount of breaths taken. The CHW counts moves bead for each breath. When one minute has passed the CHW counts back the beads to determine the respiratory rate. Due to the colour‐coding of a number of beads (e.g. every set of 10), the CHW can count back the beads per colour: e.g., 1 colour = 10 breaths, 2 colours = 20 breaths, etc. Based on the respiratory rate the CHW compares the result against the IMCI guideline, correctly remembering the age specific cut‐off rate.

Age‐specific beads with color‐coding These counting beads support the CHW not only with counting, but also with interpreting the RR against the IMCI guidelines. These beads remove the need to count by the CHWs to assess pneumonia (unless the actual RR is required for reporting purposes) because they consist of a set of two strands or rows of beads that are color‐coded to match the thresholds for the two different age groups. Depending on the age of the child, the CHW selects the matching bead strand and moves the beads for one minute. Once the minute has passed, the CHW can identify if the child has pneumonia or not, depending on the colour of the bead s/he is holding between his/ her fingers.

Primarily illiterate CHWs’ ability to count breaths and their knowledge of age‐specific cut‐off rates The IRC assessed what the key challenges were for primarily illiterate CHWs while classifying fast breathing. Findings indicated that 46% (21/46) of CHWs in Uganda were not able to apply the age‐specific cut‐off points and 33% (15/46) made an incorrect count using the ARI timer. In South Sudan, 59% (19/32) of CHWs were not able to apply the age‐specific cut‐off points and 72% (23/32) of CHWs made an incorrect count using the ARI timer.

Primarily illiterate CHWs’ ability to classify fast breathing using counting beads In South Sudan 13% (4 out of 32) of the CHWs were able to classify fast breathing using only the ARI timer, whereas this number increased to 63% (20/32) (OR=11.7, p=0.002) when they used the beads together with the timer. Findings were similar in Uganda, where the ability to classify fast breathing increased from 37% (17/46), using only the timer, to 73% (24/33) (OR= 4.4, p<0.005), using both tools. Combined the data from both countries; the ability to classify fast breathing increased from 27% (21/78) to 68% (44/65) (OR= 5.7, p<0.005), see Figure 4.1.

61 Chapter 4

by

‐ ‐ age ‐

‐ and were

cut age

used s’ non

‘ and

color challenged

Children these

& bead

shown used beads

still the

IRC,

points

counting

Ghana and preferred off When in remembering CHWs specific coded were specific Save

(NA)

toy, used

a a

be easily

only

classifying as

caregivers revealed

with support question

have

to to

to would were who

with

it communicating

useful timer

research UNICEF Uganda NA them and results Beads perceived confused but CHWs the useful that device

this

for

made times rate

timer the slow

reduces

count the

in a

more

combined

5.6 compared

to

data the

is

is

the

using

it to Consortium with

which

to beads

were

alone likely movement

beads

beads of respiratory accuracy

specific

when

easier

Use it CHWs Malaria Uganda No communicate more the correctly and caretaker Combining results timer to tasking, bead with process the

US

they

the of

do

as

of were

breathing use assess

and

beads

count

Children correctly to

60% of

10 the

fast

arithmetic

to CHWs Sudan not

to the

breathing use able

fast could simple beyond Save South NA beads not Prior The classify CHWs enabled

and

questions able fast

strings 73% age

of (IRC)

off

‐ an not beads

to the timer

count

the training locally which of

cut

classification

research the 37%

were

made the use

4

apply more

appropriate the Committee Uganda 46% to specific using The points 33% incorrect increased correct from

for package,

classification

match to

the ‐

able identify Rescue

63% beads

age request

off related ‐ an

beads a

not

to the

timer

count amoxicillin

the of

cut

classification

CHWs

simplifies Sudan was

coded

the

13% ‐

were made use

findings apply

The Beads using International South NA practice 59% specific to points 72% increased correct from breathing Color available helped There incorrect key

of

the

limited literacy table

CHWs, literate

tools,

without

beads? by fast of of

assess beads beads CHWs numeracy, with

regarding

fast of of

these

and differ

ability ability of CHWs’ Summary

classify use use those

known counting

this illiterate assess correctly

the the

is are

the the

of

count?

to to

literacy

4.2

does use

no correctly What Does Does What

1: primarily ability the 2: improve particularly or to breathing 3: improve CHWs breath 4: perceptions and level? Table

62 The use of counting beads to improve the classification of fast breathing

Using only the ARI 73 68 timer 63

37 Using both the ARI 27 timer & 'age‐specific 13 and color‐coded' counting beads South Sudan Uganda Total

Figure 4.1 Percentage of CHWs able to classify fast breathing correctly, based on findings from IRC.

Save the Children’s research did not have a component of assessing a sick child without the use of beads, as they identified that CHWs had limited numeracy, and were not able to count beyond 10. A total of 69 sick children were assessed by the 27 CHWs. Using the ‘age‐specified and color‐coded’ beads and ARI timer, the CHWs classified 25 cases as suffering from fast breathing. The senior evaluator who only used the ARI timer classified 23 as fast breathers. Of the 25 CHW classified fast breathing cases, 15 (60%) matched with the evaluator classified fast breathing cases.

CHWs were also administered a knowledge questionnaire which included questions on use of beads appropriate for age. All 27 CHWs picked the beads appropriate for the 2 age groups. With respect to classifying a child as having fast breathing, 26 out of 27 picked the right colour (red) for the 2‐11 months group and 27 out of 27 for the 12‐59 months age group.

Literate CHWs’ ability to correctly count breaths using counting beads In Uganda, Malaria Consortium assessed if the use of tools, including counting beads, improved the ability of literate CHWs to count the respiratory rate. CHWs were 5.6 times more likely to count (not classify) respiratory rate correctly (i.e. ±2 breaths) using the timer alone compared to when it is combined with beads (OR=5.6, p<0.001). There was no significant difference between the ARI timer and mobile phone application (OR=1.1, p=0.08), implying that CHWs have a similar capacity to correctly count respiratory rate using either of the assessment methods (i.e. counting themselves versus pressing a button for every breath observed).

Overall, the median difference between the “true rate” and the rate observed with the three methods was ‐1 (IQR ‐5−2) for the UNICEF mer, ‐1 (IQR ‐7−2) for the mobile phone application and ‐5 (‐12−2) for the ARI mer with beads. Using the sign test for non‐parametric data on matched pairs, the differences in rates observed using the ARI timer compared to the true rate was not significantly different from 0 (p=0.01),

63 Chapter 4 whereas for the mobile phone application and counting beads, the difference between the rates observed and the true rate were significantly different from 0 (p=0.001 and p<0.0001, respectively). Using the same test, the median difference observed between the ARI timer and mobile phone application was not significantly different (p=0.179), whereas the difference was significantly different between the ARI timer and the counting beads (p<0.001) as well as between the mobile phone timer and the counting beads and timer (p=0.001).

When analysing the accuracy of the three different methods by the characteristic of the rate of the child in the video (i.e. normal or fast), it was demonstrated that all three methods performed much better on the slower breathing rates (i.e. 40 breaths/min) than the fast rates (i.e. 65 and 66 breaths/min). All three methods tended to overestimate the rate in the slow breathing scenario whereas they all under estimated the rate in the fast breathing scenarios.

CHWs’ perceptions on the use and design of age‐specific and color‐coded beads The color‐coded beads designed by IRC in Uganda match the colour of the locally available amoxicillin packages for specific age groups and according to (mainly illiterate) CHWs this helped them to identify the correct string needed for the child, eliminating the need to recall the cut‐off points for the different age groups. Regarding the use of counting beads, CHWs interviewed by IRC were most likely to say that: 1) the use of beads eliminated the need to count or worry about forgetting the number or making mistakes while counting; 2) it was easy to move hands along the beads; 3) it was easy to know when to give the treatment; and 4) it was easy to explain to the mother that her child did not need medicine. The need for more training and practice regarding the use of the beads in combination with the ARI timer was mentioned by CHWs in South Sudan and Uganda.

In the study amongst literate CHWs by Malaria Consortium in Uganda, CHWs expressed their support for alternative methods to count the RR other than the timer which they were familiar with. Using the beads, in addition to the timer, was perceived as being advantageous because each breath being represented by movement of a bead and then counting of the beads could be done afterwards, thus giving more accurate results.

“What I have liked about this method is that I don’t have to count the beads as I move them, counting usually comes last and this gives me more concentration on observing the child and moving the beads.”

64 The use of counting beads to improve the classification of fast breathing

The method was acknowledged as one that could give a quick picture of the diagnosis by identifying the colour of the bead where the hand has stopped after the alarm has gone off (i.e. whether white or red/green bead), the next steps could follow later:

“From what I have learnt today, the combination of both the timer and beads is very helpful because it helps me to immediately know whether the child has fast breathing by looking at the colour of beads where I have stopped and then, I count the beads to confirm what I have seen, after which I write in my book.”

The use of the beads also made it easier to communicate the results to the caretaker, as the colours visually flagged if the child had a high breath rate (indicating pneumonia) or not.

However, the beads were also perceived by some as disadvantageous because combining the timer and beads was considered to be more tasking, with a slow bead movement process and therefore reduced accuracy. During observations, it was noted that CHWs often found it difficult to start moving the beads immediately after starting the timer.

“For me, I find this method very challenging because I have to observe three things at the same time: I have to look at the child, start the timer and also move the beads at the same time, which is a bit tasking. That is why you have seen that I have been forgetting to put on the timer as I move the beads.”

Related to the design, CHWs often mentioned that the space between individual beads can affect the outcome of the count. The space between the beads should be small, as the bigger the distance was between the beads, the more difficult it was to move them. The CHWs also suggested that the beads should be light and not attractive in order not to be mistaken for accessories and have a strong string, which would not break easily.

Data from the assessment in Uganda by UNICEF, where mainly illiterate CHWs only used the ARI timer, revealed that CHWs would find it useful to have a device that supports them with the classification of fast breathing and that would help them communicate the findings to the caregivers. When the CHWs in Uganda were shown counting beads they mentioned that beads might be confused with a toy; however, it would help them with the assessment and the classification of fast breathing.

“I would have preferred the beads because each time I count, I hold a bead so I will be able to know how many beads I left behind in case I forget where I was when counting”

65 Chapter 4

“Green is OK. Red is for danger.”

CHWs perceptions on the use and design of non‐age specific beads The main concern among mainly illiterate CHWs regarding the use of non‐age specific beads used in the Northern region of Ghana, was that these beads still required counting and remembering the cut off rate. The beads also have ‘separator bead’ which the CHWs thought creates confusion since “we move the beads without looking at them”. Nevertheless, the use of prayer beads in the region is common; CHWs are used to counting with beads and are at ease with this. However, the CHWs reported that this similarity could reduce the acceptance by caretakers as the beads are perceived as a tool for prayer and not as a healthcare related tool.

When shown the ‘age‐specific and color‐coded’ beads used by IRC and Save the Children, the CHWs and their supervisors in Ghana showed a strong preference to this design as it eliminated the need to count and remember the cut‐off rate.

“This one has only two colours and the colours easily tell if the child has pneumonia or not. I think that this one will be more convenient to use than the current one we have.”

DISCUSSION

This review of studies on the utility of counting beads as a tool to improve classification of fast breathing in children with cough and/ or difficulty breathing to guide pneumonia diagnosis, shows that the introduction of ‘age‐specific and color‐coded’ counting beads, in addition to an accurate timer, can help CHWs with limited numeracy and literacy to more accurately assess fast breathing. While CHWs also expressed concerns on the task‐intensity of the method, in general it was acceptable and applicable across different settings.

There are several limitations associated with the studies included in this review. First, the studies used several different research approaches, methods (including gold standards) and research questions. Moreover, programmatic settings differed as well as the levels of literacy and numeracy of the CHWs. Second, the case scenarios by the various organizations assessed children in different settings (at home, hospital or on a video screen) which resulted in different breathings patterns, e.g. the children assessed at home were unlikely to have fast breathing, although a few of them happened to have it. This review suggests that these factors influence the utility of beads.

66 The use of counting beads to improve the classification of fast breathing

Data from IRC, Save the Children and anecdotal evidence from PSI and IRC suggest that ‘age‐specific and color‐coded’ beads enable and improve accuracy in the classification of fast breathing by CHWs with limited literacy and numeracy. However, if literacy and numeracy is not an issue, findings from Malaria Consortium show that the use of beads complicates the assessment, as it resulted in more in‐accurate counts (i.e. CHWs were 5.6 times less likely to count correctly using the beads and timer). A reason for the breathing count inaccuracy by literate CHWs was that some perceived the use of beads as more task‐intensive. While for CHWs with limited literacy and numeracy (e.g. those not able to count beyond 10), the use of beads enabled them to track the breathing rates, as well as classifying the count against the age‐specific cut‐off points, which would not be possible without these beads.

It was also found that the rate of breathing influences the accuracy in the respiratory rate count. The research conducted by Malaria Consortium shows that, regardless of the tool CHWs used, they tended to overestimate the rate in the slow breathing scenario whereas they all under estimated the rate in the fast breathing scenarios.

Recommendations Due to the overall complexity of diagnosis, the still staggering mortality, lack of diagnostic aids and the growing problem of antibiotic resistance for pneumonia, there is an urgent need for more robust data on tools for pneumonia diagnosis. Regarding the use of counting beads, these data show that it is premature to conclude to which degree the beads, in addition to a timer, would improve the ability of trained CHWs to correctly classify fast breathing. A more conclusive assessment is needed amongst sick children, disaggregating data by intensity of training and supervision, levels of literacy, numeracy and comparing the final breath count to a gold standard. Previously conducted studies on pneumonia diagnosis by CHWs indicate that even if CHWs are good in counting, they still often make mistakes in classifying the breath count.6‐7 Here, the use of beads could help in classification and it should become clearer how for literate CHWs this potential positive effect is affected by the inaccuracy in counting when using beads. Is this because of the lack of familiarity in using the beads, or is the use of beads actually more complex and task‐intensive? Concurrent to the need of more evidence regarding the utility of beads, there is a need for more research assessing the effectiveness of other devices, such as automated respiratory rate counters.

Conclusion Given the overall paucity of data, this review of recent studies provides insights on a range of issues to consider when implementing counting beads in iCCM programs. This emerging evidence suggests that the introduction of well‐designed ‘age‐specific and

67 Chapter 4 color‐coded’ beads in addition to an accurate timer can help CHWs who have difficulty counting breaths and remembering age‐specific cut‐off rates to more accurately assess and classify fast breathing. It also has potential to improve communication with the child’s caretakers – particularly regarding appropriate treatment options. However, more research is needed on these and other devices to decrease the inaccuracy in pneumonia diagnosis.

Acknowledgement Special thanks to Alyssa Sharkey for reviewing and editing this article. The authors also thank all who supported the various research components, including people from local governments, colleagues from country offices and partners from various research institutes. Authors like to thank Mark Young (UNICEF) and Shamim Ahmad Qazi (WHO) for reviewing this article.

68 The use of counting beads to improve the classification of fast breathing

REFERENCES

1. United Nations Children’s Fund (UNICEF). Committing to Child Survival: A Promise Renewed Progress Report 2013 New York: UNICEF. 2. WHO and UNICEF. 2011. Integrated Management of Childhood Illnesses Caring for Newborns and Children in the Community. Geneva: WHO. 3. World Health Organization (WHO) and UNICEF. 2005. Handbook IMCI Integrated Management of Childhood Illness. Geneva: WHO. 4. Pio A. Standard case management of pneumonia in children in developing countries: the cornerstone of the acute respiratory infection programme. Bulletin of the World Health Organization 2003;81:298‐300. 5. Bang AT and RA Bang. Breath Counter: A new device for household diagnosis of childhood pneumonia. Indian J Paediatr 1992;59:79‐84. 6. Kallander K, Tomson G, Nsabagasani X et al. Can community health workers and caretakers recognise pneumonia in children? Experiences from western Uganda. Trans R Soc Trop Med Hyg 2006;100: 956‐963. 7. Mukanga D, Babirye R, Peterson S et al. Can lay community health workers be trained to use diagnostics to distinguish and treat malaria and pneumonia in children? Lessons from rural Uganda. Trop Med Int Health 2011;16:1234‐1242.

69 Chapter 4

70 PART III

A potential solution to decrease delays

71

72 CHAPTER 5

Improvement of maternal health services through the use of mobile phones

Noordam AC, Kuepper BM, Stekelenburg J, Milen A Tropical Medicine & International Health 2011;16:622–626

73 Chapter 5

ABSTRACT

OBJECTIVE: To analyse, on the basis of the literature, the potential of mobile phones to improve maternal health services in Low and Middle Income Countries (LMIC). METHODS: Wide search for scientific and grey literature using various terms linked to: maternal health, mobile telecommunication and LMIC. Applications requiring an internet connection were excluded as this is not widely available in LMIC yet. RESULTS: Few projects exist in this field and little evidence is available as yet on the impact of mobile phones on the quality of maternal health services. Projects focus mainly on the delay in recognizing the need and making the decision to seek care, and the delay in arriving at the health facility. This is achieved by connecting lesser trained health workers to specialists, and the coordination of referrals. Ongoing projects focus on empowering women to seek health care. DISCUSSION: There is broad agreement that access to communication is one of several essential components to improve maternal health services and hence the use of mobile phones has much potential. However, there is a need for robust evidence on constraints and impacts, especially when financial and human resources will be invested. Concurrently, other ways in which mobile phones can be used to benefit maternal health services need to be further explored, taking into consideration privacy and confidentiality.

74 Improvement of maternal health services through the use of mobile phones

INTRODUCTION

Progress in achieving Millennium Development Goal (MDG) 5, to improve maternal health by reducing maternal mortality and improving access to reproductive health, is lagging behind the targets. New impulses are needed to attain the goals. Two recent international initiatives recommend mobile phones as a means to improve maternal health services.1‐2

Maternal health Every 90 seconds a woman dies of complications related to pregnancy and childbirth, resulting in more than 340 000 maternal deaths a year.3 Millions of women suffer from pregnancy‐related illnesses or experience other severe consequences such as infertility, fistula and incontinence.4 Delay is considered the key factor leading to women not accessing health services. There are three phases of delay: (i) recognizing the need for health care and in the decision‐making process; (ii) arrival at a health facility; and (iii) receiving appropriate and adequate care at the health facility.5 Underlying determinants that cause the delays are the position of women in society, large geographical distances, weak health systems, poverty and lack of education.4,6

Mobile phones MDG 8 addresses the need to make benefits of new technologies available, especially those related to information and communication. The fastest growing new technology worldwide is the mobile phone. In Africa and Asia, where the burden of maternal mortality is greatest,7 the expectations are that by 2012, 50% of the people will have access to a mobile phone.8 The uptake of mobile phones varies; it is inversely proportional to poverty rates, but also influenced by the competitiveness and thus the price levels of the relevant markets.9 The use of mobile phones in health systems is called mHealth. This article discusses the potential of mobile phones to improve maternal health services in LMIC by strengthening communication throughout different levels of the health system.

METHODS

Our literature search limited to English publications combined terms linked to: maternal health, mobile telecommunication, and LMIC. Only publications considering the basic use of mobile devices (without requiring internet access) were included, as poor internet coverage, high illiteracy rates and low levels of experience in using technology make more advanced use of mobile technology difficult in LMIC.

75 Chapter 5

Searches initiated in PubMed, Embase, Cochrane Library, Scopus, Science Direct and African Journals Online retrieved a large amount of mHealth‐related publications, of which only eight were relevant; these articles address maternal health services in LMIC, the use of mobile devices and reported preliminary results. The search was subsequently expanded to grey literature, and reference lists were also screened for further relevant sources.

LITERATURE FINDINGS

A recently published paper on mobile phone technology for health care in LMIC10 reviewed literature on mHealth, such as treatment compliance, data collection and disease prevention. The authors see great potential for mHealth; however, there is not much evidence of actual and wide‐scale impacts yet. We analysed resources for the particular area of maternal mHealth and confirmed a lack of evidence‐based studies focusing on the efficacy and effectiveness of interventions. Most documentation referred to pilot studies and often lacked baseline data, a control group and clear outcome indicators.

Accessing emergency obstetric care Before the wider use of mobile phones, several project publications considered improved communication through radio systems as one component among several aimed at improving access to emergency obstetric care and referral systems. These projects mainly focused on reducing the second phase of delay. Traditional Birth Attendants (TBAs) and ⁄ or midwives were equipped with walkie‐talkies, enabling them to contact supervisors and ambulances when facing difficult situations. Concurrently, other components such as the overall quality of the health services were improved through more reliable transport means, increased capacity, medical equipment and reduction of financial barriers. Projects in Mali, Uganda, Malawi, Sierra Leone and Ghana, which implemented the above mentioned components, noted a significant reduction in maternal deaths and an increase in supervised births when comparing the situation before and after the interventions. Faster modes of communication and transport were named as important factors in improving access to emergency obstetric care.11‐16 The projects in Uganda and Ghana additionally considered the first phase of delay by connecting traditional health providers to the biomedical health system. As TBAs are frequently at the homes of pregnant women, they can speed up the process there. Krasovec (2004) concluded that studies provided only weak empirical evidence regarding the actual impact of communication systems and that access to tools of communication is not the solution for decreasing maternal deaths in isolated areas.17

76 Improvement of maternal health services through the use of mobile phones

The tight timeframe in which a woman requires emergency obstetric care (due to e.g. severe bleeding) implies that quality services need to be accessible at short notice and supported by effective infrastructure management. In a more recent review, Lee et al. (2009) confirm the need for more rigorous assessments.18 Information regarding plans for scaling‐up projects that use radio systems was only found for the pilot project in Uganda. These plans were not realized due to high costs, inability to maintain equipment and lack of integration into the health system. However, in this project the radio system was later replaced by mobile phones, which were found to be a cheaper and a more practical solution.19

Improving the capacity of lesser trained health workers More recent projects introduced mobile phones to improve the capacity of lesser trained health workers by connecting them to better trained medical staff, thus aiming to reduce the third phase of delay. In Indonesia, Chib et al. (2008) selected 15 health facilities through random sampling; midwives in eight of the facilities received a mobile phone.20 Perceived benefits reported were that: (i) mobile phones made it easier to contact patients, midwives and supervisors, (ii) time efficiency increased due to the ability to coordinate visits, and (iii) if complications occurred assistance was only a call away. Despite these advantages, constraints included the costs, poor mobile phone network infrastructure in rural areas, increased demand for consultation, difficulties in uptake of higher technology programmes for data analysis, and hesitation in contacting supervisors due to organizational hierarchy.20‐21 A recently launched project in Rwanda went a step further by using text messaging to facilitate and coordinate the communication as well as data exchange between community health workers, health centres and hospitals. Preliminary data suggested a positive effect on access to maternal health services and consequently lower death rates.22 An initiative in Tanzania designed a phone‐based application that contained forms and protocols meant to support pregnant women before, during and after delivery.23 The results of a pilot project seemed positive; however, the authors mentioned the need to further assess the impact of the project.

Empowering women to contact health services and access information To decrease the first phase of delay, several programmes aimed to empower women to contact health services and access information; however, data was still being processed at the time this article was written. In Zanzibar, a study following 2,500 women investigated the impact of both voice and text messages on maternal health.24‐25 Text messages were sent to pregnant women containing basic health education and reminders for routine health care appointments. Expectant mothers received vouchers

77 Chapter 5 and phone numbers that they could use to contact services for questions and emergencies. The study assessed the impact on quality of services, health seeking behaviour and maternal morbidity and mortality. The data was being processed at the time of writing this article; the study promised to yield useful information.26 MoTECH is an ongoing project in Ghana aiming to determine how mobile phones can best be used to increase the quantity and quality of antenatal care.27 Results from randomized treatment and control groups were not yet available.28

Gender discrepancies in access to and use of the technology The analysis of the potential of mobile phones for maternal health requires examining how mobile phones may relate to the root cause of poor maternal health, namely the position of women in society.4 Globally, a woman is 21% less likely to own a mobile phone than a man.29 This discrepancy in the uptake of mobile phones is highest in South Asia, followed by Sub‐Saharan Africa. Women who do have access to a mobile phone often use it for business, banking and employment opportunities29‐31 and thus to make themselves more independent. Several projects use mobile phones to improve access to basic education for women, for example text message‐based literacy programmes.29 The main reason for not owning a mobile phone lies in the associated costs, illiteracy and lack of electricity.29,31 Being practical, especially women in Africa are likely to borrow a phone if they do not own one.30 Other discrepancies in the ownership of mobile phones exist between countries and in rural areas versus urban areas, mainly due to poor network coverage.32

DISCUSSION

Robust studies providing evidence on the impact of introducing mobile phones to improve the quality or increase the use of maternal health services are lacking. However, there is broad agreement that access to communication is an essential component of improving the use and quality of maternal health services. The mobile phone has a high potential as it is small, portable, widely used, relatively cheap and the extending network coverage increasingly enables communication with rural and isolated areas. The extremely quick uptake of mobile phones worldwide can shorten delays in seeking and receiving health care. The available literature suggests great potential in connecting traditional and biomedical health care, as well as connecting the different levels within a health care system, provided that women are not restricted due to their position in society, lack of finance or means of transport.

78 Improvement of maternal health services through the use of mobile phones

To fully realize the benefits of mobile communication, research needs to generate the evidence‐basis for scaling up mHealth and enabling informed mHealth policy‐making, and to analyse its benefit in ensuring timely delivery of medical equipment, provide health education and improve access to reproductive health services, e.g. for family planning. So far, projects mainly focus on acute, life threatening situations, but mobile phones can also be used to deliver mass health messages to pregnant women, recalling women with risk factors to present themselves at an antenatal clinic or referring women who suffer from complications such as fistula, incontinence and infertility. Possibilities related to connecting them to specialized hospitals need to be integrated into research and project designs. In addition, all the different applications, best practices, constraints and lessons learned need to be documented. The quick uptake of the mobile phone and its use in health care requires policies and guidance of governments, especially related to issues such as privacy and confidentiality. An overuse of text messaging by the private and public sector will soon be regarded as spam, making it lose its effectiveness. In addition to privacy, governments need to ensure confidentiality of sensitive information.

79 Chapter 5

REFERENCES

1. International Telecommunication Union (ITU) (2010) Committed to Connecting the World ITU is the UN Agency for Information and Communication Technology. Available at: http://www.itu.int. 2. mHealth Alliance (2010) Maternal and Newborn mHealth Initiative. Available at: www.mhealthalliance.org. 3. Hogan MC, Foreman KJ, Naghavi M et al. Maternal mortality for 181 countries, 1980–2008: a systematic analysis of progress towards Millennium Development Goal 5. Lancet 2010;375:1609–1623. 4. United Nations Children’s Fund (UNICEF) (2009) The state of the world’s children 2009. New York: UNICEF. 5. Thaddeus S, Maine D. Too far to walk: maternal mortality in context. Soc Sci Med 1994;38:1091–1110. 6. Ronsmans C, Graham WJ. Maternal mortality: who, when, where, and why. Lancet 2006;368: 1189– 1200. 7. World Health Organization & UNICEF (2010) Countdown to 2015 Decade Report (2000–2010): Taking Stock of Maternal, Newborn and Child Survival. Countdown to 2015 Coordination Committee, Geneva: WHO. 8. International Telecommunication Union (ITU) (2009) Information Society Statistical Profiles 2009. Africa. ITU, Genève. Available at: http://www.itu.int/ITU‐D/ict/material/ISSP09‐AFR_finalen.pdf. 9. United Nations Conference on Trade and Development (UNCTAD) (2010) Information Economy Report 2010 – ICTs, enterprises and poverty alleviation. United Nations, New York and Geneva. Available at: http://www.unctad.org/en/docs/ier2010_embargo2010_en.pdf. 10. Mechael PN, Batavia H, Kaonga N et al. (2010) Barriers and Gaps Affecting mHealth in Low and Middle Income Countries: Policy White Paper. Center for Global Health and Economic Development Earth Institute, Columbia University, New York. 11. Samai O & Sengeh P. Facilitating emergency obstetric care through transportation and communication, Bo, Sierra Leone. Int J Gynecol Obstet 1997;59:S157–S164. 12. Musoke MGN. Some Information and Communication Technologies and Their Effect on Maternal Health in Rural Uganda. A summary of research findings prepared for the African Development Forum 1999, Addis Abeba. 13. Musoke MGN (2002) Maternal Health Care in Rural Uganda: Leveraging Traditional and Modern Knowledge Systems. IK Notes No.40, World Bank, Washington DC. 14. Matthews MK, Walley RL. Working with midwives to improve maternal health in rural Ghana. Canadian Journal of Midwifery Research and Practice 2005;3:24–33. 15. Lungu K, Ratsma YEC. Does the upgrading of the radio communications network in health facilities reduce the delay in the referral of obstetric emergencies in Southern Malawi? Malawi Med J 2007;19:1–8. 16. Fournier P, Dumont A, Tourigny C, Dunkley G, DraméS. Improved access to comprehensive emergency obstetric care and its effect on institutional maternal mortality in rural Mali. Bulletin of the World Health Organisation 2009;87:30–38. 17. Krasovec K. Auxiliary technologies related to transport and communication for obstetric emergencies. International Journal of Gynecology & Obstetrics 2004;85(Suppl. 1):S14–S23. 18. Lee ACC, Lawn JE, Cousens S et al. Linking families and facilities for care at birth: what works to avert intrapartumrelated births? International Journal of Gynecology and Obstetrics 2009;107(Suppl. 1): S65‐S85. 19. UNFPA (2007) Report on the regional conference on obstetric fistula and maternal health. UNFPA regional conference on obstetric fistula and maternal health, 10–13 December 2007, Nouakchott. Available at: http://www.fistulanetwork.org/FistulaNetwork/user/ReportNKKT_Final‐Version.pdf. 20. Chib A, Lwin MO, Ang J, Lin H, Santoso F. Midwives and mobiles: using ICTs to improve healthcare in Aceh Besar, Indonesia. Asian Journal of Communication 2008;18:348–364. 21. Chib A. The Aceh Besar midwives with mobile phones project: Design and evaluation perspective using the information and communication technologies for healthcare development model. Journal of Computer‐Mediated Communication 2010;15:500–525. 22. Holmes D. Rwanda: an injection of hope. Lancet 2010;376:945–946.

80 Improvement of maternal health services through the use of mobile phones

23. Svoronos T, Mjungu D, Dhadialla P et al. (2010) CommCare: Automated Quality Improvement To Strengthen Community‐Based Health. Available at: http://d‐tree.org/wp‐ content/uploads/2010/05/Svoronos‐Medinfo‐CommCare‐safepregnancy1.pdf. 24. Lund S (2009) Mobile Phones can Save Lives. Profile ⁄ Global Health, University of Copenhagen, Copenhagen, 18–19. Available at: http://www.e‐pages.dk/ku/307/18. 25. Lund S (2010a) Wired Mothers – Use of Mobile Phones to Improve Maternal and Neonatal Health in Zanzibar’. Enreca Health. Available at: http://www.enrecahealth.dk/archive/ wiredmothers/. 26. Lund S (2010b) Personal communication via email, 10 June 2010. 27. Mechael PN (2009) MoTECH: mHealth Ethnography Report. Dodowa Health Research Center for The Grameen Foundation, Washington DC. 28. Mailman School of Public Health (2010) MoTeCH: A Comprehensive Overview. Colombia University, New York. 29. GSMA Development Fund, Cherie Blair Foundation for Women&Vital Wave Consulting (2010) Women & Mobile: A Global Opportunity. A Study on the Mobile Phone Gender Gap in Low and Middle‐Income Countries. GSMA, London. 30. Macueve G, Mandlate J, Ginger L, Gaster P & Macome E. Women’s use of information and communication technologies in Mozambique: a tool for empowerment? In: African Women & ICTs, 1st edn (eds Buskens I & Webb A) Zed Books Ltd, London, 2009: 21–32. 31. Hellström J (2010) The innovative use of mobile applications in East Africa. Sida Review 2010:12, Stockholm. Available at: www.upgraid.files.wordpress.com/2010/06/sr2010‐12_sida_hellstrom.pdf. 32 Comfort K, Dada J. Rural women’s use of cell phones to meet their communication needs: a study from northern Nigeria. In: African Women & ICTs, 1st edn (eds Buskens I & Webb A) Zed Books Ltd, London, 2009:44–55.

81 Chapter 5

82 CHAPTER 6

Improving care‐seeking for facility‐based health services in a rural, resource‐limited setting: Effects and potential of an mHealth project

Higgins‐Steele A, Noordam AC, Crawford J, Fotso JC African Population Studies 2015;28:1643‐1662

83 Chapter 6

ABSTRACT

The aim of this paper was to investigate the impact of a toll‐free hotline and mobile messaging service on care‐seeking behaviours. Due to the low uptake of the services, the treatment on the treated estimate is used. For maternal health, the intervention had a strong, positive impact on antenatal care initiation and skilled birth attendance. No effect was observed for postnatal check‐ups, receiving the recommended four antenatal care visits and vitamin A uptake. A negative effect was observed on tetanus toxoid coverage. For child health, no change was seen in child immunization, and a significant decrease was observed for care‐seeking for children with fever. Different factors are associated with care‐seeking, which may explain in part the variations seen across care‐seeking behaviours and possible influence of exogenous factors. Introduction of mHealth services for demand generation require attention to local health systems to ensure adequate supply and quality are available.

84 Improving care‐seeking for facility‐based health services

INTRODUCTION

In Malawi, despite a consistent reduction over the last two decades, infant and under‐ five mortality rates remain high.1,2 Under‐five mortality in Malawi was 71 per 1,000 live births in 2012, down from 112 in 2010 and 234 in 1992.2 Maternal mortality on the other hand dropped only minimally, with the Millennium Development Goal target still about six times lower than the current level of around 675 maternal deaths per 100,000 live births.3‐5 While a package of effective, health facility and community‐based, interventions could significantly improve maternal and child health (MCH) outcomes, uptake of these services remains low.6 For example, less than half of pregnant women in Malawi receive the four recommended antenatal care (ANC) visits and only 12% of women attend ANC within the first trimester of pregnancy. Also, nearly 30% of deliveries are not attended by a skilled birth attendant (SBA) and almost half of women do not receive postnatal care (PNC). And, nearly one third of children with symptoms of malaria or acute respiratory infections (ARI) are not taken to a health facility for advice or treatment.1 To accelerate progress towards improved MCH outcomes, global commitments and national efforts triggered the implementation of new strategies and innovative solutions in low‐resource contexts.7 One such strategy is the use of mobile phones to improve the delivery of health services, also referred to as mHealth. As mobile phone ownership in low‐ and middle‐income countries has grown substantially in the last decade, its potential to improve health services by enabling faster modes of communication is widely recognized.8‐9 Use of mobile phones – for example through two‐way communication, voice messages and/or short message services (SMS) – can overcome persistent health system constraints across the continuum of care.10 In low‐ resource settings, mHealth interventions targeting groups in the general population have shown the potential to improve adherence to treatment protocols, promote healthy behaviour, increase utilization of health services, and increase access to health information.11‐13 To adopt mHealth strategies to increase the utilization of facility‐based services requires measuring and understanding health‐seeking behaviour, while also acknowledging the challenges and shortcomings of health care delivery in resource‐ limited settings.14 Appropriate utilization of facility‐based MCH services is not only linked to demand‐side barriers, such as lack of knowledge and cost of those services, but also to supply‐side barriers such as human resources and availability of medicines.15 For example, care may be sought but appropriate treatment may not be provided.16 In Malawi, among other barriers, challenges linked to accessing services include long distances to facilities, poor perceptions of quality, and lack of human resources.17‐18 Moreover, women and caregivers of young children lack access to health information for decision‐making which leads to delays in seeking care.19‐20

85 Chapter 6

Previous studies have highlighted ways in which mHealth and health promotion can be successful in improving healthy behaviours, positively influencing timely care‐ seeking.21‐22 Strategies such as SMS to deliver health‐related messages23 and hotline systems to enable two‐way communication between individuals and health workers have helped overcome some of the challenges surrounding utilization of health facility‐ based services.24‐25 More specifically for Malawi, mHealth solutions have been limited largely to strengthening provider‐to‐provider communications and data reporting. In two districts in Malawi, an intervention aimed at connecting community health workers and district level officials found that mHealth can aid in dissemination of new information down to the health worker level, improve reporting on service delivery data, and save time and money on transportation costs.26 Another study in Malawi reported on a mHealth pilot project designed to facilitate communications between community health workers and their supervisors through SMS. This study demonstrated that the SMS project improved communication systems for several common applications such as reporting patient adherence, queries to supervisors on complicated symptoms or other issues, requests for medicines, and emergency transport referrals.27 MHealth services to share information between the health worker or health system and the individual have only been documented, yet not studied extensively in Malawi.14 Despite its potential, recent reviews of mHealth projects in low‐resource settings highlight poor evaluations for desired outcomes and impact of these projects,21,28‐30 which includes effects on care‐seeking behaviours. This paper analyzes data from a study using a pre‐post‐test design with a comparison group to examine changes in care‐seeking behaviours among women and caregivers of young children after the implementation of an mHealth initiative. We investigate the impact of the intervention on facility‐based care‐seeking for MCH. These findings are important to further guide the integration of mobile technology into health services. The paper is part of a series analyzing the effectiveness of a hotline and text messaging service named Chipatala Cha Pa Foni (CCPF) or “health centre by phone”, implemented between 2011 and 2013 in the catchment areas of four health centre’s in Balaka district, Malawi. The purpose of these services was to inform women and caregivers on essential care and encourage appropriate care‐seeking during pregnancy, childbirth and infancy.

DATA AND METHODS

Data This paper draws on the evaluation data of the CCPF project in Balaka district, Malawi. The project aimed to increase knowledge and behaviour relating to recommended MCH home‐based and facility‐based care. The CCPF project included a toll‐free hotline

86 Improving care‐seeking for facility‐based health services service providing protocol‐based health information, advice and referrals. Users could access this service to seek advice and information regarding the illness of their child under the age of five. The project also included an automated and personalized tips & reminders mobile messaging service, for which subscribers could opt‐in, with a choice of two local languages. For those community members who did not have a mobile telephone, volunteers were selected from communities and provided with a mobile telephone as a way to provide access to them. Box 6.1 provides some examples of automated tips & reminders messages sent to CCPF subscribers, which include women who signed up and received messages via a volunteer equipped with a phone for community use. The project was implemented between July 2011 and June 2013 in Balaka district, an area with some of the poorest MCH indicators in the country.23 The project is described in more detail in the second paper in this series by Larsen Cooper et al, under review.31 The evaluation used a quasi‐experimental design, with catchment areas of two health centre’s in the contiguous Ntcheu district as the control site. The control district was selected based on similar characteristics with the intervention site and since other districts either had dissimilar characteristics or had other MCH projects ongoing that may influence responses of participants. Once the health centre’s in the intervention and control sites were selected for the study, GIS information was used to map catchment areas of the health facilities and create a comprehensive list of villages in each of the catchment area. GIS information also provided mean distance from each village to the nearest health center.32 The core of the evaluation data consisted of cross‐sectional baseline and end line household surveys conducted in June‐July 2011 and April‐May 2013, respectively. Three questionnaires (household, woman and under‐five) were developed, covering more than 30 MCH indicators largely drawn from the Multiple Indicators Cluster Survey (MICS).32 As it is the rule, the same questionnaires were used at both baseline and end line, with the exception of an additional exposure module administered at end line only with questions referring the CCPF project which began after baseline data was collected. At baseline, a total of 2,840 women (1,119 in the control site and 1,721 in the intervention site) and 3,605 children under‐five (1,385 in the control and 2,220 in the intervention) were surveyed. At end line, a total of 3,853 women (2,509 in the intervention) and 3,261 children were surveyed. Based on the population size in the catchment area, it was determined during planning for the baseline that for statistical significance a minimum of 1,600 subjects was needed per group for the intervention area and a minimum of 1,200 per group for the control.

87 Chapter 6

Box 6.1 Examples of automated tips and reminders messages.

Messages for pregnant women 6 weeks: “When you and your family know that you are pregnant, a visit to ANC will help you understand everything you need to do to keep the baby healthy.” 10 weeks: “The ANC is your partner in the pregnancy. It is important to go to all 4 of your visits to use the tablets that they have given to you.” 20 weeks: “At ANC visits, you will have a Tetanus Toxoid (TT) vaccine during pregnancy, to stop you or your baby from getting tetanus, which is a serious infection.” 32 weeks: “When you deliver at the hospital, the baby will get everything he needs to start life healthy. Make sure you have a plan to get to the hospital when it is time” “Baby can come anytime now. Do you have all you need for delivery packed in a bag? Pack napkins, cloths, baby soap, towels, basin and clothes for you.” 41 weeks: “Be sure to take your baby to the health centre 6 weeks after delivery. Your baby will get checked and receive more immunizations.”

Messages for mothers for their infant 1 week: “Make sure your baby has its vaccination. In the first week, your baby should get polio vaccine by mouth and the BCG vaccine against TB by injection.” 6 weeks: “This week your baby needs to receive 2 vaccines; OPV and DTP‐HepB‐Hib. She will receive these vaccines 2 more times. It very important to protect your baby.” 11 weeks: “Remember to have your baby sleep under a treated mosquito net every night to prevent malaria and take him to the clinic straight away if he has fever.” 14 weeks: “It is now time for the third and final dose of OPV and DTP‐HepB‐Hib vaccines for your baby. It is important that these are given 4 weeks apart.” 17 weeks: “Prevent pneumonia by protecting your baby from breathing smoke from rubbish or cooking fires and tobacco. Get treatment immediately if baby has fast breathing”

Variables of interest In this paper, we analyze the following aggregate and individual variables related to facility‐based care for MCH: . Facility‐based maternal health care (for women who had a live birth in the last 18 months) ‐ Received the correct dosage of the tetanus toxoid (TT) vaccine during the last pregnancy ‐ Received a Vitamin A dose during the last pregnancy ‐ Received the recommended four antenatal care (ANC) consultations during the last pregnancy ‐ Started ANC in first trimester during the last pregnancy ‐ Gave the last birth under the supervision of a skilled birth attendant ‐ Received one postnatal care (PNC) check‐up within 2 days of the last birth . Facility‐based child health care ‐ Child was fully immunized by first birthday (children aged 12‐23 months) ‐ Child with symptoms of acute respiratory illness (ARI) in the last two weeks preceding the survey sought care at facility

88 Improving care‐seeking for facility‐based health services

‐ Child with fever in the last two weeks preceding the survey sought care at facility The control variables included well known confounders at the community, household, woman and child levels, the selection of which was guided by literature reviews. These variables, as can be seen in Tables 6.1 and 6.2, include at the household level, wealth status, number of under‐five children, and ethnicity and religion of the head of household. At the woman level, these are age, marital status, education, and access to a personal phone. Child characteristics are age and sex. Finally, at the community (village) level, the analyses control for the mean distance to the nearest health centre.

Table 6.1 Percentage distribution of mothers/caretakers of children under 5 and pregnant women Baseline Control Intervention Endline Community level covariates Mean distance to the health centre (km) 5.7 4.4 4.8 Household level covariates Household wealth Poor (lowest 50%) 56.6 53.0 55.3 Rich (highest 50%) 43.4 47.0 44.7 No. of children under the age of 5 yrs. 0 5.7 4.5 29.5 1 66.6 60.8 52.3 2+ 27.7 34.7 18.1 Ethnicity of the head of household Lomwe 6.0 21.0 16.3 Ngoni 77.3 20.6 41.1 Yao 6.8 38.9 28.7 Other 9.9 19.5 13.9 Religion of the head of household Catholic 16.1 17.5 18.3 Other Christian 70.1 41.0 50.9 Muslim 5.5 37.0 25.7 Other/ No Religion 8.4 4.5 4.8 Women‐level covariates Cell phone in Household No 77.6 68.0 68.5 Yes 22.4 32.0 31.5 Education None 16.5 15.4 13.2 Primary 73.7 74.0 73.8 Secondary+ 9.7 10.6 13.1 Marital status Not in union 11.0 17.0 24.4 In union 89.0 83.0 75.6 Age in years <20 10.6 11.3 16.0 20‐29 57.2 53.9 41.7 30+ 32.2 34.7 42.3 N 1,119 1,721 3,853

89 Chapter 6

Table 6.2 Percentage distribution of children under 5 Baseline Control Intervention Endline Age in months < 12 25.4 23.2 24.2 12‐23 21.2 19.7 22.7 24+ 53.4 57.0 53.2 Sex Male 48.4 51.3 51.5 Female 51.6 48.7 48.5 N 1,365 2,220 3,261

Data analysis Descriptive and multivariate analyses are employed to quantity the impact of the intervention on the outcomes of interest. Firstly, we estimate the simple difference‐in‐ difference (DID) as follows:

Where and represent the average outcome at endline in the intervention site and control area, respectively, and and the average outcome at baseline in the intervention site and control area, respectively. Since this estimate compares the intervention and the control sites regardless of the use of the services offered by the project, it is to be interpreted as intention‐to‐treat (ITT) effect.33 Secondly, to adjust for potential confounding variables we conduct multivariate DID defined as follows: Yivt = 0 + 1 Tv + 2 Pt + 3 (T * P)vt + Wivt  + Xv  + ivt where Yivt is the outcome measure for woman/child i, in village v, at time t. Tv is a dummy variable taking the value 1 for individuals in treatment areas and 0 for individuals in control areas; Pt is a dummy variable taking the value 0 for the baseline data and 1 for the endline data; Wivt is a vector of the controls at the household, women and child levels; X is a village‐level control variable; and ivt is the idiosyncratic error, clustered by health centre catchment area. Thirdly, we further analyze the data to assess the effect of the program on its users, applying the so‐called “treatment effect on the treated” (TOT) model. This analysis is of specific importance, especially if the uptake of the intervention is not very high. The method uses instrumental variable analyses to construct a proper counterfactual – the women who would have used the services in control communities had they been offered.34 More detailed information on the research methodologies including study limitations can be found in the first paper of this series.35

90 Improving care‐seeking for facility‐based health services

Ethical clearance Ethical approval for the study was granted by the National Health Sciences Research Committee, Malawi Ministry of Health.

RESULTS

Sample characteristics Table 6.1 shows the characteristics of women interviewed at baseline in the intervention and comparison areas. The distribution of women by household wealth, number of children under the age of five years, education, marital status and age appears similar across the intervention and control communities. As can be seen, a majority of women were from households with one child under‐five years of age; about three‐quarters of women had completed primary education; and more than 80% were in union. The mean distance to the nearest health facility was slightly shorter in the intervention site compared with the control area (4.4 km versus 5.7 km). The distribution by ethnicity, religion and access to a mobile phone exhibits differences, with the control area dominated by the Ngoni ethnic group (77.3%) and non‐Catholic Christians (70.1%), while the intervention site shows a seemingly more balanced distribution across the ethnic and religious groups. Access to phone was higher in the intervention area (32%) than in control communities (22%). Table 6.1 also shows that the overall baseline and endline samples had similar characteristics. There are two noticeable exceptions. The proportion of women in union dropped slightly from an average of 86% at baseline to 75.6% at end line. Access to phone on the other hand, improved from around 28% to 31.5% during the same period. The sex and age of the sample of children under the age of five are presented in Table 6.2. As can be seen, the distribution is similar across the intervention and control groups at baseline, and between the baseline and end line samples. Table 6.3 shows that at end line, the awareness of the hotline was high in the intervention area, at around 77% among the total sample of 2,509 women. Among individuals who heard about the hotline (N=1,929), less than 24% used the services, a proportion which represent about 18% of the total sample of women at end line. Awareness of the mobile messaging system was substantially lower, at 33.3%, and use was estimated at 22.6%.

91 Chapter 6

Table 6.3 Awareness and use of the services among women of child bearing age at endline. Intervention area Control area % N % N Awareness and use of the hotline services Heard about the services 76.9 2,509 2.8 1,344 Used the services 23.8 1,929 3.5 38 Awareness and use of the Mobile messaging services Heard about the services 33.3 2,509 0.4 1,344 Used the services 22.6 835 0.0 5

Effect of the intervention on care‐seeking behavior for maternal health Table 6.4 shows the levels of the selected maternal health indicators at baseline and endline and across the intervention and control areas, as well as the resulted difference‐in‐difference, adjusted and unadjusted. Three variables – TT vaccine, skilled birth attendance (SBA) and vitamin A – had high coverage in both sites, while PNC, and to a lesser extent, ANC initiation in the first trimester of pregnancy, displayed a low coverage. The proportion of pregnancies which received the recommended four ANC visits, on the other hand, ranged from 55% to 64%. The unadjusted and adjusted DID results indicate that the intervention did not have any ITT effect on either of the indicators for facility‐based maternal care, and as a result, did not affect the aggregate maternal health indicator. Given the low use of services (only 18% of women used the services), the TOT model is a more appropriate approach, as it adjusts for the fact that some individuals in the intervention area did not use the services, and some others in the control area would not use the services even if offered. As Table 6.4 shows, the TOT model reveals a markedly different pattern. The intervention had a strong, positive impact on ANC initiation during the first trimester of pregnancy. While the increase between baseline and end line did not vary across the two areas (about 12 percentage points increase) hence a negligible DID, the focus on women who used the services reveals a strong impact of the intervention. The project also had a positive effect on SBA, despite a similar margin of increase (about 5 percentage points) across the control and the intervention sites as observed for ANC initiation. At the other end of the scale, there is a negative TOT effect of CCPF on TT vaccine during pregnancy. The indicator remained almost unchanged in the intervention site, but declined by about 4.5 percentage points in the control area, yielding a positive though not statistically significant DID. Table 6.4 also shows an absence of TOT effect on PNC check‐up, receiving the recommended four ANC visits, and vitamin A. Effect of the intervention on care‐seeking behavior for child health The content of Table 6.5 which presents the results on child health is similar to that of Table 6.4.

92 Improving care‐seeking for facility‐based health services

18

sample last ‐

in

sub

(N=2,813)

birth

Same Same Same Same Same Same live

a

months

Had Corresponding health

in

effect

maternal

0.055 0.083 0.022 0.104** 0.239** 0.110** 0.444*** for TOT

difference) (Difference analysis

care

based ‐

‐ Multivariate (Difference facility

0.012 0.079 0.069 0.085 0.016 0.008 0.040 on difference)

in effect

ITT

interventions

the

(Difference of

0.010 0.085 0.068 0.092 0.017 0.009 0.044 ‐ difference)

in analysis effect

effects

ITT

(TOT)

9.1% 5.3% 5.4% 0.5% 55.4% 63.8% 54.9% 71.5% 96.1% 96.1% 85.8% 38.9% 32.3% 88.4% Descriptive Endline

treated

‐ the 0.4% 3.9% 5.7%

61.7% 61.9% 63.2% 72.9% 91.9% 90.8% 86.0% 26.4% 20.6% 93.0% Baseline 0.0% on

***p<0.01

Control Control Control Control Control Control Control treatment

Intervention Intervention Intervention Intervention Intervention Intervention Intervention

and

of 4

**p<0.05,

during (ITT)

up ‐ for

pregnancy dose treat supervision

dosage

trimester A

*p<0.10, to check

care

the

attendant birth

first during

PNC of recommended correct

in

based Vitamin birth ‐ under

health the a one the

Intention

days

ANC

2 vaccine

significance: birth

consultations facility skilled

pregnancy

TT

a 6.4

Received ANC Received Received Gave last Received maternal of Started the within

3. Overall 2. 6. 5. 1. 4.

Table Statistical

93 Chapter 6

or

ARI

sample weeks fever

(N=1,610)

of

2

(N=4,068)

sub had

last ARI

or

in

of months

23, (N=2,194) (N=1,895) ‐ 23 symptoms

‐ fever 12

12

Had

Had Corresponding symptoms Aged Aged health.

in

child

for effect

0.083 0.012 0.499*** 0.499*** care TOT difference)

(Difference analysis

based

‐ ‐ facility ‐

Multivariate on

(Difference

0.025 0.009 difference) 0.181*** 0.171***

‐ in effect

ITT

interventions

the

of ‐

(Difference

0.030 0.013 ‐ effects difference) 0.189*** 0.172***

‐ ‐ in analysis effect

(TOT)

ITT

0.090 treated 0.781 0.627 0.521 0.758 0.641 0.708

Descriptive Endline 0.037

the

on

0.788 0.591 0.045 0.675 0.777 0.639 0.676 Baseline 0.000

***p<0.01

treatment

Control Control Control and

Intervention Intervention Intervention Intervention Control

**p<0.05,

(ITT)

by

ARI

for 2

of treat

sought

care *p<0.10, to last

care

in

who immunized

based sought

‐ fever symptoms

fully

Intention weeks

who 2

significance:

with was

with health

birthday facility

last 6.5

Child weeks first Child Child child in care

Overall 1. 3. 2.

Table Statistical

94 Improving care‐seeking for facility‐based health services

Large, negative TOT effects are seen for the aggregate facility‐based child health indicator (p<0.01). This negative effect results exclusively from, and reflects, a reduction in the rate at which children with fever in the intervention communities sought treatment at a health facility. Indeed, the proportion of children with fever who visited a health facility dropped by 15.4 percentage points in the intervention site (from 67.5% to 52.1%), and by contrast, increased by 3.6 percentage points in the control area (from 59.1% to 62.7%), hence a DID of about 19 percentage points (p<0.01). The project did not have any ITT or TOT effect on the full immunization. Its coverage did not change noticeably over time, remaining at around 78% in the intervention group and at 77% in the control area. A similar pattern was observed for facility care for ARI symptoms.

DISCUSSION

Characteristics between intervention and control samples are similar for household‐ and women‐level covariates with the exceptions of ethnicity, religion and access to a mobile phone which show differences between groups sampled in Balaka and Ntcheu districts. It is noted elsewhere that differences in ethnicity and religion do not appear to affect MCH outcomes in Malawi.32,36 Utilization and access The overall use of the CCPF services was low; only 18% of the targeted population or less than 24% of those who were aware of the service used the hotline and the use of the tips & reminders service was lower. A possible reason for the low uptake of the service may be access to a mobile phone in the household was less than one‐third. For tips & reminders, women may have been more inclined to use the hotline service via a community volunteer phone than to sign up for tips & reminders on one of these shared phones. Importantly, the rural location of Malawi was selected, not in relation to mobile phone penetration, but because of poor MCH indicators that this pilot would seek to address. The pilot illustrates that even with enablers introduced by the pilot to increase access to the mHealth services – namely through community activities to enhance knowledge of the services and volunteers to facilitate access – use of the services was not widespread in the target groups. Other studies confirm that uptake of new and potentially successful solutions aiming to improve access to health care and information can be low. Moreover, there is insufficient evidence on facilitators and barriers to the use of mobile phones for health‐ related needs.37 Besides inadequate knowledge of how the new service works and what benefits are of using it,38 reasons for non‐use of mobile phones have been linked to privacy concerns and network coverage.39‐40

95 Chapter 6

Due to the low uptake of the CCPF project, its benefits can only be measured by assessing the impact on those who actually used the service, i.e., through a TOT model, a notable finding with implications for measurement of other mHealth projects. Maternal health The TOT model showed that, among the women who used the service, there was a significant increase in ANC initiation within the first trimester and SBA during delivery. ANC as well as SBA are crucial indicators for monitoring the wellbeing of the mother and child and ensuring timely identification of danger signs, birth preparedness as well as access to life‐saving interventions when needed.41 Yet, according to Malawi’s Demographic and Health Survey (DHS 2010), the majority of women start ANC late, only 12% nationally1 and 21‐26% in our study area, attended ANC in the first trimester. SBA is much higher, nationally at around 71% and 91‐92% in our study area. Studies conducted in Malawi suggest that delays in ANC are (often) due to superstitious beliefs regarding the consequences of disclosing pregnancy in the first trimester.42‐43 Access to information through the hotline and tips & reminders service may therefore have resulted in an increase in timely utilization of these services, either by providing information or facilitating referrals. The intervention had no TOT effects on maternal indicators such as the use of Vitamin A, attending ANC at least four times during pregnancy and receiving PNC. In the long run, the increase in timely access to ANC might have a positive effect on these indicators, especially the use of Vitamin A and attending all four ANC visits. This assumes regular and adequate stock of Vitamin A and health workers administering the supplement during an ANC visit. For PNC, national coverage in Malawi is 43%, yet in our study area this was as low as 4‐6%. This large difference may be explained by variations in question phrasing between the DHS questionnaire and the Multiple Indicator Cluster Survey (MICS) indicator selected as part of this study. The PNC question for this study did not include home‐based PNC visits, for example, and focused on how long after birth PNC was received. Even with these questionnaire differences, further research is needed to better understand why coverage of PNC is so low in the study area, whether this is due to structural reasons related to the local health system (e.g., lack of health resources) or associated with demand side barriers. Finally, a negative effect was observed for women receiving TT vaccines. A decrease in TT coverage was found in both the intervention and control areas. While not specifically examined in this study, this could be related to limited availability of TT vaccines or health worker reluctance to vaccinate pregnant women. A study in Kenya suggests that improving access to ANC services is a necessary condition but not sufficient for improving uptake of TT immunization and suggests areas of future research to include the quality of provider‐client interaction, availability of stocks of TT and women’s perception of TT immunizations.44

96 Improving care‐seeking for facility‐based health services

Overall, the CCPF project had a positive impact on maternal health indicators, which is important, as a combination of these interventions – and specifically increase in SBA – are needed to reduce the still high mortality rate in Malawi. If an increased proportion of women started ANC earlier, the health system could be strengthened to deliver other pregnancy‐related interventions, such as to encourage repeat ANC visits to fulfil recommended four visits, Vitamin A supplementation and TT vaccination, and importantly encourage SBA. A multi‐country study found significant and positive effect on the number of antenatal consultations on SBA during child birth [45]. Effective strategies, which not only include mobile phone use, need to be identified to ensure timely access to and supply of these and other essential services. Adjiwanou and LeGrand suggest SBA can be enhanced through improvements in quality of ANC, notably adequate information conveyed on the importance of SBA by the health care provider and provision of services closer to populations in need.45

Child health There was no TOT effect of the CCPF project on coverage of immunization and care‐ seeking for children with signs of ARI presumably because baseline values for both variables were relatively high. The percentage of children fully immunized in the study area was estimated at around 78%, a figure comparable to the national estimate from the 2010 DHS. For care‐seeking for ARI, the national coverage was around 70% as opposed to 65% in our study area. While vaccine coverage in Malawi has improved, children are often vaccinated at a later age. Reasons for this include that caregivers believe children were too young to be vaccinated,46 poor recognition of danger signs and illness severity, and financial considerations limiting timely access to care.47 For both the increase in coverage of vaccinations as well as the recognition of ARI, SMS services could inform caregivers about appropriate timing of vaccinations. Moreover, hotline services could potentially ensure timely referral for ARI, although this was an aim of CCPF and demonstrated negative effects. Our findings show a negative effect of facility‐based care‐seeking for fever. Other studies show that children with fever in Malawi are often treated at home.48‐49 According to Ministry of Health protocols, hotline workers only refer children with a fever to the nearest village clinic or health centre for diagnosis and treatment if the fever is being presented as a danger sign and has persisted for seven days or if the fever is accompanied by other symptoms. It is possible that some children with a fever were appropriately treated at home and did not require a visit to a health facility. This is confirmed by findings on home‐based care for the CCPF project, published in the third paper of this series,50 suggesting that caregivers were well equipped to handle conditions like fever at home and avoid unnecessary trips to the facility.

97 Chapter 6

Conclusion When used by the targeted population, mHealth services providing the user options of how to access information on when and where to seek facility‐based care have potential for behaviour change. Changes were seen in a relatively short duration of this pilot, evaluated after less than two years after its introduction. Different factors are associated with care‐seeking for outcome variables measured, which may explain in part the variations seen across care‐seeking behaviours and possible influence of exogenous factors. MHealth can be a useful channel to improve care‐seeking, though reasonable expectations are required in terms of uptake and use by target groups, which can be low even with demand generation activities for the service, as well as of effects on demand for MCH services. Introduction of mHealth services for demand generation and information on care‐seeking require attention to local health systems to ensure adequate supply and quality are available.

Acknowledgement The CCPF project is part of Innovations for Maternal, Newborn & Child Health, an initiative of Concern Worldwide U.S. funded through a multi‐year grant from the Bill & Melinda Gates Foundation. The Government of Norway and the United Nations Foundation also supported the Malawi mHealth project (CCPF) through the Innovation Working Group Catalytic mHealth Grants program as part of the UN Secretary General’s Every Women Every Child strategy. The project was implemented by VillageReach, an international NGO headquartered in Seattle, USA. We would like to give special thanks to the Reproductive Health Unit and its Director, Mrs. Fannie Kachale, and the Balaka District Health Office for their support of CCPF. The evaluation was conducted by Invest in Knowledge Initiative (IKI), a Malawi‐based research institution, with the leadership of Professor Susan Watkins of University of Pennsylvania and Dr. Amanda Robinson of Ohio State University. The authors would like to thank Dr. Amanda Robinson for her contribution to data analysis, Dr. Linda Vesel of Concern Worldwide US for reviewing the manuscript, and the anonymous reviewers for their comments.

98 Improving care‐seeking for facility‐based health services

REFERENCES

1. National Statistical Office (NSO) and ICF Macro. Malawi Demographic and Health Survey 2010. Zomba, Malawi, and Calverton, Maryland, USA: NSO and ICF Macro, 2011. 2. United Nations Children’s Fund (UNICEF). Committing to child survival: A promise renewed progress report 2013. UNICEF, New York, 2013. 3. Government of Malawi (GoM). 2010 Malawi Millennium Development Goals Report. Lilongwe, Malawi, 2010. 4. Colbourn T, Lewycka S, Nambiar B, Anwar I, Phoya A, et al. Maternal mortality in Malawi, 1977‐2012. BMJ Open 2013;3:e004150. 5. United Nations (UN). The Millennium Development Goals Report 2013. United Nations, New York, 2013. 6. Darmstadt GL, Bhutta ZA, Cousens S, Adam T, Walker N, et al. Evidence‐based, Cost‐effective Interventions: How Many Newborn Babies Can We Save? Lancet 2005;365:977‐988. 7. Darmstadt GL, Marchant T, Claeson M, Brown W, Morris S, et al. A strategy for reducing maternal and newborn deaths by 2015 and beyond. BMC Pregnancy and Childbirth 2013;13:216. 8. Vital Wave Consulting. mHealth in the global south: Landscape analysis. Washington D.C. and Bershire, UK: Vodafone Foundation –United Nations Foundations Partnership, 2008. 9. Noordam AC, Kuepper BM, Stekelenburg J, Milen A. Improvement of maternal health services through the use of mobile phones. Trop Med Int Health 2011;16:622‐626. 10. Labrique AB, Vasudevan L, Kochi E, Fabricant R, Mehl G. mHealth innovations as health system strengthening tools: 12 common applications and a visual framework. Glob Health Sci Pract. 2013;1:160‐171. 11. Cole‐Lewis H, Kershaw T. Text messaging as a tool for behavior change in disease prevention and management. Epidemiol Rev 2010;32:56‐69. 12. Gurman TA, Rubin SE, Roess AA. Effectiveness of mHealth Behavior Change Communication Interventions in developing Countries: A Systematic Review of the Literature. J Health Commun 2012;17: (Suppl 1):82‐104. 13. Sloninsky D, Mechael PN. (2008) Towards the development of an mhealth strategy: A literary review. New York, USA: World Health Organization and Earth Institute. 14. Noordam AC, George A, Sharkey AB, Jafarli A, Bakshi SS, Kim JC. Assessing scale up of mHealth innovations based on intervention complexity: Two case studies of child health programmes in Malawi and Zambia. J Health Commun 2015;20:343‐353. 15. Chopra M, Sharkey A, Dalmiya N, Anthony D, Binkin N. Strategies to improve health coverage and narrow the equity gap in child survival, health, and nutrition. Lancet 2012;380:1331–1340. 16. Diaz T, George AS, Rao SR, Bangura PS, Baimba JB, et al. Healthcare seeking for diarrhoea, malaria and pneumonia among children in four poor rural districts in Sierra Leone in the context of free health care: Results of a cross‐sectional survey. BMC Public Health 2013;13:157. 17. Lohela TJ, Campbell OMR, Gabrysch S. Distance to care, facility delivery and early neonatal mortality in Malawi and Zambia. PLoS One 2012;7:e52110. 18. Kambala C, Morse T, Masangwi S, Mitunda P. Barriers to maternal health service use in Chikhwawa, Southern Malawi. Malawi Med J 2011;23:1‐5. 19. Geubbels E. Epidemiology of maternal mortality in Malawi. Malawi Med J 2006;18:208‐228. 20. Chibwana AI, Mathanga DP, Chinkhumba J, Campbell CH. Socio‐cultural predictors of health‐seeking behaviour for febrile under‐five children in Mwanza‐Neno district, Malawi. Malaria J 2009;8:219. 21. Chib A, van Velthoven MH, Car J. mHealth Adoption in Low‐Resource Environments: A Review of the Use of Mobile Healthcare in Developing Countries. J Health Commun 2014;27:1‐31. 22. Jennings L, Gagliardi L. Influence of mhealth interventions on gender relations in developing countries: A systematic review. Int J Equity Health 2013;12:85. 23. Crawford J, Larsen‐Cooper E, Jezman Z, Cunningham SC, Bancroft E. SMS versus voice messaging to deliver MNCH communication in rural Malawi: assessment of delivery success and user experience. Glob Health Sci Pract 2014;2:35‐46.

99 Chapter 6

24. Ivatury G, Moore J, Bloch A. A Doctor in your pocket: Health hotlines in developing countries. Innovations: Technology, Governance, Globalization 2009;4:119‐153. 25. Corker J. “Ligne Verte” Toll‐free hotline: Using cell phones to increase access to family planning information in the Democratic Republic of Congo. Cases in Public Health Communication & Marketing 2010;4:23‐37. 26. Lemay N, Sullivan T, Jumbe B, Perry CP. Reaching Remote Health Workers in Malawi: Baseline Assessment of a Pilot mHealth Intervention. J Health Commun 2012;17suppl 1:105‐117. 27. Mahmud N, Rodriguez J, Nesbit J. A text message‐based intervention to bridge the healthcare communication gap in the rural developing world. Technol Health Care 2010;18:137‐144. 28. Earth Institute. Barriers and gaps affecting mHealth in low and middle income countries: A policy white paper. Washington DC: mHealth Alliance, 2010. 29. Mechael PN. The case for mHealth in developing countries. Innovations, Technology, Governance and Globalization 2009;4:103‐118. 30. Tamrat, T. and Kachnowski, S. Special delivery: An analysis of mHealth in maternal and newborn health programs and their outcomes around the world. Matern Child Health J 2012;16:1092–1101. 31. Larsen‐Cooper E, Bancroft E, O’Toole M, Jezman Z. Where there is no phone: Extending the reach of mHealth to individuals without personal phones in Balaka District, Malawi.’ Under review. 32. Invest in Knowledge (IKI). IKI’s Evaluation Report of ICT for MNCHavailable at: http://innovationsformnch.org/knowledge‐center‐resources/iki‐evaluation‐report‐of‐ict‐for‐mnch (accessed May 27 2014), 2013. 33. Heckman JJ. The Scientific model of causality. Sociological Methodology 2005;35:1‐97. 34. Angrist JD, Imbens GW, Rubin DB. Identification of causal effects using instrumental variables. Journal of the American Statistical Association 1996;91:444‐455. 35. Fotso JC, Robinson AL, Noordam AC, Crawford J. Fostering the use of quasi‐experimental designs for evaluating public health interventions: Insights from an mHealth project in Malawi. African Population Studies 2015;29:1607‐1628. 36. NORAD Evaluations Department. Local perceptions, participation and accountability in Malawi’s Health Sector. Oslo, Norway, 2012. 37. Chib A, Wilkin H, Hoefman B. Vulnerabilities in mHealth implementation: a Ugandan HIV/AIDS SMS Campaign. Global Health Promotion 2011;20(Suppl):26‐32. 38. Marwa B, Njau B, Kessy J, Mushi D. Feasibility of introducing compulsory community health fund in low resource countries: views from the communities in Liwale district of Tanzania.’ BMC Health Serv Res 2013;13:298. 39. Chib A, Wilkin H, Ling LX, Hoefman B, Van Biejma H. You Have an Important Message! Evaluating the Effectiveness of a Text Message HIV/AIDS Campaign in Northwest Uganda. J Health Commun 2012;17 Supp 1:146–157. 40. Lepper AM, Eijkemans MJ, van Beijma H, Loggers JW, Tuijn CL, et al. Response patterns to interactive SMS health education quizzes at two sites in Uganda: a cohort study. Trop Med Int Health 2013;18: 516‐521. 41. Bhutta ZA, Das JK, Bahl R, Lawn JE, Salam RA, et al. Can available interventions end preventable deaths in mothers, newborn babies, and stillbirths, and at what cost?’ The Lancet 2014;384:347‐370. African Population Studies Special Edition, 2015 http://aps.journals.ac.za 1660 42. Pell C, Meñaca A, Were F, Afrah NA, Chatio S, et al. Factors Affecting Antenatal Care Attendance: Results from Qualitative Studies in Ghana, Kenya and Malawi. PLoS One 2013;8:e53747. 43. Banda CL. Barriers to utilization of focused antenatal care among pregnant women in Ntchisi district in Malawi. Available at: https://tampub.uta.fi/handle/10024/84640 (accessed May 27 2014), 2013. 44. Haile ZT, Chertok IR, Teweldeberhan AK. Determinants of utilization of sufficient tetanus toxoid immunization during pregnancy: Evidence from the Kenya Demographic and Health Survey, 2008– 2009. J Commun Health 2013;38:492‐9. 45. Adjiwanou V, LeGrand T. Does antenatal care matter in the use of skilled birth attendance in rural Africa: A multi‐country analysis. Soc Sci Med 2013;86:26‐34. 46. Minetti A, Kagoli M, Katsulukuta A, Huerga H, Featherstone A, et al. Lessons and challenges for measles control from unexpected large outbreak, Malawi. Emerg Infect Dis 2013;19:202‐209.

100 Improving care‐seeking for facility‐based health services

47. Geldsetzer P, Williams TC, Kirolos A, Mitchell S, Ratcliffe LA, et al. The recognition of and care seeking behaviour for childhood illness in developing countries: A systematic review. PLoS One 2014;9: e93427. 48. Kazembe LN, Appeleton CC, Kleinschmid I. Choice of treatment for fever at household level in Malawi: examining spatial patterns. Malar J 2007;6:40. 49. Holtz TH, Kachur SP, Marum LH, Mkandala C, Chizani N, et al. Care seeking behaviour and treatment of febrile illness in children aged less than five years: a household survey in Blantyre District, Malawi. Trans R Soc Trop Med Hyg 2003;97:491‐497. 50. Fotso JC, Bellhouse L, Vesel L, Jezman Z. Strengthening the home‐to‐facility continuum of newborn and child health care through mHealth: Findings from an intervention in rural Malawi. African Population Studies 2015;29:1663‐ 1683.

101 Chapter 6

102 CHAPTER 7

Assessing scale up of mHealth innovations based on intervention complexity: Two case studies of child health programmes in Malawi and Zambia

Noordam AC, George A, Sharkey AB, Jafarli A, Bakshi SS, Kim JC Journal of Health Communication 2015; 0: 1–11

103 Chapter 7

ABSTRACT

As interest in mHealth (including Short Message Services or SMS) increases, it is important to assess potential benefits and limitations of this technology in improving interventions in resource‐poor settings. We analysed two case studies (early infant diagnosis of HIV and nutrition surveillance) of three projects in Malawi and Zambia using a conceptual framework that assesses the technical complexity of the programmes, with and without the use of SMS technology. We based our findings on literature and discussions with key informants involved in the programmes. For both interventions, introducing SMS reduced barriers to effective and timely delivery of services by simplifying the tracking and analysis of data and improving communication between healthcare providers. However, the primary implementation challenges for both interventions were related to broader programme delivery characteristics (e.g. human resource needs and transportation requirements), which are not easily addressed by the addition of SMS. The addition of SMS technology itself introduced new layers of complexity. This article points the reader to technological complexities to consider when implementing health‐related SMS interventions so as to help encourage programmes that can be successful and sustainable. We argue that before deciding whether and how to introduce SMS, it is important to understand the underlying challenges within the existing programme, the potential contribution of SMS, where it might introduce new challenges, and how these can be addressed. Conceptual frameworks that analyse technical complexities can be useful to help reveal strategic needs regarding scale up in existing health systems.

104 Assessing scale up of mHealth innovations based on intervention complexity

INTRODUCTION

There is increasing interest in the use of information and communications technology (ICT) for health (eHealth) in low‐ and middle‐ income countries (LMIC), especially in the use of mobile phone technology for health (mHealth). Emerging evidence suggests that mHealth in the form of Short Message Service (SMS) or text messages can have positive implications in LMIC for improving routine service delivery in terms of timely data collection1 and quality case management.2 In addition, it has been argued that SMS technology presents an opportunity to reduce the complexity of health interventions and reduce inequities in service delivery.3 However, the evidence that is available is based on implementation of pilot projects that were not brought to scale.4 As more resources are invested in the specific use of SMS, there is an urgent need to better understand the implications of pairing this technology with existing interventions,5 including assessing the implications that this technology has for intervention coverage and the impact it has on health outcomes.6 In addition, it is important to better understand why most of these projects stay at a ‘small pilot project level’,4 with few progressing to large‐scale coverage in rural Africa.1 Such analysis is critical, especially in ensuring that investments effectively address the challenges of operating at scale in disadvantaged settings in LMIC. Using a previously developed framework to assess intervention technical complexity,7 we examine mHealth interventions implemented as part of three maternal, newborn and child health (MNCH) ‐related projects in Malawi and Zambia and use this evidence to explore why mHealth projects face constraints in going to scale.

METHODS

Conceptual framework to analyze the challenges of scaling up health programs A number of conceptual frameworks have been proposed to help analyse the challenges of bringing interventions to scale.7‐11 Although economic considerations are important, they are not the only limiting factor, as feasibility analyses can make an important contribution to increasing the likelihood of successfully scaling up interventions.12‐13 In 2005, Gericke et al presented a framework to assess the feasibility of scaling up an intervention.7 In addition to financial resources, this framework asserts that feasibility depends on the degree of the intervention’s technical complexity in four domains: 1) intervention characteristics, 2) delivery characteristics, 3) government capacity requirements and 4) usage characteristics (described in Table 7.1). This framework has previously been used to assess complexity of a range of health interventions including condom promotion,7 tuberculosis DOTS programs,7 diet

105 Chapter 7 improvement,14 and food poisoning risk reduction strategies,12 as well as identifying strategies to improve nutrition mainstreaming.15

Table 7.1 Four domains of technical complexity that can impact scale‐up of health interventions (Gericke et al, 20057). The domain Category Criteria Intervention Basic product design* Stability, standardizability, safety profile, ease of storage, ease characteristics of transport Supplies Need for regular supplies Equipment The need for high‐technology equipment and infrastructure, number of different types of equipment and maintenance Delivery Facilities Retail sector, outreach services, first‐level care and hospital Characteristics care Human resources Skill level required for service provision and supervision, intensity of professional services in terms of frequency or duration and management and planning requirements Communication and Dependence of delivery on communication and transport Transport infrastructures Government Regulation/ legislation Need for regulation, monitoring and regulatory measures and capacity regulation enforcement requirements Management systems Need for sophisticated management systems Collaborative action Need for intersectoral action within government, partnership between government and civil society and partnership between government and external funding agencies Usage Ease of usage Need for information/ education and supervision characteristics Pre‐existing demand Need for promotion Black‐market risk Need to prevent resale/ counterfeiting * The basic product design for the assessment of adding RapidSMS to the project is the mobile phone.

Selection of case studies We selected three projects supported by UNICEF (two in Malawi, one in Zambia) which combine an SMS platform called RapidSMS, a free and open‐source framework for dynamic data collection, logistics coordination and communication, with MNCH interventions. These projects were selected based on three criteria: 1. Representation of different public health challenges (early infant diagnosis (EID) of HIV and child malnutrition) of relevance to the Millennium Development Goals 2. Representation of diverse applications of SMS technology with different stakeholders (e.g. improving reporting of HIV test results and follow‐up with healthcare workers and mothers, and strengthening nutritional data surveillance) 3. Availability of data relevant to the four domains of technical complexity in the Gericke et al framework.7

106 Assessing scale up of mHealth innovations based on intervention complexity

Methods used to apply framework First a review of both published and unpublished literature was conducted to identify any existing evidence relating to the selected case studies as well as issues relating to implementation of EID and nutrition surveillance projects. This search was limited to English publications combining terms linked to: EID, HIV, nutrition surveillance, including linking these terms to Zambia and Malawi. Searches were initiated in PubMed and expanded to grey literature and reference lists. Literature was gathered to understand why the project was implemented and to assess the program’s complexity with respect to the four domains described by Gerick et al. This search was then expanded to include programs incorporating RapidSMS technology. Combined with the search terms above, the following terms were added: SMS, mHealth, mobile phones and RapidSMS. Because the existing evidence was limited, additional data were collected through discussions with two key informants: those individuals responsible for implementation of the RapidSMS component in the Malawi and Zambia programmes. In particular, these informants were asked an open‐ ended question regarding what they perceived to be the main strengths and challenges they faced in implementation. Following application of the framework, the informants were also asked to review and provide their inputs on the preliminary findings. The literature search was conducted by two independent researchers. To improve validity of the analysis, preliminary findings were shared and discussed with UNICEF technical experts and country‐level program managers and implementers in Malawi and Zambia. These inputs were incorporated into the final analysis. With this information, we then applied Gericke et al’s conceptual framework7 in order to: 1. Assess the interventions (prior to the addition of the RapidSMS technology) for their degree of technical complexity, the extent to which they were able to scale‐ up, and, as relevant, the most significant constraints to scale up and 2. Assess the interventions (following the addition of the RapidSMS technology) for their degree of technical complexity, making note of whether SMS technology had addressed any key constraints to scale‐up identified earlier, and whether any additional constraints had been introduced.

RESULTS

See the boxes at the end of the text for overviews of the case study projects, before and after the integration of RapidSMS, box 7.1 and 7.2.

107 Chapter 7

Case Study 1: Early Infant Diagnosis

Basic program intervention (without SMS) As illustrated in Table 7.2, Column 1 and based on application of four domains of the framework, the main sources of technical complexity for the basic EID program in both Malawi and Zambia are linked to the lack of human resources and poor communication and transportation infrastructures. Although Dried Blood Spot (DBS) tests themselves are stable and safe to transport, complexities are linked to supplies and maintenance of laboratories, diagnostic equipment and computers. EID requires outreach services to ensure follow‐up of HIV positive mothers and their infants, a network of laboratories with regular quality control for DBS analysis, trained nurses, adequate care management and couriers to transport DBS samples and test results. In Malawi and Zambia a critical shortage of skilled health workers has resulted in overstressed staff and inadequate supervision. At the level of providing guidance, EID requires a strong national HIV control strategy and coordinating actions across national and local governments, different tiers of the health sector, as well as NGOs and other actors. For example, in Malawi the frequent changes in HIV policies have restricted full implementation of these strategies. Finally, at community level, awareness and demand for PMTCT services need to be stimulated in order to identify HIV positive mothers and promote EID. Given the stigma surrounding HIV and AIDS, this can be difficult.

Intervention with SMS When the analytical framework was applied to the EID projects in Malawi and Zambia following introduction of RapidSMS technology, the reductions in technical complexity are primarily found in reducing communication and transportation needs. The main benefits of SMS technology are seen in its ability to simplify the tracking of DBS tests, potentially reducing transport needs, improving efficiency of communication of test results to health workers and their ability to remind mothers/caretakers to return to the clinic for test results, by contacting them via phone. However, applying the framework also suggests that this innovation can add complexities to the existing program. For example, despite the fact that RapidSMS can be used on any mobile device, unreliable network coverage and limited access to electricity (which influence the connectivity) have constrained the adaptation of the programme to other geographic areas. In addition securing privacy of data, when using personal mobile phones, is a potential concern. Complexities increase as the use of RapidSMS requires technical expertise. Moreover, the application requires national eHealth policies and government leadership for coordination with public sectors and implementing organizations to ensure sustainability. For example, in Malawi, delays in scalability of the project were linked to government ownership and prioritization.

108 Assessing scale up of mHealth innovations based on intervention complexity

Widespread use of mobile phones in both Zambia and Malawi, however, suggest that given adequate training, accessibility of phones should not increase program complexity.

Case Study 2: Nutrition Surveillance

Basic program intervention (without SMS) The application of the framework to nutrition surveillance (Table 7.3, Column 1) indicates that the main source of technical complexity includes the lack of human resources, supervision and training of health workers, which results in poor data quality.. In contrast to this, the basic product design (survey tool) is relatively simple, although the paper‐based system is reliant on couriers to reach national level. Other challenges are linked to the lack of capacity in data analysis at national level and utilizing the collected data. Similarly, there are challenges in raising community awareness regarding the importance of nutrition surveillance and ensuring that mothers and their children enrolled return on a monthly basis for monitoring.

Intervention with SMS When applying the framework to nutrition surveillance with SMS intervention, (Table 7.3, Column 2), the main reduction in complexity is found in simplifying constraints linked to communication and transportation of survey data, by reducing the delay in data transmission and improving data quality and analysis as well as bypassing the labour‐intensive paper based system. As with the EID case study, similar complexities relating to mobile phone connectivity are also identified. However, for nutrition surveillance in Malawi, a lack of available computers and internet, (required for data collection at districts and national level) is also a challenge. Although SMS simplifies the labour intensive paper‐based survey process (which was kept in place after adding the SMS component), capacity to maintain the software is lacking, thus limiting full utilization of the potentials offered by the program. Advantages for the use of SMS are that it limits errors which can occur when aggregating the data to national level using paper based systems, as it allows raw data to directly enter the central server at national government level. Errors made by health workers will, remain (e.g. incorrectly taking a child measurement, sample bias), unless the entered data is physically impossible and an automated SMS response prompts the health worker to correct the data entered. The use of SMS reduces delays in data transmission and manpower requirements for data entry and analysis. In regard to providing guidance, there is a need for mHealth policies and strong leadership. A key constraint prior to the addition of SMS was that survey data needs to be analysed at central level government ensuring timely action – a constraint which Rapid SMS does not alleviate.

109 Chapter 7

is

or

1

2 the

to 1 .

to of if 1

].

and The have

level

‐ or

health 24 1 indicates

[ collect due

from

2

in manage

airtime providers

to

phones

nature

human to evaluations available

managers

district green systems

the partners (Sharpey

1

2, the outages services

1

phones phone

phone evaluated

and

from

to

constrained.

widely

personal cellphone or

based

project

communication,

differ replacements servers

project

(e.g. Zambia,

system

use

and

consultants

Column and mobile mobile

mobile simplified

SMS

in highly

and national the may based In

unreliable

SMS

by addition

up and

are ‐ up in

personal 2). using

could

using

and both

workers K, of evidence localized

software documented locally menus

chargers) at scale

is SMS

which

scaling when and care for

recharging

not

but transportable caregivers

the experience

providers

local

, (Column

in solar

of Schafer, database there or

designers), of

Rapid

1 but

] issue

need

if

delays

health 23 project, easily SMS (& (e.g.

a

[

phone part an repairs, 1:

1 service

2011) effectiveness that

a development the bridge as

be

use, contacting

also

the

the challenges Rapid (Sharpey of to

is

to

electricity 1

by rapidsms

availability can

with mobile for to

chargers

however,

to ] of

contract coverage

experience suggests

with

technicians

safe

the

capacity

any

5,25 There

data to lead [ however,

needed delivery to 2 ,

]. can EID

are 2

on

regarding

data the

is

25 example

battery access

operators

[

communication, 1 internet, need and

addition

and network

speculated/potential, for

responsible a

categorized patient areas used

linked

phones RapidSMS capacity is 1) need is

and

is ].

phones of

be with platform limited support

16

health 2

[ Evidence

network quality

2 were

mobile or to

personal

equipment EID preliminary can

claim

for

there

services, mobile

K, the phones 2: therefore

phones

Mobile requires

(Column

technical

but +

and/ the geographic

+ + +

column

results mobile

of are if RapidSMS

EID

in

increased. Zambia. ] ]

2: communication

24 25

2011) Some [ [ the lack Constraints committed Confidentiality resources local However, insufficient Column RapidSMS their Same, Additional Same and Same manage Same Schafer, personal outreach Same for a

and was

].

and

equipment 41 claims

of

and

[

and

if for

3)

],

].

And,

] via the 42

41

EID lack

[ [

39

use

Malawi [ are stable, the

PCR

to complexity

up for in to

transport

complexity for health

health heat

3) supplies,

due behind

using

consistent

Malawi computers nurses 7.2)

control

2)

is

are follow

ART, HIV needed

transported where skilled

workers

and

management. needed EID, to rapidsms

to

of be

for is

red of

analysis, intervention

(DBS)

quality

evidence For

design, lost

analysis, resources,

can (Table health

of

evidence use

and supplies

Zambia DBS

tests tested DBS

relation

for

to assess

treatment

the shortage the for

regular tests

personal reviews. in

tracing

level

human

for

product are

to

spot

safe 2)

and for

and supply DBS

and with

reduced,

supervision. The

including other related Other

critical

basic

issues

and

of without

], a

in blood The

]

21 diagnosis 1) drug was

[

personnel is ].

41

key facility, [

needed

and 39 lack specimens

equipment

[

robust. framework analysis

diagnosis reviews)

1)

essential

dried

are

manner are

the

there

outs and Africa for

with regular ] the ‐

less

DBS are

constraints

includes

the

41

challenging. infant taken

: [

of of

complexity for laboratory various staff

contradicted

be and

includes stock

are

tests where

EID

: laboratory

Sahara

in

services ‐

applies

design services are

and

of

may program lab logistical resources, Early

are DBS drug

where

sub a

7.2

planning standardized

limited confidentiality care of

in EID

a this 1: tests

to

face face

maintenance and in

product

laboratories outreach reflected findings areas still Characteristics and

Table Analysis human overstressed precautions

DBS needed

Basic of

health use in and to

mainly

often

often Characteristics

1:

supplies for basic HIV

courier,

to study

personal

7.2 privacy or

level

‐ the facilities,

regard

network

a regular technology. Facilities And, Resulting mail first needed database Colomn Intervention For simple Regular Equipment Delivery For In health workers, facilities management Case Table

110 Assessing scale up of mHealth innovations based on intervention complexity

in

of

term still

is

ethical needs

literacy Schaefer,

long

2011)

too

and

] for be and

phone

(e.g. 16

and [

1

to communication 2

collection legal

transport Malawi. need ]

communication

not phone 16 and

[ mobile of

Health for

data 2

and partners

the

of

the caregiver for

reduce

regulate

communication,

markets and

use

to use the

use

lab,

transport

to

mentioned

reliability needed to

1

scalability its to

black

is

] need government sector in Ministries

16 and how as personal

action [

this 1 tests making

K,

trained (e.g.

potentially by such

public policy, providers

delays

know

speed

however, DBS ownership be

national ,

in 1

and of to

, by with

1 Schafer, level ‐

factors

] transparent already experience, eHealth collaborative

network

28 need widely

to

[

improved government 1

on

resulting

tracking

worker and

field management

‐ more due often

used

the national (Sharpey and

ownership m mobile

2

workers coordination are

from

eHealth

and health

users

and high results

SMS),

within

improve improved

care strong

is

efficiency by dependence national

is

strong

and

can has phones

3)

facility/

for and however,

them

sent regarding communication),

health

identified to

there tests SMS SMS

needs

commitment as stage

partnership Need Mobile The

but but DBS

+ + + +

among systems,

results efficacy

results

personal early

Zambia,

an However, In applications, Same about Same, M, rates complex (DBS test Same, Same Same Telecommunication) Same frameworks financial Same

is

HIV

diminished

and

management

not

formal of

local is 2) the

couriers treatment the and risk and national

surrounding

a setting and as stewardship

the and

communication implementation epidemiological receive community for;

specifically

markets 3)

and

results

capacity,

stigma

between

And, national EID, they

awareness

standard

test testing need

black the

a of

requires to supplies,

is

that and

restricted labs.

demand,

financing

sector,

infrastructure;

regulatory given 2)

due

between supervisors

which have from there regarding

drugs drugs,

and

community health

(demand)

counselling,

and ensuring

usage, and HIV

transport the policy

of

to

testing EID

government required efficiency for

in of challenging create

capacity, training, NGOs during in

caregiver

for HIV to legislative be

ease

assurance and

utilization

tiers in 1)

results

and 1)

action

can enroll change

functional

nurses

need

and licensing

to a

of needed providing labs, quality

stakeholders

providers, regulatory tests/

is

women

which

efficiency includes

] different includes are

and : need HW, :

42 and

DBS

[ requires

regulate Frequent

EID, there

private various collaborative the

awareness

program to

child

Supervision pregnant systems

and

between supply Capacity and

HIV ] with

]

mothers activities

usage,

diminished of 42

41

transport [ legislative [

their

regarding increase AIDS. for

drug to between

strategy,

monitoring. of

sector

Characteristics

to

national risk positive and

increase a

Malawi needed.

results surveillance health demand government, applicable Regular Government Regarding control Management Dependence Usage To Need HIV The mobilization of in quality HIV needed enrollment is needed treatment

111 Chapter 7

1:

the but or

a

For

SMS ]

and

the 24

Some [

closely the .

1

local

1 and/

with local

weeks, and to

the

in

Rapid

mobile by

inputs

lacking.

from regarding

Additional therefore two

prompts 2011) airtime

]. with manage

available delays 16 for

are

experts

[

systems

to manage

2 as enabled categorized

detected

to

services

and

data

is insufficient communication, quality

sometimes

Evidence

response widely

2 well

health based not speculated/potential, 1

cellphone

is to

is as

data and

Surveillance

SMS report was platform and

SMS

due

phones

differ 1 consultants

it public communication, managers

to personal unreliable

column

replacements claim

internet K,

providers in

may increased.

using and

the and messages servers

failed

outages and improving

rapidsms

Nutrition

if personal

chargers) project

RapidSMS

software

personal however

automated

of 2: when

phone claims

SMS

SMS menus transportable

K,

Schafer,

sites a

and an

experience , ‐

and

their

solar

partly

local

the system based York, designers) or

1) ‐

and

the issue sent

complexity

]

and of

recharging use easily

SMS mobile

23 (& (e.g. computers

addition [ phone

an of

New Schafer,

lack to ‐ ] (Sharpey

locally be

use, behind

and in

23 entered, a

1

one

[ using where with

localized impossible,

(Column

is to 1

for database repairs,

repeated

workers

can

mobile

consistent chargers data that ]

in and Access

red, the feedback) safe is 1

(Sharpey project technicians

any

faster care 5,25 need there

workers data

[

the evidence

fact bridge providers

a 2 for

are on and the

However, and

of physically

to

delays battery

the surveillance of is

and

much

internet. health Surveillance

(resulting

Malawi patient used UNICEF evidence

correct electricity

level

service

phones

operators

RapidSMS

need

In phones of

and be

by to to Malawi

needed the

The reduced, in

responsible happen delivery entered is

example, Nutrition can platform phones,

experience

confirmation

mobile and

to Nutrition coverage

equipment

2: communication Malawi.

of

phone phones contract Mobile reviews.

access

network

and

requires

was

For

worker

data

+ + + + +

can 3) in together.

for

robust. identified

use and

other and

less was reviews) personal Confidentiality analyses 2011). equipment mobile Same Column When quality limited network areas Same Same the RapidSMS health consultants receiving themselves RapidSMS work Same Same

in

rapidsms and

complexity

into transport

of

complexity for easy and supplies

various

3)

)

roles

to 2)

in use at of increase

limited ] illegibility of

numbers collectors

where

are and 23

7.3 to national

and contradicted

[ entered

the

to

ID

needed lack

still data supplies

e

a be data

needed

is are

lead

bl are is areas growth

reach of

to reflected intervention child

division

standardized, services linked are design,

unclear

to can Ta

for or without

is resources

regular (

the

health there

analysing are

findings needs

servers

for

and

in assess

errors

tools

it

indicates

incorrect and product the workers

to

human outreach system couriers

ance supervision if ] with reporting

and

Malawi bias,

ll 2)

level, or

evaluations ministries, and

phones In

23,43 basic

planning

[ green database

health maintain based equipment

children

level,

1) finding

based entered,

program entering forms ]

2,

‐ to

from and 23

sample

facility, [ the

of forms

servers.

for framework Survei

level mobile various

national values case

1)

paper

surveillance easy evaluated

(e.g. paper

addition, by

qualified

national

enrolled the the

the monitoring. includes: many or

In

Column

supply using

at of

includes

analysis active

delays. missing surveillance of national In database

up

includes

for owned computers

maintaining

data nutrition and to

management

measurements

generally

data applies by

evidence

reaches

however, is design, 2).

with regular of requires

accurate Nutrition

is

resources: for of cause

are for

are 7.3

data follow

entry are data

monitoring. 2: nutrition

documented

data wrong can

needed,

system entered y there

product analysing

or Characteristics: the system ensure

procure;

quality

which

human data

if not effectiveness d (Column Analysis Table

needed needed

Basic

to to

these is

and needed growth for Characteristics: ensuring level sheet.

and collected 1:

and/

basic once

Poor

Stu

INFSSS

based facilities

7.3 and

responsibilities the

excel for the store handwriting, regard

needed

Personal Paper level Column Intervention For Supplies Equipment Delivery As In needed, perform to monitoring, sending Finally, and capacity written national As errors. entering of an is Case Table

112 Assessing scale up of mHealth innovations based on intervention complexity

2 to key

a

term

of

of to

be

efficiency needs

long

to ethical

for

and

given 1 challenging

still for personal

the

and accuracy

K, be

by (PPP),

proves increase

forms

need

still effectiveness legal

analysis and

and and

can

the Schafer, ]

system

paper information ‐

increased

19 2011) data ,

[

1

basis of

the markets ]

regulate potentially

partnership

users

19 Health of [

to

1

increase action of can 1

(Sharpey

black

needed ]

analysis regular 2

some public

23

accuracy collection a to [ and need collection,

costs

transporting Min

2 for and

on to

due data

ownership data

SMS literate

(e.g. communication,

feedback

and private ownership

trends

policy,

providers

for

done full

access are

collaborative children

increased

related .

lower with level

on

23] differs and assess [

and

data or efficiency

being

their personal

]

,

taking

eHealth

phones 1

streamline 28 network costs

[ K, of aggregated

in

government

not 1 continues and

to

of

time time

and

‐ is better national ‐

phones

internet the air

time

m

to

‐ mobile

faced

real mobile

2 health

coordination helps and ; dependence

1 Schafer,

of system and eHealth

efficacy

‐ free real

effectiveness are calculation)

3)

the

and

mobile

analysis within

use of data

elimination

strong national 2011) and

based

use

Platform Such monitored

for for provide (Sharpey

regarding terms 1 computer can

negotiate

management

to except regarding diminished commitment challenges

in

(automatic paper

entering

to

systems use of Need Need partnership HW promoting

+ + + + +

to

RapidSMS ability risk

sufficiently before,

Malawi program

the in Currently manually communication, Skills (financial) The The reporting government In example Same As Same frameworks Same telecommunication) Same caregivers Same be Same programming. challenge

3)

be

management the

not

and ] level

is 2) as

23

the

[

must faced

enrolled

data

such

of food

are

demand quality

national

human

markets

43] regarding

capacity,

[ government,

2)

to and Malawi regarding

nutrition

and ownership

systems

Nutrition black

local benefits

children then

improve

usage, to

(as initiatives and To of

around

and

nutrition

regulatory

surveillance due

awareness on in

monitoring. National utilizing ownership

infrastructure

and

ease

government

surveillance

1) district and policies

national

fully to child ]

crises

challenge. to 44 on nutrition

[ to

efficiency

participants

and

(NSP) transport community of

linked legislative government

needed and local,

government

place identify based

is capacity: up

between

1)

is ‐

related

of the in

to ]

regarding are significant create

23 representative

from

[ be

a

and lack to

follow INFSSS

functional is

surveillance order

efficiency as

includes

required

these of

Programmes

: for : in regulatory

is

need should data

constraints security

a loss

nationally awareness infrastructure of and

require

in

is

(continued)

transported to

nutrition

systems, action food ]

(NNP)

Capacity and

analysis

be

of organizations 23 ]

diminished

challenge

[ due

23 result

there

Surveillance

quality [ and

and increase

for supplies

must Characteristics legislative

to

7.3

year)

society poor risk

usage

a existing

for

An importance enforced Nutrition Programmes challenges civil nutrition applicable Government As Management resources Collaborative Usage For Need The systems communications and Forms for Regular Table

113 Chapter 7

DISCUSSION

Interest in incorporating and improving technology within health systems in resource poor settings is increasing, and therefore it is essential that we first critically analyse existing systems and programmes in order to understand how and where such technology may make a meaningful contribution. If communication and flow of information are identified to be a weak link, for example, SMS or other forms of mHealth technology can have a positive effect on the system or programme. However, if after analysing the existing programme, the main weaknesses are not related to exchanging or improving the quality of information and data, the benefits of SMS technology, for example RapidSMS may be limited and other solutions may need to be prioritized. Further, the application of a new technology itself may create additional (perhaps even negative) effects that influence the ability of the program to be sustained or scaled up. In this paper, we applied an analytical framework to assess the technical complexity of three projects prior to adding SMS (two on EID and one on nutrition surveillance) and conclude that each contained varying levels of complexity that posed significant challenges to operating at scale. Within both EID and nutrition surveillance, the key challenges identified were linked to human resources and infrastructure requirements. For both interventions, the addition of RapidSMS proved to be effective in helping to overcome constraints relating to communication and transportation infrastructure. For EID, the use of SMS decreased the turnaround time for receipt of results by the facility. For nutrition surveillance, the use of RapidSMS replaced the labour intensive paper based process with a simplified one, and improved data quality, enabling more sophisticated analysis and decreased delays in data transmission. Nevertheless, some key challenges remained and new constraints emerged when adding RapidSMS, including some that have implications for sustainability and scale‐up of the projects. This analysis shows that in regard to implementing RapidSMS, an important constraint was the shortage of local technical staff to maintain the associated data and database in both countries. For the initial phase, Zambia and Malawi depended on external consultants to design and oversee the software programs which added additional management complexities. Another key constraint found in this analysis was the difficulty created when the government did not take full ownership of the system. This was a problem in Malawi due to competing government priorities; in Zambia government ownership only occurred when maintenance of the RapidSMS service was incorporated into the Ministry of Health‘s information technology infrastructure.16 Second, for both case studies there were difficulties implementing RapidSMS in some geographic areas, a phenomenon that has been found in other health‐related SMS

114 Assessing scale up of mHealth innovations based on intervention complexity programmes as well.17‐20 For example, an assessment in Ethiopia identified poor network coverage as the number one obstacle in relation to implementing RapidSMS.20 Third, in both case studies, key challenges were linked to a critical shortage of skilled health workers.21 Given that this was a key operating challenge prior to the introduction of RapidSMS, it would be critical to identify a strategy to address the needs of these (often) overstressed health workers, ensuring that any new technology introduced had a positive impact on their workload.22 The effectiveness of using technology itself to address this particular problem is still under debate. For example, the use of RapidSMS for nutrition surveillance suggests that it can have a positive impact by decreasing the workload,23 however preliminary data on health workers using SMS to trace mothers/ caregivers in other contexts suggests that this may not always be the case.24 A related point is that all health workers and their supervisors must be trained to ensure they are able to use the newly available data in a meaningful way so that it contributes to better programming and health outcomes. In addition to these constraints highlighted in our analysis, other constraints that influence the effectiveness and scalability of mHealth projects have been identified in the literature. These include price levels of relevant products,25‐26 the frequency in change of mobile phone numbers of individuals,25,1,27 data security25,5 and lack of regulations, for example to protect personal identifiable data.28 The literature shows that many of these challenges are not only linked to the use of RapidSMS, or the broader concept of SMS technology, in LMIC. Findings from a study conducted for the European Commission on the use of eHealth in high income countries (HIC) argued that key challenges were linked to political, legal and practical obstacles and that almost everywhere the use of ICT was ‘proven to be much more complex and time‐consuming than initially anticipated’.29 In addition, a study assessing the effects of using mobile phones for diabetes care in the USA concluded that the use of mobile phones can have great potential, however, it is a complex process and the technology itself ‘is not sufficient to make a difference’.30 It is therefore important to carefully plan the use of technology, establish technical working groups that involve all partners implementing eHealth initiatives and exchanging experiences and lessons learned between LMIC, as well as from HIC. At a local level, mHealth projects should be designed from the beginning with local partners, governments and users, as they can best identify challenges which may be influential for the success of the project.9,27‐28,31 Efficiencies could be gained by 'upgrading' the technological infrastructure. Even though these may not target the specific identified program challenges, they hold potential for creating efficiencies that will benefit multiple programs using a common mHealth surveillance or workflow management system. Improving the technological infrastructure is crucial to overcome the need for maintaining the phone and paper based parallel systems due to challenges mentioned earlier due to possible failures in electronic systems. Further, upgrading the infrastructure may encourage increased

115 Chapter 7 availability and use of diverse types of technology. This will in turn enable utilization of more complex technological tools such as smartphones or laptops that can support more complex programs, rather than basic mobile phones. However, this also has important implications for scalability, sustainability and cost which should be taken into consideration during planning.

Limitations The key limitation of this analysis is that there was a dearth of relevant project level data available. For this reason, project data were supplemented by key informant interviews. A second limitation is that the analytic framework used does not take economic considerations and cost effectiveness into consideration. Addressing costs‐ effectiveness is essential as some mHealth projects may be based on very expensive technology which may not be scalable due to that reason. And, as one of the key informants involved in the mHealth projects mentioned, funding sources (and limitations) can have an important impact: “Barring connectivity and other logistic issues, communication and information flow are real bottlenecks in a country like Malawi. However, community level health information systems as well as mHealth initiatives almost entirely are partner‐initiate driven and most come to an abrupt halt at the end of many a project's lifecycle.” Another limitation is that our analysis does not address contextual issues such as social and gender dynamics (e.g. who has the right to use a phone in a household), cultural issues (e.g. how is health care perceived), economic aspects (e.g. price of a phone and calls, repair) and usability (e.g. who can use the phone and for which purposes). Despite high penetration of mobile phones world‐wide, women, the elderly, and the poorest populations32 remain those most likely to not have access to technology, raising important considerations regarding equity.

Conclusion In conclusion, scaling up existing health programmes in which use of mobile phone applications such as RapidSMS have been piloted is not a simple issue. SMS technology has the potential to facilitate implementation of MNCH interventions by simplifying elements of their technical complexity. However, in deciding whether and how to introduce SMS technology to MNCH programmes, it is critical to first understand what the underlying implementation challenges are, what SMS technology can contribute, where it may introduce new challenges, and how these can be addressed within the local context. Conceptual frameworks that describe and analyse technical complexity can be useful for clarifying these factors. In this paper, we applied a framework by Gericke et al. (2005) to analyse the underlying system complexity of three MNCH programmes before assessing the contribution (positive and negative) of SMS technology.7 We argue that such analysis is an important first step in assessing the

116 Assessing scale up of mHealth innovations based on intervention complexity potential contribution of SMS (and eHealth more broadly), a second step would be to select mHealth strategies which are worthy of scale as part of a strength, weakness, opportunities and threats (SWOT) analysis or other strategies used by governments and implementing agencies to make these decisions. With such complexities identified and strategies in place to attempt to manage some complexities, robust monitoring and evaluating frameworks are needed to ensure that SMS investments effectively address identified challenges and equip key actors to handle existing and emerging complexity – capacities which are critical for sustaining and scaling up interventions in resource poor settings.33‐36 To date, rigorous evaluation of programmes using SMS technology has been limited, and further research in this area will be vital to ensure that potential benefits of mHealth innovations reduce health inequities and reach those most in need.3

117 Chapter 7

Box 7.1 Rapidsms and early infant diagnosis of HIV in Malawi and Zambia.

Without timely diagnosis and treatment, about one third of HIV positive children will die in their first year and almost 50% by their second year of life.37 Efforts to strengthen early infant diagnosis (EID) of HIV are integrated with prevention of mother to child transmission (PMTCT) programmes. In Malawi and Zambia, EID programmes begin by offering counselling and HIV testing to pregnant women through routine antenatal care services. Those who are HIV‐positive are enrolled into the PMTCT program. They are then counselled and instructed to return to the clinic to test their newborn 4‐6 weeks after birth.38 The advent of dried blood spot (DBS) tests has advanced the field of EID by allowing healthcare workers to more easily obtain blood samples from infants (via heel or finger prick) for diagnosis.39 For EID, the health worker sends the DBS tests to a laboratory by courier, where the specimens are tested; a process that can take up to 2 weeks. At laboratory, lab technicians process the tests and update the records in an excel database, which can also take 2 weeks. The laboratory then sends the paper‐based results back to the health facility, by courier, which can take up to 3 weeks.16,24 Subsequently, the mother or caregiver is traced and brought back to the clinic for counselling so that both mother and infant can be enrolled for appropriate treatment and care. At every step of the EID process the loss to follow‐up can be high. Incorporation of RapidSMS into the program

In 2010, a RapidSMS platform was added to EID programmes in both Malawi and Zambia with the aim of facilitating bi‐directional communication around DBS tests and results between health workers and laboratories using the ‘Result160’ application and to simplify follow‐up of mothers/ caregivers and their newborns through the ‘RemindMi’ application.24 Within this program, health workers and volunteers send and receive messages free of charge using their personal phones. After sending DBS samples to the laboratory, health workers send an SMS (text message), to a central database reporting the number of samples that have been sent to the laboratory. If the samples haven’t reached the laboratory within a specific timeframe, the central database system sends a SMS to the health worker, informing him or her that the samples have not yet arrived. Once the lab technician enters the test results into the computer at the laboratory, the system sends a SMS to all health workers in the health facility informing them that the results are ready. The health worker designated to collect the data then logs in with a pin code and receives the results on his or her phone. Finally, all health workers in the facility receive a SMS informing them that their colleague collected the results successfully.

Through RemindMi, the health worker sends a SMS to a designated community health volunteer or a traditional birth attendant to trace the caregiver and newborn, in order to refer them to the health facility. The central database system continues to send reminders for tracing caregivers, until the designated volunteer successfully refers the caregiver and stops the automatized reminders by sending a deactivation code to the central database.

118 Assessing scale up of mHealth innovations based on intervention complexity

Box 7.2 Rapidsms and nutrition and foor security surveillance (INFSS) in Malawi.

Chronic malnutrition and growth stunting affects close to 50% of the overall population of Malawi, and these high rates of malnutrition are compounded by persistent food shortages and high disease burdens, including HIV and AIDS. Of all children under five in Malawi, 46% suffer from moderate to severe stunting and 25% are moderately to severely underweight.40

After a famine in 2002, the Integrated Nutrition and Food Security Surveillance System (INFSSS) was set up to address chronic malnutrition in Malawi. INFSSS monitors the nutritional status for approximately 9100 children through five district growth monitoring clinics (GMCs) on a monthly basis for one year.23 The children that are monitored are randomly selected from those visiting GMCs; as a result the sample includes healthy, malnourished, and ill children. Health Surveillance Assistants (HSAs) collect and record malnutrition data from these children, and the GMCs sends the nutrition data to a local district office each month. The local district then forwards this information to the national level for data entry and analysis. By monitoring child malnutrition through children attending GMCs, the Government of Malawi is thus able to assess and subsequently respond to trends in malnutrition levels among those attending GMCs.

Incorporation of RapidSMS into the program

In 2009, a RapidSMS platform was added to INFSSS to streamline and improve the quality, speed, and accuracy of data collection.23 Using SMS, HSAs are trained to enter and send child nutrition data with mobile phones. Health workers receive immediate confirmation that their information was received. A central server analyses the data based on automated algorithms to screen for child malnutrition, with follow up directions provided to health workers if the data indicates child malnutrition. A website created by INFSSS provides the Malawian government and other stakeholder’s real‐time access to the data and its analysis, allowing timely analysis and follow‐up to changes in malnutrition trends. Data sent by the HSA is then directly analysed by the central server and feedback is sent back to the HSA.

Acknowledgement Many individuals reviewed drafts of this work and provided helpful feedback. We thank Christian Salazar, Mickey Chopra, Theresa Diaz, Khassoum Diallo, Kumanan Rasanathan, Ariel Higgins‐Steele, and Ahmet Afsar. The manuscript was also reviewed by colleagues from the UNICEF office in Zambia; we thank Nilda Lambo and Lastone Chitembo. From the office in Malawi we thank Luula Mariano. For the inputs regarding the use of SMS technology for both projects, we thank Erica Kochi, Merrick Schaefer, and Kieran Sharpey‐Schafer.

119 Chapter 7

REFERENCES

1. Asiimwe C, Gelvin D, Lee E, Ben Amor Y, Quinto E, Katureebe C, Sundaram L, Bell D, Berg M. Use of an Innovative, Affordable, and Open‐Source Short Message Service–Based Tool to Monitor Malaria in Remote Areas of Uganda. Am J Trop Med Hyg 2011;85:26‐33. 2. Zurovac D, Sudoi RK, Akhwale WS, Ndiritu M, Hamer DH, Rowe AK, Snow RW. The effect of mobile phone text‐message reminders on Kenyan health workers’ adherence to malaria treatment guidelines: a cluster randomized trial. Lancet 2011;378:795‐803. 3. Patil DA. Mobile for health (mHealth) in developing countries: Application of 4 Ps of social marketing. Journal of Health Informatics in Developing Countries 2011;5(2) 4. Heerden A, Tomlinson M, Swartz L. Point of care in your pocket: a research agenda for the field of mHealth. Bull World Health Organ 2012;90:393‐394. 5. Noordam AC, Kuepper BM, Stekelenburg J, Milen A. Improvement of maternal health services through the use of mobile phones. Trop Med Int Health 2011;16:622‐626. 6. Tomlinson M, Rotheram‐Borus MJ, Swartz L, Tsai AC. Scaling Up mHealth: Where Is the Evidence? PLoS Med 2013;10:e1001382. 7. Gericke CA, Kurowski C, Ranson MK, Mills A. Intervention complexity‐ a conceptual framework to inform priority‐setting in health. Bull World Health Organ 2005;83:285‐293. 8. Hanson K, Ranson MK, Oliveira‐Cruz V, Mills A. Expanding access to priority health interventions: a framework for understanding the constraints to scaling‐up. J Int Dev 2003;15:1‐14. 9. Mangham LJ, Hanson K. Scaling up in international health: what are the key issues? Health Policy Plan 2010;25:85‐96. 10. WHO. Nine steps for developing a scaling‐up strategy. Geneva, Switzerland: WHO ISBN 978 92 4 150031 9, 2010. 11. Yamey G. Scaling Up Global Health Interventions: A Proposed Framework for Success. PLoS Med 2011;8:e1001049. 12. Wu F, Khlangwiset P. Evaluating the Technical Feasibility of aflatoxin risk reduction strategies in Africa. Food Addit Contam Part A Chem Anal Control Expo Risk Assess 2010;27:658‐676. 13. Ross‐Degnan D, Backes‐Kozhimannil K, Payson A, Aupont O, LeCates R, Briggs J, Chalke J. Improving Community Use of Medicines in the Management of Child Illness: A Guide to Developing Interventions. Submitted to the U.S. Agency for International Development by the Rational Pharmaceutical Management Plus Program. Arlington, VA: Management Sciences for Health, 2008. 14. Snowdon W, Potter JL, Swinburn B, Schultz J, Lawrence M. Prioritizing Policy Interventions to improve diets? Will it work, can it happen, will it do harm? Health Promot Int 2010;25:123‐133. 15. Menon P, Frongillo EA, Pelletier DL, Stoltzfus RJ, Ahmed AM, Ahmed T. Assessment of epidemiologic, operational, and sociopolitical domains for mainstreaming nutrition. Food Nutr Bull 2011;32(2 Suppl):S105‐14. 16. Seidenberg P, Nicholson S, Schaefer M, Semrau K, Bweupe M, Masese N, Bonawitz R, Chitembo L, Goggin C, Thea DM. Early infant diagnosis of HIV infection in Zambia through mobile phone texting of blood results Bull World Health Organ 2012;90:348‐356. 17. Kallander K. Landscape analysis of mHealth approaches which can increase performance and retention of community based agents. InScale Innovations at Scale for Community Access and Lasting Effects. Kampala, Uganda: Malaria Consortium Resource Centre, 2010. 18. Buys P, Dasgupta S, Thomas TS. Determinants of a Digital Divide in Sub‐Sahara Africa: A Spatial Econometric analysis of Cell Phone Coverage. Elsevier Ltd. World Development 2009;37:1494‐1505. 19. Schackleton SJ. Rapid Assessment of Cell Phones for Development. UNICEF. Women’s Net, 2007. [http://www.unicef.org/southafrica/SAF_resources_cells4dev.pdf] 20. UNICEF, RapidSMS Ethiopia Assessment: Improved Nutrition and RUTF Monitoring. Unpublished, UNICEF, (no date). 21. Christopher JB, Le May A, Lewin S, Ross DA. Thirty years after Alma‐Ata: a systematic review of the impact of community health workers delivering curative interventions against malaria, pneumonia and diarrhoea on child mortality and morbidity in sub‐Saharan Africa. Hum Resour Health 2011;9:27. 22. Keeton C. Measuring the impact of e‐Health Bull World Health Organ 2012;90:326‐327.

120 Assessing scale up of mHealth innovations based on intervention complexity

23. Blaschke S, Bokenkamp K, Cosmaciuc R, Denby M, Hailu B, Short R. Using Mobile Phones to Improve Child Nutrition Surveillance in Malawi. UNICEF Malawi and UNICEF Innovations and SIPA, University School of International and Public Affairs, New York, 2009. 24. Gonzales M, Iiams‐Hauser C, Jeffers JW, Kastner K, Raué A, Vaisben M. Improving Maternal, Newborn, and Child Health in High HIV‐burden Areas Through Mobile Technology. UNICEF Malawi and UNICEF Innovations and SIPA, Columbia University School of International and Public Affairs, New York, 2010. 25. Mechael P, Batavia H, Kaonga N, Searle S, Kwan A, Goldberger A, Fu L, Ossman J. Barriers and Gaps Affecting mHealth in Low and Middle Income Countries: Policy White Paper. New York: Center for Global Health and Economic Development, Earth Institute, Colombia University, m‐Health Alliance; 2010. 26. UNCTAD. Information Economy Report 2010 ‐ ICTs, enterprises and proverty alleviation. New York and Geneva: United Nations. [http://www.unctad.org/en/docs/ier2010_en.pdf], 2010. 27. Coomes CM, Lewis MA, Uhrig JD, Furberg RD, Harris JL, Bann CM. Beyond reminders: a conceptual framework for using short message service to promote prevention and improve healthcare quality and clinical outcomes for people living with HIV. AIDS Care, 2011. 28. WHO. Atlas eHealth Country Profiles Global Observatory for eHealth series ‐ Volume 1. Geneva, Switzerland: WHO ISBN 978 92 4 156416 8. [http://whqlibdoc.who.int/publications/2011/ 9789241564168_eng.pdf], 2010. 29. Watson R. European Union leads way on e‐health but obstacles remain. BMJ 2010;341:c5195. 30. Katz R, Mesfin T, Barr K. Lessons from a community‐based mHealth diabetes self‐management program: “It's not just about the cell phone”, J Health Commun 2012;17:sup1:67‐72. 31. Levine D, McCright J, Dobkin L, Woodruff AJ, Klausner JD. SEXINFO: a sexual health text messaging service for San Francisco youth. Am J Public Health 2008;98:393‐395. 32. GSMA. Women and mobile: A global opportunity. A study on the mobile phone gender gap in low and middle‐income countries. GSMA Development Fund, Cherie Blair Foundation for Women and Vital Wave Consulting, 2010. 33. Paina L, Peters DH. Understanding patways for scaling up health services through the lens of complex adaptive systems. Health Policy Plan 2012;27:365‐373. 34. Subramanian S, Naimoli J, Matsubayashi T, Peters DH. Do we have the right models for scaling up health services to achieve the Millennium Development Goals? BMC Health Serv Res 2011;11:336. 35. Leon N, Schneider H, Daviaud E. Applying a framework for assessing the health system challenges to scaling up mHealth in South Africa. BMC Medical Informatics and Decision Making 2012;12:123 36. Mair FS, May C, O’Donnell C, Finch T, Sullivan F, Murray E. Factors that promote or inhibit the implementation of e‐health systems: an explanatory systematic review Bulletin of the World Health Organization 2012;90:357‐365. 37. Newell ML, Coovadia H, Cortina‐Borja M, Rollins N, Gaillard P, Dabis F.Mortality of Infected and Uninfected Infants Born to HIV‐infected Mothers in Africa: A pooled Analysis. Lancet 2004;364: 1236‐1243. 38. UNICEF. Scaling up Early Infant Diagnosis and Linkage to Care and Treatment. A Briefing Paper. [http://www.unicef.org/aids/files/EIDWorkingPaperJune02.pdf], 2009. 39. Ciaranello AL, Park J, Ramirez‐Avila L, Freedberg KA, Walensky RP, Leroy V. Early infant HIV‐1 diagnosis programs in resource‐limited settings: opportunities for improved outcomes and more cost‐effective interventions. BMC Med 2011;9:59. 40. UNICEF. The Situation of Women and Children Retrieved May 2012 from: [http://www.unicef.org/malawi/children], 2012. 41. UNGASS. Malawi HIV and AIDS Monitoring and Evaluation Report 2008‐2009, 2010. 42. Braun M, Kabue MM, McCollum ED, Ahmed S, Kim M, Aertker L, Chirwa M, Eliya M, Mofolo I, Hoffman I, Kazembe PN, van der Horst C, Kline MW, Hosseinipour MC. Inadequate coordination of maternal and infant HIV services detrimentally affects early infant diagnosis outcomes in Lilongwe, Malawi. J Acquir Immune Defic Syndr 2011;56:e122‐8. 43. Trowbridge FL, Wong FL, Byers TE, Serdula MK. Methodological issues in nutrition surveillance: the CDC experience. The Journal of Nutrition 1990;120:1512‐1518.

121 Chapter 7

44. Akhter N, Haselow N. Using data from a nationally representative nutrition surveillance system to assess trends and influence nutrition programs and policy. Journal of Field Action. Field Actions Science Reports, 4. 2010.

122 PART IV

Discussion

123

124 CHAPTER 8

General discussion

125 Chapter 8

126 General discussion

GENERAL DISCUSSION

To ensure equitable access to health care, we need to know more about a) the key determinants in the delays that prevent a child from receiving adequate care, and b) how these delays can be addressed. The exemplary case studies in this thesis are based on children with symptoms of pneumonia (of which most occur in sub‐Saharan Africa1), and mHealth (of which sub‐Saharan Africa is the fastest growing mobile region2). This discussion aims to address these issues based on the findings from the individual studies in this thesis, in order to optimize programmatic coverage focussed on delays in care seeking, possibly facilitated by mHealth, with its successes and challenges.

To better understand the key determinants which delay care seeking, we first need to know which child with symptoms of pneumonia is currently delayed in accessing care Sub‐Saharan Africa has the highest under‐five mortality rates, including those due to pneumonia, and the lowest levels of care seeking.1,3‐4 It is estimated that only 2 out of every 5 children with pneumonia specific symptoms are taken to an appropriate care provider,4 hence knowing which child is delayed in accessing care, is crucial. The use of large national household surveys is specifically useful in resource limited settings for several reasons. This thesis shows that the USAID‐supported Demographic and Health Surveys (DHS) and the UNICEF‐supported Multiple Indicator Cluster Surveys (MICS), are helpful in identifying marginalized children. For all countries, these surveys are sufficient to produce reliable estimates at national, urban‐rural and regional levels.5 Depending on the sample, additional sub‐national or local level surveys are required to help identify poor and marginalized communities at (sub‐) district level; e.g. those living in urban slums, religious and ethnic minorities, and/ or those living in conflict zones. Such analyses are essential to improve coverage of effective health interventions, as national averages often conceal broad disparities.6 With both DHS and MICS conducted on a routine basis (every 3‐5 years), these surveys can also be used to monitor if programs have been successful in decreasing inequities in coverage, or not. Another reason to better use these existing datasets in resource limited settings is because routine health management information systems5 and other program level data7 are often of poor quality, lack periodicity to track changes and trends, or are simply not available. Finally, while these surveys consist of various interviews, including women’s, men’s and children’s questionnaires; Chapter 2 shows that merging these datasets enabled us to conduct analyses on both caregivers’ characteristics as well as child health outcomes. Based on both DHS and MICS, the two chapters show that, even between neighbouring countries in sub‐Saharan Africa, care seeking patterns vary widely. In Uganda almost 80% of the children with symptoms of pneumonia are taken to an appropriate care provider, whereas this is less than 30% in Ethiopia and Chad. We also found variations

127 Chapter 8 within the countries, for example between urban‐rural residence, geographical locations, religious groups, caregivers’ levels of education, number of children ever born, the child’s age and sex, and others. In spite of these differences, the main similarity is that a child belonging to the poorest household is the least likely to obtain adequate care, regardless of where he or she lives. Still, knowing where these children live is crucial to leverage resources and to ensure that they are not lagging behind when attempting to improve coverage of health interventions overall.

After having identified the child which is delayed in accessing care, we need to know why this child with symptoms of pneumonia is not receiving antibiotics (or other required treatment) in time To untangle delays which ultimately lead to high pneumonia mortality rates, the ‘three delays’ model by Thaddeus and Maine8 is useful. The first phase is the decision to seek care on the part of the individual, the family or both (deciding to seek care). The second phase deals with reaching an adequate health care facility (reaching adequate care), while the third phase presents if adequate health care is received at the health facility (receiving adequate care). While the model was initially designed to recognise the barriers women face when trying to reach adequate care in time, it is not the first time it is applied more broadly.9‐10 Based on the model, Chapter 3 shows that for Ethiopia and Chad the main causes of delays occur at household level, while in Uganda delays are most likely due to quality of care being sub‐standard. Despite these variations, causes of delays often affect more than one phase of delay; e.g. wealth is linked to whether care is sought or not, if care can be reached, but also from which quality.11‐12 Therefore, researchers should not aim to identify one specific phase in which most delays occur, but rather unravel the complete chain of challenges associated with care seeking; starting at home, the delay‐model serving as a way to structure the inventory of the chain. For children with symptoms of pneumonia, the first phase of delay (deciding to seek care) was the most complex to measure, as the decision of caregivers to seek appropriate care or not is influenced by a range of factors such as their ability to recognise symptoms, socio‐economic aspects, beliefs, and the perceived quality of care provided by health workers.8 Based on the model of three delays, Chapter 2 shows that, while pneumonia is the main cause of childhood mortality; caregivers in high mortality settings are often unaware of its symptoms. Chapters 2 and 3 also reveal large differences in the reported prevalence of pneumonia (this ranged from 1.9% in Burkina Faso to 14.8% in Uganda). As causes of child death,13 including those due to pneumonia,14 differ substantially from one country to another, it is not possible to reveal to which extent the range in reported cases is due to caregivers’ (in‐) ability to recognize symptoms. One of the ways to further assess the accuracy in which caregivers recognize symptoms is by interviewing caregivers at hospital level, where children can be clinically classified as having pneumonia or not. Yet, again, this

128 General discussion approach would miss those caregivers’ perspectives of children who were never brought to receive care. Hence, the main challenges of assessing this within the formalized health structure is that it will exclude the poor and marginalized child, as they are more likely to seek care elsewhere, e.g. from the informal sector (e.g. private pharmacies) or at home.15 A way to include these children could be done by conducting verbal autopsies at community level.16

To develop effective programs for multiple countries, we need to identify common challenges across these high pneumonia mortality settings, as revealed by multi‐country reviews Chapter 4 illustrates the most concerning cause of the third phase of delay (receiving adequate treatment), namely; a health worker is too often not able to correctly distinguish if a child has pneumonia or not. With no other cross‐country reviews on the ability (or inability) of community health workers to correctly count and classify respiratory rates in children with a cough or difficulty breathing being published before, our findings illustrate the challenges they face; especially those with limited training, literacy and numeracy. Failure to properly classify a sick child is especially problematic; also as it influences caregivers’ decision to seek care upon a next illness episode. These challenges reveal an urgent need to design, test and scale‐up potential pneumonia diagnostic devices. While counting the respiratory rate is still the WHO/UNICEF criteria for classifying pneumonia in resource limited settings, more emphasis is currently put on the need to develop point‐of‐care diagnostic test which can differentiate between viral and bacterial pneumonia.17‐18 One of the reasons behind the need for more advanced diagnostic devices is the changes in the epidemiology of pneumonia; with an increase in coverage of the pneumococcal conjugate vaccine (PCV),19 amongst other reasons, it is expected that the proportion of pneumonia cases attributed to viral infections will increase.20

With the fast uptake of mobile phones, programmers need to know why mobile technology based interventions succeed or fail; in order to rank such interventions in their effectivity to decrease delays in accessing care Chapter 5 illustrates that mobile phones have great potential to overcome fundamental communication challenges, shortening delays in deciding to seek care, reaching and receiving health care. The expectations are especially high in sub‐Saharan Africa; where many communities live in isolation and the uptake of mobile phones has been surprisingly fast. Nevertheless, various studies have underlined important challenges associated with the use of mobile phones to improve health outcomes (i.e. mHealth), this as most of the mHealth initiatives piloted in sub‐Saharan Africa fail to go to scale.21 Simultaneously,

129 Chapter 8 little is known on what’s working, what’s not and why not.22 Aiming to build on this evidence gap, Chapter 6 illustrates the effectiveness of a hotline and messaging service based on the “treatment effect on the treated” (TOT) model. The TOT model was used as the uptake of the initiative was low; only 18% of the targeted population actually used the toll‐free hotline service. The use of the SMS messaging service was even lower. Low uptake of mHealth initiatives has been reported elsewhere.23 Having a phone or not (despite the fact that caregivers could borrow a phone), is likely to be associated with the low uptake of the initiative; only 32% of the caregivers in the intervention area had a phone. The GSMA report on gaps in mobile phone ownership24 found that the most important reason for not owning a phone is related to its costs. The report also reveals that women are less likely to own a phone, as compared to men. In sub‐Saharan Africa the gender discrepancy is 13%; of the three countries included in the GSMA report this difference was the largest in Niger (45%), followed by DRC (33%) and Kenya (7%). In Kenya the gender gap widened to 16% amongst the poorest households.24 Hence, programmers should be careful when designing projects that use these devices as a strategy to overcome delays in accessing health care, especially since women are the caregivers primarily responsible for taking their child for care. With access to phones, network coverage, and other factors playing a significant role in the success of mHealth project; Chapter 7 illustrates that the intervention complexity is what determines scalability of the initiative. We found that primary implementation challenges before adding the SMS component were related to broad programmatic challenges (e.g. resources and governance), and these are not (easily) addressed by text messaging or faster communication systems. Nevertheless, the mHealth applications did address barriers related to effectiveness and timeliness as it enabled faster communication of results amongst health workers and data collection. In other words, the addition of SMS is not the solution, but part of a strategy to overcome communication barriers. Translating this to care seeking for pneumonia; ultimately the sick child still requires antibiotics, therefore the use of mobile phones can only help overcome part of the inefficiency and timeliness in reaching the required antibiotics. This can be achieved by using mobile phones to support caregivers to recognize the illness (delay in deciding to seek care), coordinate referral (delay in reaching care) and by facilitating consultation with more senior health staff (delay in receiving adequate care). Despite the expectations, the TOT ‐ as explained in Chapter 6 ‐ revealed no effect on care‐seeking for children with symptoms of pneumonia. While these findings may not be as we anticipated, they are in‐line with our findings presented in Chapter 2, where we found a lack of significant associations between knowledge and care seeking. Both studies highlight the complexities surrounding care seeking behaviour and they

130 General discussion illustrate a need for integrated health system and community based approaches; as they go beyond targeting individuals and the challenges they face.

Recommendations To conclude, this thesis aimed to answer the two main research questions which were posed in the introduction, namely: 1) How do the delays in care affect care seeking behaviour, and how can better knowledge of these delays lead to improved programming? And, 2) what is the potential of mobile technology to improve health outcomes by addressing these delays, and what are programmatic challenges in implementing mHealth strategies? The first step to increase coverage of health interventions is by identifying the child which is most likely to die due to pneumonia, as he or she fails to reach acceptable, affordable and appropriate health care in time. While national surveys are essential, local level research is required to identify why these children fail to reach adequate care. As the model of three delays is shown to be helpful to untangle determinants which affect care seeking, verbal autopsies at community level, based on this framework can help reveal causes of delays, starting with those which occur at household level. This thesis illustrates that care seeking behaviour is very much context specific, hence programmers (and donors) need to be careful when generalizing or duplicating effective strategies from one context to another. Despite the large differences, this thesis highlights that there are similarities across high pneumonia mortality settings. The main similarity is that caregivers and health workers fail to know and recognize pneumonia specific symptoms. Further research should focus on improving caregivers and health workers knowledge of pneumonia specific danger signs. Better knowledge on how to recognize pneumonia is crucial, especially in remote areas where the risk of dying due to pneumonia is the highest and health workers are the least equipped to classifying illnesses correctly. Such research should also focus on how knowledge, illness perception and other caregivers’ characteristics affect the accuracy in which caregivers are able to recognize pneumonia specific symptoms. In regard to pneumonia diagnosis, simple solutions ‐ such as counting beads ‐ need to be recognized and implemented more broadly, while waiting for other diagnostic aids which are preferably also low‐tech and more robust. To overcome challenges in timely accessing care, the focus should be on children under the age of two, as the incidence is higher amongst this age group. Finally, with timely care seeking being so complex, and child mortality due to pneumonia being so high, research needs to focus on preventing a child from getting ill in the first place, for example by increasing vaccine coverage, breastfeeding practices and improved nutrition. To finish, while the fast uptake of mobile phones has potential to shorten delays in seeking and receiving health care; programmers, donors and policy need to critically

131 Chapter 8 assess when and how such intervention is most likely to be successful. And, while further research should be robust and well documented aiming to build on the evidence‐base of mHealth initiatives, resources simultaneously need to be leveraged to test and scale‐up other interventions which are not (necessarily) technology driven. Especially equity focussed programs should be careful in technology driven interventions, since, despite the impressive uptake of these devices in sub‐Saharan Africa, (high) technology will stay out of reach for the poorest and most marginalized for still some time.

132 General discussion

REFERENCES

1. United Nations Children’s Fund (UNICEF). Levels & trends in child mortality. Report 2015. New York: UNICEF. 2. GSMA. Mobile for development. Available: http://www.gsma.com/mobilefordevelopment/ Accessed 2016, 2015. 3. UNICEF. Committing to child survival: A Promise Renewed. Progress report 2015. New York: UNICEF. 4. United Nations Children's Fund (UNICEF) global databases 2015. Available: http://data.unicef.org/child‐ health/pneumonia.html. Accessed 2016, Jan 12. 5. Hancioglu A, Arnold F. Measuring coverage in MNCH: tracking progress in health for women and children using DHS and MICS household surveys. PLoS Med 2013;10:e1001391. 6. UNICEF (2010) Narrowing the gaps to meet the goals. New York: UNICEF. 7. Noordam AC, George A, Sharkey AB, Jafarli A, Bakshi SS, Kim JC. Assessing scale up of mHealth innovations based on intervention complexity: Two case studies of child health programmes in Malawi and Zambia. Journal of Health Communication: International Perspectives 2014;20:1‐11 8. Thaddeus S, Maine D. Too far to walk: Maternal mortality in context. Soc Sci Med 1994;38:1091‐1110 9. Mbaruku G, van Roosmalen J, Kimondo I, Bilango F, Bergstrom S. Perinatal audit using the 3‐delays model in western Tanzania. Int J Gynaecol Obstet 2009;106:85‐88. 10. Waiswa P, Kallander K, Peterson S, Tomson G, Pariyo GW. Using the three delays model to understand why newborn babies die in eastern Uganda. Tropical Medicine and International Health volume 2010;15:964–972 11. Kahabuka C, Kvåle G, Hinderaker SG. Care‐seeking and management of common childhood illnesses in Tanzania—Results from the 2010 Demographic and Health Survey. PLoS One 2013;8:e58789. 12. Onwujekwe O, Hanson K, Uzochukwu B. Do poor people use poor quality providers? Evidence from the treatment of presumptive malaria in Nigeria. Tropical Medicine & International Health 2011;16:1087‐ 1098. 13. Black RE, Morris SS, and Bryce J. Where and why are 10 million children dying every year? The Lancet 2003;361:2226–2234. 14. Rudan I, Boschi‐Pinto C, Biloglav Z, Mulhollandd K, Campbelle H. Epidemiology and etiology of childhood pneumonia. Bulletin of the World Health Organization 2008;86:408–416 15. Kerber KJ, de Graft‐Johnson JE, Bhutta ZA, Okong P, Starrs A, et al. Continuum of care for maternal, newborn, and child health: from slogan to service delivery. Lancet 2007;370:1358–1369 16. Serina P, Riley I, Stewart A, Flaxman AD, Lozano R. A shortened verbal autopsy instrument for use in routine mortality surveillance systems BMC Medicine 2015;13:302. 17. Rambaud‐Althaus C, Althaus F, Genton B, D'Acremont V. Clinical features for diagnosis of pneumonia in children younger than 5 years: a systematic review and meta‐analysis. Lancet Infectious Diseases 2015;15:439‐50. 18. Qazi S, Were W. Improving diagnosis of childhood pneumonia. Lancet Infectious Diseases 2015;15: 372‐373. 19. Global Alliance for Vaccines and Immunizations (GAVI). Saving children’s lives and protecting people’s health by increasing access to immunization in poor countries. The 2014 annual progress report. Available at: http://www.gavi.org/progress‐report, 2015. 20. Usuf E, Bottomley C, Adegbola RA, Hall A. Pneumococcal carriage in sub‐Saharan Africa—A systematic review. PLoS One 2014;9:e85001. 21. Leon, N., Schneider, H. and Daviaud, E. Applying a framework for assessing the health system challenges to scaling up mHealth in South Africa. BMC Medical Informatics and Decision Making 2012;12:123 22. Tomlinson M, Rotheram‐Borus MJ, Swartz L, Tsai AC. Scaling up mHealth: Where is the evidence? PLoS Medicine 2013;10:e1001382.

133 Chapter 8

23. Chib A, Wilkin H, Hoefman B. Vulnerabilities in mHealth implementation: a Ugandan HIV/AIDS SMS Campaign. Global Health Promotion 2011;20(Suppl):26‐32. 24. The GSMA Connected Women Global Development Alliance. Bridging the gender gap: Mobile access and usage in low‐ and middle‐income countries. Available from: gsma.com/gender‐gap‐2015 Accessed 15 February 2016, 2015.

134

Summary

135 Chapter 8

136 Summary / Samenvatting

SUMMARY

Despite an increase in coverage of effective interventions to improve child health outcomes, millions of children still die before their fifth birthday each year. Children living in sub‐Saharan Africa are most likely to die, mainly due to infectious diseases ‐ of which most are attributed to pneumonia. With health outcomes most adversely affected by a delay in treatment, as described in Chapter 1, the aim of this thesis is to better understand what needs to be done to ensure timely access to adequate care for children with symptoms of pneumonia. Firstly, by examining the causes of delays based on three phases: the delay in deciding to seek care (phase 1), and the delay in reaching (phase 2) and receiving (phase 3) care. Secondly, this thesis examines what the potential is of mHealth strategies to overcome these delays.

The three phases of delay in care While there are a lot of factors which influence if a caregiver seeks care, we only examined the association between knowledge and care seeking. While pneumonia is the main cause of childhood mortality, Chapter 2 shows that only 30% of the caregivers living in sub‐Saharan African are aware of its symptoms (i.e. fast or difficulty in breathing). This chapter also shows that, the caregivers who are aware of these symptoms are not necessarily more likely to seek care. This illustrates the complexities surrounding care seeking, as it is interlinked to various aspects (including ability to recognize symptoms, empowerment, distance, etc.) and for that reason, knowledge of symptoms alone is insufficient to address the delay in deciding to seek care (phase 1). Linked to this, Chapter 3 illustrates that care seeking behaviour varies widely across high pneumonia mortality countries. For example, in Tanzania 85% of the children with symptoms of pneumonia are taken to a care provider, whereas this is less than 30% in Ethiopia. Despite the majority attending to primary health care facilities, many caregivers visit ‘non‐appropriate’ (or unauthorized) services, such as private pharmacies and traditional practitioners ‐ especially in Democratic Republic of Congo (DRC) and Nigeria. This chapter also illustrates that ‐ with an exception for DRC and Uganda ‐ caregivers from poor households are more often delayed in reaching care (phase 2), as compared to those from wealthier households. Our analyses illustrate that care seeking is lower in rural settings, which may (partly) be explained by lower education and income levels of those living in these areas. Finally, to ensure that children with symptoms of pneumonia receive adequate care (phase 3), especially in rural areas, governments ‐ in collaboration with partners ‐ train community health workers (CHWs) to assess, classify and treat these children. However, Chapter 4 shows that these health workers are often challenged in counting and determining how the respiratory rate relates to the age‐specific cut‐off points. More specifically, Save the Children found that the CHWs had limited numeracy, and were not able to count beyond 10. For these reasons, various organizations tested the utility of counting beads to help overcome the

137 Chapter 8 challenges these health workers face while determining if a child has fast breathing or not. We found that beads enable and improve the assessment and classification of breathing rates for illiterate CHWs. This study illustrates that there is an urgent need to better equip (community) health workers, especially those working in remote settings with high childhood mortality rates.

A potential solution to decrease delays To decrease delays in accessing care, Chapter 5 of this thesis shows that mobile phones, as a way to strengthen communication systems, have a lot of potential. The literature review illustrates that mobile phones are mainly used to reach and receive care in a timelier manner, this by coordinating referrals and supervision. While access to communication is one of the essential components to improve health services, fundamental challenges associated with care remain, e.g., gender discrepancies, organizational hierarchies, and remoteness. Therefore, while it is evident that connecting health workers within the health system is crucial, robust evidence on constraints and the impact of mHealth projects is limited. To build on the evidence base, Chapter 6 evaluates the impact of a toll‐free hotline and mobile messaging service on care seeking behaviour. Due to low uptake of the services, the impact could only be measured by assessing the effect of the service on those who actually used it, rather than comparing the differences between intervention and control area. The assessment shows an increase in facility‐based care for maternal health, yet an overall decrease of facility‐based care for children. For children with symptoms of pneumonia, there was no observed difference. One of the possible explanations could be that – after consulting with a health worker by phone – caregivers were able to treat the child at home and did not require a visit to a facility. Finally, as the interest in Short Message Services (or SMS) is increasing, we assessed the potential benefits and limitation of this specific technology in improving three projects. The assessment is based on a conceptual framework which assesses the technical complexities of programs. We assessed the complexities of the programs before and after adding the SMS component. Chapter 7 shows that despite the potential of mHealth projects, the primary implementation challenges for the existing programs can be thus complex that mHealth – or more specifically the use of Short Message Services (SMS) – may not necessarily be the best solution. In addition, there is a challenge that the technology itself may introduce new layers of complexities. Finally, the discussion presented in Chapter 8 aims to address the delays and potential of mHealth, based on the individual studies presented in this thesis. The discussion focusses on the importance of utilizing existing data, as well as the context specific challenges in accessing adequate care. Lastly, while communication means are needed to overcome delays, it highlights the need to assess how adding a technological layer is going to affect the intervention complexity of the existing program.

138

List of publications

139 Chapter 9

140 List of publications

LIST OF PUBLICATIONS

Part of dissertation Noordam AC, Kuepper BM, Stekelenburg J, and Milen A. Improvement of maternal health services through the use of mobile phones. Trop Med Int Health 2011;16:622–626. Noordam AC, George A, Sharkey AB, Jafarli A, Bakshi SS, and Kim JC. Assessing scale up of mHealth innovations based on intervention complexity: Two case studies of child health programmes in Malawi and Zambia. J Health Commun 2015;0: 1–11. Noordam AC, Barberá Laínez Y, Sadruddin S, van Heck PM, Chono AO, Acaye GL, Lara V, Nanyonjo A, Ocan C, and Kallander K. The use of counting beads to improve the classification of fast breathing in low‐resource settings: a multi‐country review. Health Policy and Planning 2015;30:696–704. Higgins‐Steele A, Noordam AC, Crawford J, and Fotso JC. Improving care‐seeking for facility‐based health services in a rural, resource‐limited setting: Effects and potential of an mHealth project. African Population Studies 2015;29:1634‐1654. Noordam AC, Carvajal‐Velez L, Sharkey AB, Young M, and Cals JWL. Care Seeking Behaviour for Children with Suspected Pneumonia in Countries in Sub‐Saharan Africa with High Pneumonia Mortality. PLoS One 2015;10:e0117919. Noordam AC, Sharkey AB, Hinssen P, Dinant GJ, Cals JWL. Associations between Knowledge and Care Seeking Behaviour for Children with symptoms of Pneumonia in six sub‐Saharan African Countries. Submitted.

Additional writing activities Palmer AC, Diaz T, Noordam AC. and Dalmiya, N. Evolution of the child health day strategy for the integrated delivery of child health and nutrition services. Food and Nutrition Bulletin 2013;34:412‐419. Fotso JC, Robinson AL, Noordam AC, and Crawford J. Fostering the use of quasi‐ experimental designs for evaluating public health interventions: insights from an mHealth project in Malawi. African Population Studies 2015;29:1597‐1615.

141 Chapter 9

142