J Manage Vol. 32, No. 2, pp. 150–171 Copyright c 2009 Wolters Kluwer | Lippincott Williams & Wilkins The Impact of Primary Healthcare on in Low- and Middle-Income Countries

James Macinko, PhD; Barbara Starfield, MD, MPH; Temitope Erinosho, PhD

Abstract: This article assesses 36 peer-reviewed studies of the impact of primary healthcare (PHC) on health outcomes in low- and middle-income countries. Studies were abstracted and assessed according to where they took place, the research design used, target population, measures, and overall conclusions. Results indicate that the bulk of evidence for PHC effective- ness is focused on infant and child health, but there is also evidence of the positive role PHC has on population health over time. Although the peer-reviewed literature is lacking in rigorous experimental studies, a small number of relatively well-designed observational studies and the con- sistency of findings generally support the contention that an integrated approach to primary care can improve health. A few large-scale experiences also help identify elements of good practice. The review concludes with several recommendations for future studies, including a focus on better con- ceptualizing and measuring PHC, further investigation into the advantages of comprehensive over selective PHC, need for experimental or quasi-experimental research designs that allow testing of the independent effect of primary care on outcomes over time, and a more detailed conceptual framework guiding overall evaluation design that places limits on the parameters under consider- ation and describes relationships among different levels and types of data likely to be collected in the evaluation process. Key words: developing country, evaluation, population health, primary healthcare

HE WORLD HEALTH ORGANIZATION dergone dramatic changes, including democ- T(WHO) formalized its commitment to pri- ratization, economic liberalization in an in- mary healthcare (PHC) in 1978, when it was creasingly globalized world, redefining the identified as central to the achievement of the role of the state, and reforming their health goal of “”and as a key instrument and social services systems. Health reforms, in for improving health throughout the world particular, have aimed at streamlining health- (WHO, 1978). care financing and decentralizing authority In the decades following Alma Ata, many for planning and implementation. There is low- and middle-income countries have un- increasing evidence that not all of these re- forms have strengthened PHC, nor have they uniformly contributed to improving health or equity in its distribution (Infante & de Author Affiliations: Department of Nutrition, Food Studies, and , New York University, Mata, 2000; Mackintosh, 2000; Varatharajan & New York (Drs Macinko and Erinosho); and Thankappan, 2004). Department of and Management, The In many high-income countries, various at- Johns Hopkins Medical Institutions, Baltimore, Maryland (Dr Starfield). Dr Erinosho is now with tributes of primary care have been shown Research Branch, National to exert a positive influence on health costs, Cancer Institute, Rockville, Maryland. appropriateness of care, and outcomes for Corresponding Author: James Macinko, PhD, Depart- most of the major health indicators (Bindman ment of Nutrition, Food Studies, and Public Health, New et al., 1996; Forrest & Starfield, 1996, 1998; York University, 35 W 4th St, 12th Floor, New York, NY 10012 ([email protected]). Starfield, 1998; Starfield et al., 2005a, 2005b). 150 Primary Healthcare in Low- and Middle-Income Countries 151

There is also evidence that countries charac- tion of PHC programs, systems, and services terized by a strong primary care orientation and to exclude articles that (1) did not explic- have better and more equitable health out- itly define the scope of the PHC intervention; comes than those systems that are oriented (2) evaluated only one component of selective toward specialty care (Macinko et al., 2003; primary care services (eg, immunization, oral Starfield, 1996; Starfield & Shi, 2002). Nev- rehydration ); or (3) did not include ertheless, there is considerable debate about data on changes in health outcomes attributed how effective PHC has been in improving to the PHC intervention. Overall, 36 key arti- population health in low- and middle-income cles were retrieved and abstracted. countries (Filmer et al., 2000; Lewis et al., The Appendix contains a synthesis of the 2004). main objectives, study designs, outcomes, The 30-year anniversary of the Alma Ata PHC measures, and results of the reviewed ar- meeting, the changing health challenges in ticles. We adopt the term “selective”to charac- the developing world, and the widespread terize interventions directed at selected indi- dissatisfaction with the status quo have gen- vidual health conditions (such as control of di- erated interest in a renewed and reinvigo- arrheal ) and “integrated” to describe rated approach to health systems develop- approaches that are more directed at health in ment based on PHC (“Margaret Chan puts general. In the presentation of results, we dis- primary ,” 2008; Pan American tinguish between PHC tasks or services (ie, di- Health Organization, 2005; WHO, 2008). To rected at a specific health problem) and PHC aid in this process, the present review as- functions (ie, directed at assuming the main sesses the peer-reviewed literature for evi- role of PHC within health systems, regardless dence of the effectiveness of previous PHC of the specific health problem). experiences with the goal of identifying lessons learned and providing suggestions for RESULTS strengthening the PHC evidence base. Figure 1 shows the distribution of new ar- METHODS ticles by year on the topic of PHC, which has increased each year and, after a relatively sta- The literature review was conducted ble period from 1995 to 2003, now averages by searching the US National Library of about 500 new articles per year. ’s PubMed database, the Cochrane Table 1 shows characteristics of the 36 Database of Systematic Reviews, and the In- abstracted studies. Geographically, they are ternet (via Google) for articles that contained fairly evenly distributed: slightly more than a the phrases “primary care”or “primary health third are from Africa, about a third are from care” along with the terms “evaluation” or Latin America and the Caribbean, a quarter “impact” in either the title or the abstract. from Asia, with the remaining representing Several journals that publish on healthcare multiple regions. In terms of study design, in the developing world were also hand most (45%) use a pre- and postintervention searched. All articles were then culled to cross-sectional design with controls or com- identify additional references. This process parison groups, about 14% use a case-control revealed more than 10 000 potential articles design, another 11% use multivariable longitu- as of July 2008. dinal analyses of ecological data, 1 study uses From the large potential pool of articles, an experimental design, and 1 uses a cohort we excluded all commentaries and non–peer- approach. All remaining studies employed reviewed works and all articles related to Eu- a variety of observational designs without ropean or other Organization for Economic controls. Cooperation and Development countries. Ab- In terms of outcomes, more than three- stracts and study designs were then reviewed quarters of the studies focused on infant or to identify articles that addressed the evalua- under-5 mortality, with the remainder dealing 152 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009

Figure 1. The number of new articles published in PubMed with “”in the abstract or title, by year (1975–2007).

Table 1. Characteristics of studies reviewed with maternal mortality, life expectancy, all- (N = 36) cause mortality, and cause-specific mortality in adults. All but 3 studies measured PHC ex- Number posure by residence in a geographic area in Domain of studies which PHC services were being delivered. The other 3 assessed individual use of spe- Geographic region cific PHC services. Nearly all studies point to a Latin America and the 12 positive impact of the PHC intervention stud- Caribbean Sub-Saharan ied: only 5 articles show no improvement at- Africa 16 tributable to PHC. Asia 8 Other (or multiple regions) 1 The magnitude of impact also varied con- Study design siderably. Reductions in infant and under-5 Experimental or 5 mortality attributed to PHC averaged more quasi-experimental than 40% and varied from 0 to as high as Prospective study with 1 71%, with interventions lasting from 2 years control group to more than 10 years. Repeated (pre/post) 16 cross-sectional design Studies on specific PHC tasks with control Several studies analyzed the association of Case-control 5 specific primary care tasks with health out- Repeated (pre/post) 5 comes. Moore et al. (2003) conducted a panel cross-sectional design study of 22 Latin American countries over without control Systematic literature review 1 the period from 1990 to 1998. The study Observation/qualitative 3 found that the most important contributor study/single cross-section to lower under-5 mortality was women’s lit- Main outcome studied eracy, followed by coverage and Infant or under-5 mortality 28 use of . A similar anal- Other (child) 1 ysis conducted by using Demographic and Other (adult) 7 Health Surveys from 5 East African countries estimated that nearly three-quarters of the Primary Healthcare in Low- and Middle-Income Countries 153 attributable risks for mortality in children IMCI programs in Brazil, Peru, Uganda, Egypt, younger than 1 year might be amenable to and Tanzania showed that although the ap- preventive services, including antenatal care, proach was in many cases more compre- immunizations, fertility regulation, and use of hensive and effective than individual vertical potable water (Brockerhoff & Derose, 1996). interventions, poor access, low levels of uti- In both studies, the extent to which these ser- lization, and structural weaknesses in health vices were part of an integrated PHC system systems limit its impact on population health or the result of an effort targeted only at major (Bryce et al., 2003; WHO, 2004) A recent re- causes of infant and is unclear. view suggested that a more comprehensive Dugbatey (1999) assessed the relationship approach to PHC and health systems devel- between a set of “Health for All” policies opment will be required for strategies such as (, nutrition, water and san- the IMCI program to flourish (Freedman et al., itation, and maternal/child health services) 2005). and health outcomes at the national level in Rohde et al. (2008) identify 13 coun- 4 African countries in the 1990s. Through tries that have implemented comprehen- a comparative case study design, the author sive PHC (Thailand, Turkey, Vietnam, Brazil, showed that PHC-sensitive conditions (such Sri Lanka, El Salvador, Tunisia, Dominican as ) were improved in the Republic, Iran, Kazakhstan, Turkmenistan, 2 countries with more comprehensive PHC China, and Cuba). Their analysis suggests that policies (Botswana and Zimbabwe), as op- these countries experienced important health posed to those with a less coherent set of PHC gains and that in comparison with coun- policies ( and Cote d’Ivoire), in spite of tries having more selective PHC approaches, the latter having higher gross national product health improvements—particularly for condi- per capita (Dugbatey, 1999). tions that require sustained and coordinated The Bellagio Child Survival Study Group care—seem to “depend on progression to concluded that nearly 10 million child deaths comprehensive primary health care with a re- worldwide could be averted by tasks or ser- liable system linking to functioning fa- vices including combined use of oral rehy- cilities”(Rohde et al., 2008, p. 958). dration therapy (Victora et al., 2000), immu- nization (England et al., 2001), micronutrient Studies of specific primary supplementation, promotion of exclusive care programs breast-feeding (Arifeen et al., 2001), and oth- The Navrongo experiment in Ghana was ers, all but one of which (neonatal intensive the only experimental study identified. In it, care) would be expected to be delivered by villages received 1 of 4 different interven- a PHC system. This estimate is supported by tions: professional community nurses; vol- another study (Berman, 2000) that estimated untary workers (CHWs); that about 62% of all disability adjusted life a combination of both; and nothing (con- years (lost in the adult and child population trol). In the nurse-only intervention areas, of developing countries) would be amenable under-5 mortality fell by 14% during 5 years to primary care services (termed “ambulatory of program implementation, compared with healthcare” by Berman). Access to primary that before the intervention period (Pence care appears to be particularly important in et al., 2007; Phillips et al., 2006). In the Africa; some authors suggest that up to 80% volunteer-only villages, under-5 mortality in- of child deaths occur at home, without the creased by 14%. The professional nurse inter- child having any contact with the health sys- vention added approximately $2 per capita to tem (Oluwole et al., 2000) the $6.80 per capita budgeted for PHC ser- Integrated management of childhood ill- vices. Note that the study used a “plausibil- ness (IMCI) reflects a horizontal primary care ity” rather than a “probability” design, mean- approach in the sense that it combines sev- ing that treatments were not truly randomly eral specific interventions. An evaluation of assigned. 154 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009

In the Gambia, Hill et al. (2000) compared • A large NGO-delivered PHC program PHC and non-PHC communities over a 15- (focused on maternal and child health year period. Primary care was assessed by the services) targeted about 340 000 poor presence of a community health nurse as a households in Bangladesh. Services were supervisor to village health workers (VHWs) delivered through trained family health and traditional birth attendants; presence of visitors and included regular household an expanded program of immunization and a visits, illustrating the importance of a fam- basic package of maternal and child health ser- ily focus in PHC. After 5 to 6 years, vices; and community participation in health- program areas experienced a 52% reduc- care provision, priority setting, and fund rais- tion in infant mortality and a 49% re- ing. The study found that although child mor- duction in under-5 mortality, larger de- tality declined in intervention and control creases than those experienced in control villages, the decline was generally steeper areas (Mercer et al., 2004). Another study in PHC villages (Hill et al., 2000). As a type in Matlab, Bangladesh, showed the im- of natural experiment, the authors report pact of a community-based PHC approach that once PHC services were stopped in the employing supervised and trained VHWs villages (because of lack of funds), the trend (with referrals to health centers staffed reversed and infant mortality increased to by healthcare professionals) on lowering levels higher than those in control (non-PHC) under-5 mortality from acute lower respi- villages. ratory tract by 32% in 2 years Velema et al. (1991) studied access and lon- (Fauveau et al., 1992). Prior studies in gitudinality of primary care services to a pop- Matlab had also documented reductions ulation of about 13 000 in Benin. In a matched in under-5 and maternal mortality due to case-control study, 2 factors were partic- different PHC interventions (Chen et al., ularly influential in predicting death risk: 1983; Fauveau et al., 1991). measles vaccination before their first birthday • In a cohort study conducted in (Odds ratio [OR] = 0.4) and regular con- Pondicherry, India, provision of a broad tact with VHWs (OR = 0.36). The authors range of PHC services, including home concluded that regular contact with VHWs, visits by PHC nurses in 12 villages (total which is consistent with the person-focused population of about 16 000), decreased care over time function of primary care, im- infant and child mortality by more than proves the likelihood of child survival. 65% (Dutt & Srinivasa, 1997). Another In , the activities of an integrated lo- smaller-scale Indian study assessed the cal , based on a PHC model effects of VHW provision of primary (the Hopitalˆ Albert Schweitzer or HAS), were and maternity care and health education associated with infant and under-5 mortality to pregnant women and grandmothers, that are about half of those in other areas resulting in reductions in neonatal mor- with similar income levels (Perry et al., 2007). tality by 62% and infant mortality by 71%, This was accomplished through a decades- as compared with preintervention levels long partnership with local communities. In (Bang et al., 1999). terms of resources, HAS had fewer physicians • In Liberia (Becker et al., 1993) and Zaire and fewer beds per capita than did (Chahnazarian et al., 1993; Taylor et al., the rest of Haiti but more nurses, CHWs, and 1993), a more selective PHC approach other outreach and support staff. The HAS was attributed to reductions in under-5 system costs about $19 per capita, includ- mortality by as much as 28% over a 5-year ing community development initiatives (Perry period, an improvement that was signifi- et al., 2006). cantly greater than that reported in com- Other studies of PHC programs using pre- parison areas. and postintervention measures and control or • A study using 2 waves of nationally repre- comparison groups include the following: sentative surveys in Indonesia found that, Primary Healthcare in Low- and Middle-Income Countries 155

while holding other village- and maternal- 1986). PHC improvements beginning in the level variables constant, the addition of a 1970s were estimated to have reduced in- maternity and a physician to a vil- fant mortality by between 40% and 75% lage was found to decrease the odds of (Rosero-Bixby, 1991). infant death (relative to an infant born be- In the 1990s, additional Costa Rican fore the clinic existed) by about 15% and reforms sought to improve PHC coverage 1.7%, respectively (Frankenberg, 1995). and efficiency. A quasi-experimental study • In , a comprehensive community- comparing 3 groups of districts on the basis based PHC program (delivered by paid of when they adopted PHC reforms found nurses and community volunteers with that, in addition to other improvements in some physician support) serving a pop- living standards, PHC reforms significantly ulation of about 15 000 successfully re- reduced mortality in both adults and children duced under-5 mortality by more than (Rosero-Bixby, 2004a). For every 5 additional 52% over a 5- to 6-year period, as com- years after PHC reforms, child mortality pared with control areas (Perry et al., declined by 13% and adult mortality by 1998, 2003). Costs for the program were 4%. The proportion of the population with estimated at about $10 per person. insufficient access to PHC services declined • In Pakistan, a case-control study of chil- by 15% in reformed districts compared with dren who had diarrhea or acute respira- only a 2% decline in districts that did not tory tract infections showed that the use undergo reforms. The reforms additionally im- of a traditional healer (as opposed to a proved equity in access by targeting the least trained VHW) raised the odds of a child’s privileged population first (Rosero-Bixby, death by a factor of 14 (OR = 14.5; 95% 2004b). confidence interval [CI] = 4.23–49.8), Brazil’s family health program (FHP) is now and frequent changing of providers (ie, perhaps the largest community-based PHC lack of continuity with a PHC provider) system in the world. In 2007, the program raised the odds of death 8 times (OR = 8; encompassed more than 27 000 community- 95% CI = 2.22–28.8) (D’Souza & Bryant, based teams responsible for providing care 1999). to about 85 million people (Brazilian Min- istry of Health, Department of Primary Care, 2006). The FHP is based on an explicit strat- Studies of countrywide PHC experiences egy to provide all core primary care functions, There have been only a few studies that di- including first-contact access for each new rectly test the hypothesis that health systems need, long-term person-focused care, compre- based on a strong PHC orientation (based hensive care for most health needs, coordi- on PHC principles) lead to better overall nated care when it must be sought elsewhere, indicators. and a focus on the family and the community. By 1985, Costa Rica’s life expectancy had These functions are achieved through the reached 74 years, and infant mortality de- program’s decentralized organization, elimi- clined from 60 per 1000 in 1970 to 19 per nation of copays for services, incentives to lo- 1000, levels comparable with those in more cal government for increasing access to the developed countries. Explanations for this program, and multidisciplinary teams com- rapid progress include the development of a posed of a physician and a nurse who deliver universal social security system and a mul- clinic-based care along with CHWs who make tidimensional approach to health improve- regular home visits and perform community- ment, which included expanding PHC ser- based health-promotion activities (Ministry of vices, investing in education and sanitation, Health of Brazil, 2003). Costs for the pro- and improving access to secondary and ter- gram (which includes access to pharmaceu- tiary healthcare services (Haines & Avery, ticals) are estimated at between $25 and $35 1982; Klijzing & Taylor, 1982; Rosero-Bixby, (Macinko et al., 2007). A panel data analysis 156 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009 of Brazilian states from 1990 to 2002 showed grams and expanded to a broader strategy that that an increase in FHP coverage by 10% was combined vertical programs with more com- associated with an average 4.6% decrease in prehensive PHC and human development ap- infant mortality, controlling for other health proaches, including legislation making access determinants, including water supply, living to maternal and child health services a citi- conditions, doctor and hospital supply, and zen’s right (Frenk et al., 2003). Reyes et al. women’s education (Macinko et al., 2006). (1997) also found that in Mexico, primary A follow-up study conducted at the microre- care characteristics (such as adequate refer- gional level for 1999–2004 showed that the ef- ral processes, continuity of care, being seen fect of FHP coverage was especially strong for by the same provider, and being attended in a conditions that are known to be sensitive to public facility) had an important, independent primary care (such as postneonatal mortality effect on reducing a child’s odds of dying. and deaths from diarrheal diseases) (Macinko Similarly, Gutierrez´ et al. (1999) point to the et al., 2007). Several studies also demon- importance of access to primary care (as mea- strated associations between CHWs and lower sured by nurse and ) as well infant mortality in specific Brazilian states as investments in public health (immunization (Emond et al., 2002; Svitone et al., 2000). and improved water and sanitation) and edu- A few other countrywide observations are cation as particularly important for reducing suggestive of the role of PHC in population infant mortality in Mexico. health improvements, although these studies In Thailand, decreases in under-5 infant do not explicitly quantify the contribution of mortality occurred after primary care reform, PHC to health improvements or explicitly test which included developing at least 1 pri- the impact of specific PHC interventions. mary care health center for every rural village Cuba’s universal PHC program uses family by 1990 and a government medical welfare health physicians and nurses, who provide scheme started in 1993. In the correspond- universal, comprehensive, integrated, and in- ing decade, under-5 mortality declined by 44% tersectorial care to geographically defined ar- in the poorest population quintile, 41% in the eas with a focus on families (Evans et al., next poorest quintile, 22% in the third, 23% in 2008; Waitzkin, 1997). Changes in PHC ac- the fourth, and 13% in the wealthiest quintile cess, organization, and delivery over the past (Vapattanawong et al., 2007) 40 years correspond to about a 40% decline In Indonesia, a 20% reduction in infant mor- in infant mortality over the same period, even tality during the early 1990s has been at- while other indicators such as gross national tributed to improvements in PHC (Simms & product per capita have not substantially in- Rowson, 2003). Some evidence for this attri- creased (Riveron´ Corteguera, 2000). Invest- bution comes from the observation that in ments in prevention integrated into PHC may the later 1990s, once primary care spend- also have contributed to the control of car- ing declined substantially (and hospital spend- diovascular diseases, resulting in lower-than- ing increased by almost 25%), infant mortal- expected mortality and fewer avoidable hos- ity actually increased by 14% in almost every pitalizations for these and related conditions province of the country (Simms & Rowson, (Spiegel & Yassi, 2004). Lessons learned from 2003). the Cuban experience suggest the potential Finally, the 2008 on benefits of organizing an entire health sys- PHC presents numerous case studies of PHC tem around the PHC approach (Franco et al., experiences. Although it does not contain a 2007). systematic review of the evidence on the ben- In Mexico, child mortality declined from 64 efits of PHC, it reviewed the evidence for the per 1000 live births in 1980 to 23 per 1000 benefits of PHC components and concluded in 2006 (Sepulveda et al., 2006). These re- that there is an overwhelming justification for ductions were consequent to a strategy that a focus on developing and strengthening PHC began with a number of -specific pro- in all countries (WHO, 2008). Primary Healthcare in Low- and Middle-Income Countries 157

Studies finding little or no impact of posed to those in municipalities with low PHC on health outcomes or no coverage. Infant mortality declined by In Niger, a prospective study found that 42% and 45.5% in the intervention and con- there was no additional survival advantage trol groups, respectively, a nonsignificant dif- for children in villages with a “village health ference (Morsch et al., 2001). A possible ex- team”present, although the presence of a dis- planation for the lack of an effect might have pensary lowered the odds of death by 32%, been the inability to control for variables re- as compared with villages with no services lated to the performance of primary care ser- (Magnani et al., 1996). In explanation for the vices, such as the technical quality of care or apparent lack of an effect, Magnani et al. sug- accessibility, which vary by municipality. gest the need for more comprehensive pack- Finally, in their systematic review of “in- ages of health services than those delivered by tegration” of primary care in developing the project, because the single interventions countries, Briggs et al. (2001) discuss an es- were possibly offset by continued high levels sential feature of primary care: the extent to of exposure to other unchanged factors. which it provides a range of services meant to In the Philippines, the Bohol project pro- attend to most common healthcare problems. vided very low-cost PHC services to a popula- This feature of primary care is more often tion of about 400 000 residents for 5 years. termed “comprehensiveness.” The review of The evaluation included pre- and postin- Briggs et al. (2001) contains only 4 studies, tervention surveys and comparison with a and they conclude that no overall conclusions control village. The project increased the can be drawn from their results. As a possible utilization of some health services but did explanation for the lack of a conclusive not significantly decrease infant mortality finding, the authors point to the poor quality (Williamson, 1982). Williamson suggests that of many of the studies conducted, including potential reasons for the lack of an effect in- poor recording of outcomes, inadequate clude the generally poor quality of health ser- randomization processes, and control groups vices provided, a selective rather than a com- that were not entirely comparable with exper- prehensive approach to PHC with a strong imental groups. Moreover, each study defined emphasis on (fertility did de- and measured integration in a different cline), and overworked and/or inadequately way. trained staff. In a retrospective study with control com- DISCUSSION munities in the Gambia (De Francisco et al., 1994), there were no significant differences This review of the evidence of the ef- (P = .88) in under-5 mortality between vil- fectiveness of PHC on population health in lages with VHWs and those without them low-income countries has shown that sev- (35.5/1000 vs 35.8/1000). De Francisco et al. eral analyses provide consistent evidence suggest that different health service utiliza- of the impact of PHC on improved health tion patterns (based on the type of child ill- outcomes. Nevertheless, many studies suf- nesses) and preferences for traditional heal- fered from important methodological weak- ers may partially explain the lack of effect. In nesses, including inadequate controls for addition, there was no indication that these individual- or community-level confounders VHWs were achieving PHC functions, includ- in multivariable analyses. Reductions in in- ing provision of good-quality care and refer- fant mortality (the most frequently studied ral to trained healthcare professionals when outcome) attributed to PHC actions averaged indicated. about 40% and varied from 0% to as high In Brazil, one study found that participa- as a 71% over intervention periods ranging tion in FHP between 1994 and 1998 did not between 2 and 10 or more years. Costs for significantly improve child health indicators comprehensive PHC programs ranged from in municipalities with high coverage, as op- about $10 to $35 per capita per year. 158 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009

Despite the apparent consistency of results, PHC tasks. They also point to the importance analysis of the studies revealed that PHC has of accurately measuring variations in the tech- only rarely been evaluated in a consistent and nical quality of primary care delivered, a topic reproducible way. Rather, it is often only var- that certainly deserves far more attention in ious aspects of health services that are as- the literature reviewed here. sumed to be part of PHC that have been the fo- In view of the limitations of these studies, cus of attention. For example, all but 3 studies an agenda for the evaluation of the contribu- measured PHC exposure as residence within tion of PHC tasks and functions to population a geographic area in which the PHC program health would benefit from the following con- or project was implemented. In addition, the siderations. definition of the PHC program also varied con- First, a clear conceptualization of primary siderably, from the mere presence of a VHW care is needed, including specification of in a community to the use of specific ser- each of its component features, for example, vices to the development of an integrated first-contact access and use, longitudinality network of health and social services in the (person-focused care over time), comprehen- community. For this reason, there is little that siveness (addressing the breadth of common can be gleaned regarding the mechanisms by health needs), and coordination (integration which these PHC approaches might achieve of services with other levels of care). important primary care functions, such as Second, studies should start with a con- first-contact access, longitudinality, compre- ceptual framework to guide the overall eval- hensiveness, and coordination of care. uation, design the characteristics under con- The general failure to use an operational sideration, and describe relationships among conceptualization of PHC has also made iden- different levels and types of data to be col- tification of studies about PHC difficult. For lected in the evaluation process. This frame- this reason, the literature probably contains work should provide a model of how primary more evidence than is discoverable from the care is conceptualized in relation to biologi- abstracts or titles of published articles. cal, social, and environmental influences on Publication bias is also likely to have lim- health (Starfield, 2001). ited the scope of this review. Many suc- Third, future studies require clear speci- cessful (as well as unsuccessful) experiences fication and measurement of the PHC sys- have simply not been documented in peer- tem, including specific structural character- reviewed journals. As an illustration, we were istics (input and policy), process (service unable to retrieve any peer-reviewed articles delivery modalities), and relevant health that adequately assessed the impact of PHC outcomes and outputs. on population health in Sri Lanka, China, or Fourth, as noted throughout the PHC lit- Vietnam, although each country’s PHC ap- erature (Chen et al., 1993; Hill et al., 2000), proach has been discussed elsewhere as “suc- there is still an urgent need for more rig- cessful”(Bloom, 1998; Fritzen, 2007; Halstead orous research designs that allow testing of et al., 1985). the independent effect of primary care on In addition, there is little peer-reviewed evi- outcomes over time. This should include dence on the role of PHC on improvements in individual- and community/contextual-level adult health in low- and middle-income coun- data derived from longitudinal sources, appro- tries, because most published studies have fo- priate control or comparison groups, and con- cused only on infant and under-5 mortality. trol for relevant individual- and contextual- Thus, the potential for PHC to help control level covariates. Such evaluations will require adult chronic and infectious diseases in the de- a commitment from donor organizations and veloping world remains largely unexplored. national governments to provide necessary re- The studies that found no effect of PHC on sources and to ensure the scientific integrity health indirectly provide support for a com- of the research process. prehensive approach to PHC: most involved In the short term, 3 approaches could interventions that focused only on selective be implemented to aid in providing more Primary Healthcare in Low- and Middle-Income Countries 159 systematic evaluation of primary care, as opment Goals (WHO, 2008). Basing health follows: systems more strongly on PHC represents an 1. Existing or planned cohort studies could important strategy to address emerging health begin to incorporate PHC measures problems (Fuster & Voute, 2005), scale up through the use of validated instruments existing interventions, and effectively com- such as the Primary Care Assessment bat health threats such as HIV/AIDS (Buve Tools (Harzheim et al., 2006; Macinko et al., 2003), (Mahendradhata et al., 2007; Pasarin et al., 2007). et al., 2003), chronic illnesses (Rothman & 2. Standardized surveys such as the Demo- Wagner, 2003), and others. graphic and Health Surveys or Living These observations are also relevant to the Standards Measurement Surveys could renewal of primary care in the . include modules derived from the Pri- Recently, the American Academy of Family mary Care Assessment Tools along with Physicians, the American Academy of Pedi- health system variables to identify how atrics, the American College of Physicians, and where populations are receiving ef- and the American Osteopathic Association fective PHC services. (2007) united to endorse the “Joint Princi- 3. Researchers could direct their attention ples of the -Centered ,” to countries that are currently undergo- which describes characteristics of a patient- ing reform of their primary care system, centered medical home (PCMH) as including thus opening the possibility for analy- a personal physician, physician-directed med- sis of natural experiments in which re- ical practice, whole-person orientation, coor- formed states or municipalities could dinated and integrated care, quality and safety, be compared with otherwise similar re- enhanced access to care, and payment that gions without reformed primary care “appropriately recognizes the added value systems. Better yet, experimental assign- provided to ....”Lessons learned from ment of different PHC approaches could the evaluation of PHC in the developing world be used to help phase in reforms and may also have relevance to the assessment of more rigorously evaluate their impact the PCMH, as it is apparent that definitions (King et al., 2007). and tools of measurement should be consis- Finally, there is a need to encourage the tent, standardized, and based on evidence of publication of evaluations of PHC experi- effectiveness of primary care components. ences, both successful and unsuccessful, so Without greater attention to these aspects, the that the PHC approach can be guided by a PCMH model may fall short of reaching its wider body of high-quality evidence. goal of renewing a PHC approach to health- care organization and delivery in the United CONCLUSION States. As national governments, the WHO, and The WHO proposal for renewing PHC re- other international organizations move to re- inforces the idea that strengthened health sys- new their PHC strategies, greater clarity in tems should be viewed as a necessary (though specifying PHC in terms that allow for more not sufficient) condition for meeting interna- standardized measurement and investment in tionally agreed-upon development goals such rigorous evaluation of PHC effectiveness and as those contained in the Millennium Devel- its effects on equity will be essential.

REFERENCES

Afari, E. A., Nkrumah, F.K., Nakana, T., Sakatoku, H., Hori, experience. Central African Journal of Medicine, H., & Binka, F. (1995). Impact of primary health care 41(5), 148–153. on childhood and mortality in rural Ghana: The Gomoa American Academy of Family Physicians, American 160 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009

Academy of , American College of Physi- health systems for more effective HIV/AIDS preven- cians, & American Osteopathic Association. (2007). tion and care. The International Journal of Health Joint principles of the patient-centered medical Planning & Management, 18(Suppl. 1), S41–S51. home. Patient Centered Primary Care Collabora- Chahnazarian, A., Ewbank, D. C., Makani, B., & Ekouevi, tive. Retrieved November 5, 2008, from http://www. K. (1993). Impact of selective primary care on child- medicalhomeinfo.org/joint%20Statement.pdf hood mortality in a rural health zone of Zaire. Interna- Arifeen, S., Black, R. E., Antelman, G., Baqui, A., Caulfield, tional Journal of Epidemiology, 22(Suppl. 1), S32– L., & Becker, S. (2001). Exclusive re- S41. duces acute respiratory and diarrhea deaths Chen, L. C., Hill, A. G., Murray, C. J., & Garenne, M. among infants in Dhaka slums. Pediatrics, 108, E67. (1993). A critical analysis of the design, results and Berman, P. (2000). Organization of ambulatory health implications of the mortality and use of health ser- care provision in developing countries. A critical de- vices surveys. International Journal of Epidemiol- terminant of health system performance in developing ogy, 22(Suppl. 1), S73–S80. countries) Bulletin of the World Health Organization, Chen, L. C., Rahman, M., D’Souza, S., Chakraborty, J., Sar- 78(6), 791–802. dar, A. M., & Yunus, M. (1983). Mortality impact of Bang, A. T., Bang, R. A., Baitule, S. B., Reddy, M. H., & an MCH-FP program in Matlab, Bangladesh. Studies in Deshmukh, M. D. (1999). Effect of home-based neona- Family Planning, 14(8/9), 199–209. tal care and management of sepsis on neonatal mor- Coleman, R., Gill, G., & Wilkinson, D. (1998). Noncom- tality: Field trial in rural India. The Lancet, 354(9194), municable disease management in resource-poor set- 1955–1961. tings: A primary care model from rural South Africa. Bang, A. T., Reddy, H. M., Deshmukh, M. D., Baitule, S. Bulletin of the World Health Organization, 76(6), B., & Bang R. A. (2005). Neonatal and infant mortality 633–640. in the ten years (1993 to 2003) of the Gadchiroli field De Francisco, A., Schellenberg, J. A., Hall, A. J., Green- trial: Effect of home-based neonatal care. Journal of wood, A. M., Cham, K., & Greenwood, B. M. (1994). Perinatology, 25(Suppl. 1), S92–S107. Comparison of mortality between villages with and Becker, S. R., Diop, F., & Thornton, J. N. (1993). Infant without primary health care workers in Upper River and child mortality in two counties of Liberia: Results Division, The Gambia. The Journal of Tropical of a survey in 1988 and trends since 1984. Interna- Medicine and Hygiene, 97(2), 69–74. tional Journal of Epidemiology, 22(Suppl. 1), S56– den Besten, L., Bac, M., Glatthaar, I. I., & , A. R. S63. (1995). Changes in the anthropometric status of rural Bindman, A. B., Grumbach, K., Osmond, D., Vranizan, African under-fives during a decade of primary health K., & Stewart, A. L. (1996). Primary care and receipt care. The Journal of Tropical Medicine and Hygiene, of preventive services. Journal of General Internal 98(6), 361–366. Medicine, 11(5), 269–276. D’Souza, R. M., & Bryant, J. H. (1999). Determinants of Bloom, G. (1998). Primary health care meets the market childhood mortality in slums of Karachi, Pakistan. Jour- in China and Vietnam. Health Policy, 44, 233–252. nal of Health and Population in Developing Coun- Bohler, E. (1994). Has primary health care reduced in- tries, 2(1), 33–44. fant mortality in East Bhutan? The effects of primary Dutt, D., & Srinivasa, D. K. (1997). Impact of care and birth spacing on infant and child mor- and child health strategy on child survival in a rural tality patterns in East Bhutan. Journal of Tropical community of Pondicherry. Indian Pediatrics, 34(9), Pediatrics, 40, 256–260. 785–792. Brazilian Ministry of Health, Department of Primary Dugbatey, K. (1999). National health policies: Sub- Care. (2006, October). Atenc¸ao˜ Basica´ e a Saude´ da Saharan African case studies (1980–1990). Social Sci- Fam´ılia [Primary care and family health]. Retrieved ence & Medicine, 49(2), 223–239. November 20, 2006, from http://dtr2004.saude. Edward, A., Ernst, P., Taylor, C., Becker, S., Mazive, E., & gov.br/dab/abnumeros.php Perry, H. (2007). Examining the evidence of under-five Briggs, C. J., Capdegelle, P., & Garner, P.(2001). Strategies mortality reduction in a community-based programme for integrating primary health services in middle- and in Gaza, Mozambique. Transactions of the Royal Soci- low-income countries: Effects on performance, costs ety of Tropical Medicine and Hygiene, 101(8), 814– and patient outcomes. Cochrane Database of System- 822. atic Reviews, 4, CD003318. Emond, A., Pollock, J., Da Costa, N., Maranh˜ao, T., & Brockerhoff, M., & Derose, L. (1996). Child survival in Macedo, A. (2002). The effectiveness of community- East Africa: The impact of preventive health care. based interventions to improve maternal and infant World Development, 24(12), 1841–1857. health in the northeast of Brazil. Revista Panameri- Bryce, J., el Arifeen, S., Pariyo, G., Lanata, C., Gwatkin, cana de Salud Publica,´ 12(2), 101–110. D., Habicht, J. P., & Multi-Country Evaluation of IMCI England, S., Loevinsohn, B., Melgaard, B., Kou, U., & Jha, Study Group. (2003). Reducing child mortality: Can P. (2001). The evidence base for interventions to re- public health deliver? The Lancet, 362(9378), 159– duce mortality from -preventable diseases in 164. low and middle-income countries (CMH Working Pa- Buve, A., Kalibala, S., & McIntyre, J. (2003). Stronger per WG5:10). Commission on Macroeconomics and Primary Healthcare in Low- and Middle-Income Countries 161

Health. Retrieved January 3, 2004, from http://www. Haines, M., & Avery, R. (1982). Differential infant and cmhealth.org/docs/wg5 paper10.pdf child mortality in Costa Rica. Population Studies, 36, Evans, R. (2008). Thomas McKeown, Meet Fidel Castro: 31–43. Physicians, population health and the Cuban paradox. Halstead, S. B., Walsh, J. A., & Warren, K. S. (1985). Good Healthcare Policy, 3(4), 21–32. health at low cost. New York: The Rockefeller Foun- Fauveau, V., Stewart, M. K., Chakraborty, J., & Khan, S. dation. A. (1992). Impact on mortality of a community-based Harzheim, E., Starfield, B., Rajmil, L., Alvarez-Dardet, C., programme to control acute lower respiratory tract in- & Stein, A. T. (2006). Internal consistency and reliabil- fections. Bulletin of the World Health Organization, ity of Primary Care Assessment Tool (PCATool–Brasil) 70(1), 109–116. for child health services. Cadernos de Saude´ Publica,´ Fauveau, V., Stewart, M. K., Khan, S. A., & Chakraborty, 22(8), 1649–1659. J. (1991). Effect on mortality of community-based Hill, A. G., MacLeod, W. B., Joof, D., Gomez, P., & Wal- maternity-care programme in rural Bangladesh. The raven, G. (2000). Decline of mortality in children Lancet, 338(8776), 1183–1186. in rural Gambia: The influence of village-level pri- Filmer, D., Hammer, J., & Pritchett, L. (2000). Weak links mary health care. Tropical Medicine & International in the chain: A diagnosis of health policy in poor coun- Health, 5(2), 107–118. tries. The World Bank Research Observer, 15(2), 199– Infante, A. I., & de Mata, A. (2000). Lopez-Acuna,˜ Re- 224. forma de los sistemas de salud an America Latina y el Forrest, C. B., & Starfield, B. (1996). The effect of first- Caibe: Situacion´ y tendencias. Revista Panamericana contact care with primary care clinicians on ambula- de Salud Publica,´ 8(1/2), 13–20. tory health care expenditures. Journal of the Family King, G., Gakidou, E., Ravishankar, N, Moore, R. T., Lakin, Practice, 43(1), 40–48. J., Vargas, M., et al. (2007). A “politically robust”exper- Forrest, C. B., & Starfield, B. (1998). Entry into primary imental design for public policy evaluation, with appli- care and continuity: The effects of access. American cation to the Mexican universal pro- Journal of the Public Health, 88(9), 1330–1336. gram. Journal of Policy Analysis and Management, Franco, M., Kennelly, J. F., Cooper, R. S., & Ordu´nez-˜ 26(3), 479–506. Garc´ıa, P. (2007). Health in Cuba and the millennium Klijzing, F., & Taylor, H. (1982). The decline of infant development goals. Revista Panamericana de Salud mortality in Costa Rica, 1950–1973: Modernization or Publica,´ 21(4), 239–250. technological diffusion? Malaysian Journal of Tropi- Frankenberg, E. (1995). The effects of access to health cal Geography, 5, 22–29. care on infant mortality in Indonesia. Health Transi- Lewis, M., Eskeland, G., & Traa-Valerezo, X. (2004). Pri- tion Review, 5(2), 143–163. mary health care in practice: Is it effective? Health Pol- Freedman, L. P., Waldman, R. J., de Pinho, H., & Wirth, M. icy, 70, 303–325. E. (2005). Who’s got the power? Transforming health Macinko, J., Almeida, C., & de S´a, P.(2007). A rapid assess- systems to improve the lives of women and chil- ment methodology for the evaluation of primary care dren (Millennium Project Task Force 4: Child Health organization and performance in Brazil. Health Policy and Maternal Health—Interim Report, UN Millennium & Planning, 22(3), 167–177. Project). United Nations: New York. Macinko, J., Souza, F., Guanais, F.C., & da Silva Simoes,˜ C. Frenk, J., Sepulveda,´ J., Gomez-Dant´ es,´ O., & Knaul, F. C. (2007). Going to scale with community-based pri- (2003). Evidence-based health policy: Three genera- mary care: An analysis of the family health program tions of reform in Mexico. The Lancet, 362(9396), and infant mortality in Brazil, 1999–2004. Social Sci- 1667–1671. ence & Medicine, 65(10), 2070–2080. Fritzen, S. (2007). Legacies of primary health care in an Macinko, J., Guanais, F., & Souza, F. (2006). An evaluation age of health sector reform: Vietnam’s commune clin- of the impact of the family health program on infant ics in transition. Social Science & Medicine, 64, 1611– mortality in Brazil, 1990–2002. Journal of Epidemiol- 1623. ogy and Community Health, 60, 13–19. Fuster, V., & Voute, J. (2005). MDGs: Chronic diseases Macinko, J., Starfield, B., & Shi, L. (2003). The contri- are not on the agenda. The Lancet, 366(9496), 1512– bution of primary care systems to health outcomes 1514. within Organization for Economic Cooperation and Greenwood, B. M., Bradley, A. K., Byass, P., Greenwood, Development (OECD) countries, 1970–1998. Health A. M., Menon, A., Snow, R. W., et al. (1990). Evaluation Services Research, 38(3), 831–865. of a primary health care programme in The Gambia. Mackintosh, M. (2000). Do health care systems contribute II. Its impact on mortality and morbidity in young chil- to inequalities? In D. Leon & G. Walt (Eds.), Poverty, dren. The Journal of Tropical Medicine and Hygiene, inequality and health (pp. 175–193). Oxford: Oxford 93(2), 87–97. University Press. Gutierrez,´ G., Reyes, H., Fern´andez, S., Perez,´ L., Perez-´ Magnani, R. J., Rice, J. C., Mock, N. B., Abdoh, A. A., Mer- Cuevas, R., & Guiscafre,´ H. (1999). Impact of health cer, D. M., & Tankari, K. (1996). The impact of pri- services, sanitation and literacy in the mortality of chil- mary health care services on under-five mortality in dren under 5 years of age. Salud Publica´ de M´exico, rural Niger. International Journal of Epidemiology, 41(5), 368–375. 25(3), 568–577. 162 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009

Mahendradhata, Y., Lambert, M. L., Van Deun, A., Matthys, Impact of a community-based comprehensive primary F., Boelaert, M., & van der Stuyft, P. (2003). Strong health care programme on infant and child mortality in general health care systems: A prerequisite to reach Bolivia. Journal of Health, Population, and Nutrition, global tuberculosis control targets. The Interna- 21(4), 383–395. tional Journal of Health Planning & Management, Phillips, J. F., Bawah, A. A., & Binka, F. N. (2006). Acceler- 18(Suppl. 1), S53–S65. ating reproductive and child health programme impact Margaret Chan puts primary health care centre stage at with community-based services: The Navrongo exper- WHO. (2008). The Lancet, 371(9627), 1811. iment in Ghana. Bulletin of the World Health Organi- Mercer, A., Khan, M. H., Daulatuzzaman, M., & Reid, zation, 84(12), 949–955. J. (2004). Effectiveness of an NGO primary health Pison, G., Trape, J. F., Lefebvre, M., & Enel, C. (1993). care programme in rural Bangladesh: Evidence from Rapid decline in child mortality in a rural area of Sene- the management information system. Health Policy & gal. International Journal of Epidemiology, 22(1), Planning, 19(4), 187–198. 72–80. Ministry of Health of Brazil. (2003). The family health pro- Reyes, H., Perez-Cuevas, R., Salmeron, J., Tome, P., Guis- gram: Broadening coverage to consolidate the change cafre, H., & Gutierrez,´ G. (1997). Infant mortality due in the primary care model. Revista Brasileira Saude´ to acute respiratory infections: The influence of pri- Materno Infantil, 3(1), 113–125. mary care processes. Health Policy & Planning, 12(3), Moore, D., Castillo, E., Richardson, C., & Reid, R. J. (2003). 214–223. Determinants of health status and the influence of Riveron´ Corteguera, R. (2000). Estrategias para reducir la primary health care services in Latin America, 1990– mortalidad infantil, Cuba 1959–1999. Revista Cubana 98. The International Journal of Health Planning & de Pediatria, 72(3), 147–164. Management, 18(4), 279–292. Rohde, J., Cousens, S., Chopra, M., Tangcharoensathien, Morsch, E., Chavannes, N., van den Akker, M., Sa, H., & V., Black, R., Bhutta, Z. A., et al. (2008). 30 years af- Dinant, G. J. (2001). The effects of the Family Health ter Alma-Ata: Has primary health care worked in coun- Program in Cear´a state, northeastern Brazil. Archives tries? The Lancet, 372(9642), 950–961. of Public Health, 59, 151–165. Rosero-Bixby, L. (1986). Infant mortality in Costa Rica: Ex- Oluwole, D., Mason, E., & Costello, A. (2000). Man- plaining the recent decline. Studies in Family Plan- agement of childhood illness in Africa. Early evalua- ning, 17, 57–65. tions show promising results. British Medical Journal, Rosero-Bixby, L. (1991). Socioeconomic development, 320(7235), 594–595. health interventions and mortality decline in Costa Pan American Health Organization, World Health Organi- Rica. Scandinavian Journal of Social Medicine: Sup- zation. (2005). Renewing primary health care in the plementum, 46, 33–42. Americas [Position paper]. Washington, DC: Author. Rosero-Bixby, L. (2004a). Evaluacion´ del impacto de la re- Pasarin, M. I., Berra, S., Rajmil, L., Solans, M., Borrell, C., & forma del sector de la salud en Costa Rica mediante un Starfield, B. (2007). An instrument to evaluate primary estudio cuasiexperimental. Revista Panamericana de health care from the population perspective. Atencion Salud Publica,´ 15(2), 94–1003. Primaria, 39(8), 395–401. Rosero-Bixby, L. (2004b). Spatial access to health care in Pence, B. W., Nyarko, P., Phillips, J. F., & Debpuur, C. Costa Rica and its equity: A GIS-based study. Social Sci- (2007). The effect of community nurses and health ence & Medicine, 58, 1271–1284. volunteers on child mortality: The Navrongo Com- Rothman, A. A., & Wagner, E. H. (2003). Chronic illness munity Health and Family Planning Project. Scan- management: What is the role of primary care? Annals dinavian Journal of Public Health, 35(6), 599– of Internal Medicine, 138(3), 256–261. 608. Sepulveda, J., Bustreo, F., Tapia, R., Rivera, J., Lozano, Perry, H., Berggren, W., Berggren, G., Dowell, D., R., Ol´aiz, G., et al. (2006). Improvement of child sur- Menager, H., Bottex, E., et al. (2007). Long-term reduc- vival in Mexico: The diagonal approach. The Lancet, tions in mortality among children under age 5 in ru- 368(9551), 2017–2027. ral Haiti: Effects of a comprehensive health system in Simms, C., & Rowson, M. (2003). Reassessment of an impoverished setting. American Journal of Public health effects of the Indonesian economic crisis: Health, 97(2), 240–246. Donors versus the data. The Lancet, 361(9366), 1382– Perry, H., Cayemittes, M., Philippe, F., Dowell, D., Dor- 1385. tonne, J. R., Menager, H., et al. (2006). Reducing under- Spiegel, J. M., & Yassi, A. (2004). Lessons from the mar- five mortality through Hopital Albert Schweitzer’s in- gins of globalization: Appreciating the Cuban health tegrated system in Haiti. Health Policy & Planning, paradox. Journal of Public Health Policy, 25(1), 85– 21(3), 217–230. 110. Perry, H., Robison, N., Chavez, D., Taja, O., Hilari, C., Starfield, B. (1996). Is strong primary care good for health Shanklin, D., et al. (1998). The census-based, impact- outcomes? In J. Griffin (Ed.), The future of primary oriented approach: Its effectiveness in promoting care (pp. 18–29). Office of : London. child health in Bolivia. Health Policy & Planning, Starfield, B. (1998). Primary care: Balancing health 13(2), 140–151. needs, services and technology. New York: Oxford Perry, H. B., Shanklin, D. S., & Schroeder, D. G. (2003). University Press. Primary Healthcare in Low- and Middle-Income Countries 163

Starfield, B. (2001). Basic concepts in population health tres under decentralized government in Kerala, India. and health care. Journal of Epidemiology and Com- Health Policy & Planning, 19(1), 41–51. munity Health, 55(7), 452–454. Velema, J. P., Alihonou, E. M., Gandaho, T., & Hounye, Starfield, B., & Shi, L. (2002). Policy relevant determinants F. H. (1991). Childhood mortality among users and of health: An international perspective. Health Policy, non-users of primary health care in a rural West African 60(3), 201–216. community. International Journal of Epidemiology, Starfield, B., Shi, L., Grover, A., & Macinko, J. (2005a). 20(2), 474–479. The effects of specialist supply on populations’ health: Victora, C. G., Bryce, J., Fontaine, O., & Monasch, R. Assessing the evidence. Health Affairs (Millwood), (2000). Reducing deaths from diarrhoea through oral (Suppl. Web Exclusives), W5-97–W5-107. rehydration therapy. Bulletin of the World Health Or- Starfield, B., Shi, L., & Macinko, J. (2005b). The contribu- ganization, 78(10), 1246–1255. tion of primary care to health systems and health. The Waitzkin, H. (1997). Primary care in Cuba: Low- and Milbank Quarterly, 83(3), 457–502. high-technology development pertinent to family Svitone, E., Garfield, R., Vasconcelos, M. I., & Araujo medicine. Journal of Family Practice, 45, 250– Craveiro, V. (2000). Primary health care lessons from 258. the northeast of Brazil. Pan American Journal of Pub- Williamson, N. (1982). An attempt to reduce infant and lic Health, 7(5), 293–302. child mortality in Bohol, Philippines. Studies in Fam- Taylor, W. R., Chahnazarian, A., Weinman, J., Wernette, ily Planning, 13(4), 106–117. M., Roy, J., Pebley, A. R., et al. (1993). Mortality and use World Health Organization. (1978, September). Primary of health services surveys in rural Zaire. International health care. Report of the international conference on Journal of Epidemiology, 22(Suppl. 1), S15–S19. primary health care, Alma-Ata, USSR, September 6–12, Vapattanawong, P., Hogan, M. C., Hanvoravongchai, P., 1978. Geneva: Author. Gakidou, E., Vos, T., Lopez, A. D., et al. (2007). Reduc- World Health Organization. (2004). Multi-country eval- tions in child mortality levels and inequalities in Thai- uation of IMCI: Main findings. Retrieved Jan- land: Analysis of two censuses. The Lancet, 369(9564), uary 29, 2004, from http://www.who.int/imci-mce/ 850–855. findings.htm Varatharajan, D., Thankappan, R., & Jayapalan, S. (2004). World Health Organization. (2008). Primary health Assessing the performance of primary health cen- care—Now more than ever. Geneva: Author. 164 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009 ) 0.4); = 2 R continues ( .002 for 1991). = P .001). During both periods .01). From 1988 to 1989, the < P < 0.3). P = 2 .003) by 32% between 1986/1987 and R = .001 for 1984 and P < P indicator of access) reduced risk of death ( mortality was 48% lower inthan the in intervention the area control area.children ALRI-specific younger mortality than in 5 all ydeclined in ( the intervention area 1988/1989. area, ALRI-specific mortality from 198628% to lower 1987 in was the interventioncomparison area area than ( in the 1999–2002 from 52.8 to 28.3/1000poorest (among children) the and from 41.6other to children. 28.2/1000 From among 2000 tomortality 2002, declined under-5 from 15.8 toand 9.7/1000 from (poorest) 10.8 to 9.2 /1000 among other children. (1984 and 1991), children ofbirth mothers frequency with had a significantly high highermortality infant (268 in 1984 andwith 93 low in birth 1991) frequency than (47 those 1991) in ( 1984 and 12 in children in regular contact withrisk a of VHW death had ( lower 1984 to 49 in 1991 ( Measles vaccination before the first birthday (an For all children younger than 5 y in the intervention Infant mortality declined in the 12 NGO areas in mortality under-5 mortality and maternal mortality Infant, under-5, Infant mortality Infant mortality was reduced significantly from 145 in family health visitor to provide family planning counseling, contraceptives (health education, vaccination, treatment) Visit to VHWs Under-5 Presence of CHWs Cause-specific Presence of a Population measures outcome Impact/result of age than 5 y in Matlab children a C Children younger C Women and D Children 4–35 mo et al., 1991) et al. 1991, 1992) et al., 2004) Benin (rural) (Velema Bangladesh (Fauveau Appendix Review of Literature on the Evaluation of Primary Healthcare in Developing Countries Country/region(reference)Bangladesh (Mercer Type PHC Main Bhutan (Bohler, 1994) E Infants VHW activities Primary Healthcare in Low- and Middle-Income Countries 165 ) .001, < continues ( = P .01). Access to clean < P 141.3–72.7) or 52.1% (95% CI = did not significantly improve childcompared health with indicators no participation. Theinfant proportionate mortality declined by 42%intervention and and 45.5% control in groups, the respectively, and the differences were not significant. 4.5% decrease in infant mortality, controllingother for health all determinants ( water and hospital beds perassociated 1000 with were infant negatively mortality, whereas female illiteracy, fertility rates, and mean incomepositively associated were with infant mortality. The program may reduce infant mortalitythrough at reductions least in partly diarrhea deaths. d’Ivoire and Ghana in termsThis of relationship health did outcomes. not appearincome to only be because related richer to countriesscored (Cote worse d’Ivoire) than Zimbabwe andauthor Ghana. concludes The that “policies formulatedimplemented and in accordance with keyprinciples PHC could account for improvementsnational in health status.” (comparison areas) and 98.5/1000 (intervention areas), a difference of 107/100095% deaths CI ( 35.2–68.8%) lower mortality in theareas. intervention Program costs estimated atcapita. about $10 per Botswana and Zimbabwe performed better than Cote In 1992–1993, under-5 mortality was 205.5/1000 life expectancy mortality Infant mortality Participation in the program between 1994 and 1998 Infant mortality A 10% increase in FHP coverage was associated with a Infant mortality, Under-5 agents program coverage population covered by the FHP food, nutrition, water, sanitation, maternal/child health policies public health services, including improvement of clean water and sanitation Community health Proportion of Health education, Comprehensive Population measures outcome Impact/result different municipal areas population (states and microregions) of 4 countries intervention and noninter- vention areas a A Brazilian C Infants in C Children in G National policies 2001) 2006, 2007) Ghana, and Zimbabwe (Dugbatey, 1999) 1998, 2003) Brazil (Morsch et al., Brazil (Macinko et al., Botswana Cote d’Ivoire, Country/region(reference)Bolivia (Perry et al., Type PHC Main 166 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009 ) continues ( reduction in infant mortality from37/1000; 60/1000 decreased to maternal mortality; improved infant feeding and caretaking practices;immunization improved rates; increased rates ofwith consultation physicians; and a greatertaking percentage place of in births community clinicshospitals. rather than in prenatal care, vaccination coverage, institutional deliveries, and infant mortality improved.success Program includes longitudinality; family and community orientation, first contact, and intersectorial collaboration. Weaknesses include referral mechanisms, access barriers, andconditions. living in both adults (2%) andadditional children years (8%). of For reform, every child 5 reduced mortality by was 13% and adultPopulation mortality with declined limited by access 4%. todecreased health 15% services in reform areasnonreform (only areas). 2% In in multivariate analyses,1970 from to 1980, PHC accountedmortality for declines, 41% other of medical infant caresocioeconomic (32%), progress (22%), and fertility(5%). decline over 40 y associated withmortality 40% in decline the in 1970s, infant 1980s,socioeconomic and conditions 1990, remained while similar. After 30 mo, community surveys reported a significant Percentages of breast-feeding, giving ORT, receiving Reform of primary care significantly reduced mortality Changes in access, organization, and delivery of PHC infant mortality (by cause) and fertility cardiovascular disease mortality (adults) Maternal and Infant mortality Infant mortality Infant mortality, program community health agents program on when they began the health reform process services, organization and service delivery, and policies Coverage by VHW Coverage by District level, based a state ´ Population measures outcome Impact/result women, and children younger than 5y children in Cear women a E Mothers and C Women, pregnant A, A Infants and G, G Infants and adults Description of PHC ´ on a) (Svitone ´ et al., 2000) (Rosero-Bixby, 1991, 2004a, 2004b) 2007; River Corteguera, 2000) et al., 2002) Brazil (Cear Costa Rica Cuba (Franco et al., Country/region(reference)Brazil (Natal) (Emond Type PHC Main Primary Healthcare in Low- and Middle-Income Countries 167 ) continues ( .88) in under-5 mortality = P 35.5 per 1000/y vs 35.8 per 1000/y). improved the health of childrenInfant in mortality the declined villages. from 114.6/1000in live 1987 births to 40.8/1000 livemortality births decreased in from 1990. 155.6/1000 Under-5 live1987 births to in 61.2/1000 live births in 1990. (under-5 mortality declined by 14%health in volunteers 5 (under-5 y), mortality local increased14%); by a combination of bothincreased (under-5 by mortality 8%); and nomortality new interventions decreased by (under-5 4%). between villages with VHWs and= those without (rate non-PHC villages. The decline wasvillages sharper (infant in mortality PHC 134/1000 tonon-PHC 69/1000) villages vs (155/1000 to 91/1000).1994, Since after support and fundingthe for trend PHC reversed declined, (infant mortalityvillages 98/1000 vs in 78/1000 PHC in non-PHC villages). PHC services decreased infant and child mortality and The study combined 4 arms: professional nurses Infant and under-5 mortality declined in both PHC and under-5 mortality under-5 mortality under-5 mortality Infant and Infant and Child mortality No significant difference ( Infant and services (health education on infant feeding, disease prevention, and drug distribution) professional nurse, volunteers, and both in different communities healthcare use, health education, environmental health, nutrition, treatment, and referrals by VHWs and TBAs PHC program (nurses, CHWs, TBAs, supplies, services, participation) Provision of Presence of a Vaccination, Population measures outcome Impact/result than 5 y in 3 villages children in intervention areas than 5 y a A Women and C Children Presence of the C, C Children younger 2007; Phillips et al., 2006) 2000) Francisco et al. 1994; Greenwood et al., 1990) Ghana (Afari et al., 1995) C Children younger Ghana (Pence et al. Country/region(reference)The Gambia (Hill et al., Type PHC Main The Gambia (De 168 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009 ) continues ( .01). Similarly, infant < .01). Gains were sustained P < P services improved child survival, decreasedand infant child mortality. Resulted in highindex of child 91.2%; survival infant mortality201/1000 declined in from 1967 to 64/1000whereas live child births death in rates 1989, decreasedin from 1970 39.4/1000 to 18/1000 inage. 1992 among children 1–4 y of mortality declined from 75.5/1000 live38.8/1000 births live to births (45.7% netperinatal reduction) mortality and declined from 68.3/1000to births 47.8/1000 births (71.0% netintervention reduction) areas in ( after 7 y. at the end of thethan third the year targeted of 25%. intervention In byneonatal intervention more mortality areas, reduced from 62.0/1000births live at baseline (1993–1995) tobirths 25.5/1000 at live year 3 (1997–1998),percentage reflecting reduction a ( 62.2% net HAS areas than that inmortality the in rest HAS of areas Haiti. declined Infant has in remained the at mid-1970s around and 62–66/1000.in Infant the mortality rest of Haitihigher has (224 declined in from 1971–1973) nearly to 4(119/1000) about times in 2 1999. times HAS higher includescommunity-based comprehensive PHC, intersectorial actions, and referral . Provision of adequate maternal and child health Home-based neonatal care reduced neonatal mortality Infant and under-5 mortality reported 50% lowering in mortality; child survival index and perinatal mortality under-5 mortality Infant and child Infant and medical facilities, antenatal and under-5 , home visits by public health nurses, health education, nutrition supplements integrated primary, secondary, and tertiary care to people living within HAS catchment areas Availability of Activities of CHWs Neonatal, infant, Provision of Population measures outcome Impact/result from birth to 5 yofage infants in intervention and noninter- vention areas covered by HAS a B Infants followed C Population C, C Neonates and 1997) 2005) 2007) India (Dutt & Srinivasa, India (Bang et al. 1999, Country/region(reference)Haiti (Perry et al., 2006, Type PHC Main Primary Healthcare in Low- and Middle-Income Countries 169 ) continues ( 3.59–26.1). Significant predictors = .1) lower odds of death than an infant born < 9.68, 95% CI P intervention areas and infant mortality25% declined and by under-5 mortality declinedbaseline by levels. 28% Note from that thisintervention was focused a on selective tetanus PHC andimmunizations, childhood ORT, and treatment. (controlling for individual and family socioeconomics, access to care) on= ARI deaths (OR included inadequate referral, lack of(attended continuity by multiple physicians), andattended being by a private providerpublic (as provider). opposed to a before the addition of health workers. decreases the odds of deathto of that an clinic infant by with about access born 15%, before relative the to clinic the existed. infant health An workers infant are born added after to( a village has about 1.3% outputs; in another study, integrated programsoutcomes had similar to those ofthe vertical remaining programs. 2 In studies, integratedperformed programs worse than vertical ones. Immunization and malaria treatment increased in Primary care processes had an independent effect under-5 mortality from ARI Infant and Infant mortality Varied In one study, integration showed a positive effect on Combating Childhood Communicable Disease project of physicians seen, no. of visits, private or public physicians, antibiotics, hospital referral, access) care delivery vs vertical primary care services Presence of the Population measures outcome Impact/result children in program and nonprogram areas a F Published studies Integrated primary C Women and D Infants Primary care (no. D Infants Service availability Infant mortality Within a village, an increase of 1 maternity clinic 1997) 1993) 1995) Togo (Briggs et al., 2001) Mexico (Reyes et al., Liberia (Becker et al., Indonesia (Frankenberg, Country/region(reference)India, Nepal, Tanzania, Type PHC Main 170 JOURNAL OF AMBULATORY CARE MANAGEMENT/APRIL–JUNE 2009 ) .03) over continues 8; 95% CI ( = = P = 14.52; 95% CI = 2.22–28.81) as increasing the odds of infant death 2 y. Decreased utilization of hospitalscare. for routine Potential reasons include poor qualityhealth or services, quantity a of selective approachemphasis to on PHC family with planning (fertilityoverworked did and/or decline), inadequately trained staff. live births from 1970 toprimarily 1993. from Reductions diseases seen prevented byDiarrhea immunization. and ARIs are lowerof than Senegal; in only other 4% rural of areas deaths attributed to malaria. control clinical conditions of 68%, of 82% patients of with those with non–insulin-dependent , and 84% ofwith those . Patient-reported adherence to treatment increased from 79% to 87% ( 4.23–49.83) and frequent switching ofproviders healthcare (lack of longitudinality) (OR VHWs or doctors) (OR = from respiratory infections or diarrhea. likely to have died duringwere the children study in period villages than withsurvival no advantage services. in No villages withteam the present. village health program resulted in a 66%mortality reduction and in a infant 62% reduction in under-5 mortality. Under-5 mortality declined from 350 to 81 deaths/1000 The utilization of protocols enabled local nurses to Children in villages with a dispensary were 32% less Implementation of activities associated with the IMCI mortality chronic diseases mortality under-5 mortality Infant mortality The program had no effect on infant mortality. Under-5 Control of Infant mortality The use of traditional healers (as opposed to trained Under-5 Infant and (, (drugstores, hospital units) integrated PHC program in the area of study updated protocols for the management of chronic diseases (type of provider, longitudinality, ability to explain clearly) proximity and the use of health services coverage PHC centers Presence of Clinics using Health service use Geographical IMCI program Population measures outcome Impact/result children in the project area intervention areas chronic illnesses Karachi slums than 5 y than 5 y a E Infants in E Adults with C Women and C Children younger C Children younger D Children in 6 1993) et al., 1998) (Williamson, 1982) Bryant, 1999) 1996) et al., 2007) Senegal (Pison et al., South Africa (Coleman The Philippines Pakistan (D’Souza & Niger (Magnani et al., Country/region(reference)Mozambique (Edward Type PHC Main Primary Healthcare in Low- and Middle-Income Countries 171 althcare; TBA, traditional control study; E, repeated ; FHP, family health program; .01). The prevalence of severe < P .01). The prevalence of stunting decreased < P mortality: female literacy, BCG vaccination rate,safe water, access use to of ORT, and gross national product per capita. survival. In terms of attributableimmunizations risk, are incomplete responsible for 34%clean of water deaths, for lack 16%, of inadequatefor prenatal 11%, care and and high delivery fertilityother for health 12%, determinants while (covariates controlling include for socioeconomic and biologic status and infrastructure). than 5 y between 1982underweight and increased 1990. from The 28% prevalence to of and 31% then from declined 1982 to to 23% 1984, respectively and ( 19% in 1988 andsteadily 1990, from 33% in 1982prevalence to of 17% low in weight 1990. for The5% height low in (wasting) 1982 declined to from 1% in 1990 ( ORT, malaria treatment. Under-5 mortality declinedfrom by baseline 33% levels. The interventionhave was reduced “estimated mortality to at agesassociated 6–35 with mo the by change at in leastmay the 18%–23% have incidence been of responsible measles for and recorded the between full 1980–1984 28% and reduction 1985–1989.” increased from 3% inthen 1982 declined to rapidly 14% to in 5% 1984 in and 1988 and to 4% in 1990. Five main variables were associated with reduced under-5 Selective PHC services have an important impact on child PHC activities improved anthropometric of children younger Intervention included tetanus and childhood immunizations, mortality and under-5 mortality morbidity mortality Under-5 Neonatal, infant, Under-5 Under-5 use of ORT, breast-feeding, physician supply and immunization coverage services (growth monitoring, ORT, breast-feeding, immunization, family planning, food supplementation, female education) Combating Childhood Communicable Disease project Vaccination coverage, Use of prenatal care Provision of GOBI-FFF Presence of the Population measures outcome Impact/result Latin American countries and children younger than 5 y from Demographic and Health Surveys than 5 y children in program and nonprogram areas a E Children younger A Children in 22 C Women and G Neonates, infants, countries) (Moore et al., 2003) Malawi, Tanzania, Zambia (Brockerhoff & Derose, 1996) Besten et al., 1995) et al., 1993; Taylor et al. 1993) A, experimental or quasi-experimental; B, prospective design with control; C, repeated cross-sectional design with comparison/control; D, case- Latin America (22 Kenya, Madagascar, cross-sectional design without comparisoncontrol; F, systematic literature review; G, observation/qualitative/single cross-section. birth attendant; VHW, village health worker. Country/region(reference)South Africa (den Type PHC Main Abbreviations: ALRI, acute lower respiratoryHAS, infection; Hospital ARI, Albert-Schweitzer; acute IMCI, respiratory integrated infection; management CHW, of communitya childhood health illness; worker; ORT, CI, oral confidence rehydration interval therapy; OR, odds ratio; PHC, primary he Zaire (Chahnazarian