Internal Geographical Imbalances

This article discusses internal imbalances of health care in low-and middle-income countries. Throughout this article, ‘internal’ refers to within country, and the emphasis will lie on the differences between rural and urban areas. Much of the work in this area focuses on the imbalance in the quantity of health workers, but recent evidence indicates that imbalances in the quality of health care are as important. The paper discusses both, focusing throughout on the human resources aspect of health-care delivery.

What Geographical Imbalances?

With poverty reduction taking a prominent place on the international agenda in the early-1990s – later resulting in a consensus around the 2015 Millennium Development Goals (MDGs) – there has been an increased interest in rural service delivery. Many of the poor live in the countryside, where poverty is at its deepest, and the emphasis placed on coverage of cardinal interventions makes access to services in rural areas key to reach the MDGs. But although health outcomes are unfavorable in rural areas, there is also less care provided in those areas. This is sometimes referred to as ‘the inverse care law.’ These geographical imbalances have mostly been discussed in terms of shortages of health-care workers, which is perhaps best illustrated by the World Health Organization guideline recommending 2.28 health professionals – including doctors, nurses, and midwives – per 1000 inhabitants to allow the delivery of quality health services. Contemporary work is concerned with both the quality and quantity of services. Evidence indicates that low numbers are not the single constraint for the delivery of appropriate services. A narrow focus on the numbers of health personnel is therefore misguided. It also stands in the way of thinking critically about health care in remote areas, particularly in the context of rapid urbanization, as is the case in most developing countries, which may require more fundamental changes to rural health policies, as discussed in the section Encourage and Support Self-Help among Rural Populations. In what follows, the paper discusses the evidence on quantitative and qualitative imbalances in human resources for health (HRH).

Imbalances In The Number Of Health Workers

Although a focus on the quantity of health-care providers is not enough, considering the figures does provide a starting point and reveals striking differences. Table 1 illustrates the within-country geographical imbalances across the world for the countries for which there are data available.

Internal Geographical Imbalances

The contrasts are stark. On average, more than 80% of doctors work in urban areas, and the remaining 20% works in rural areas. The figures are more favorable for nurses, midwives, and medical assistants, of whom approximately 40% work in rural areas. The distribution is more skewed for dentists, pharmacists, and radiographers, of whom 18%, 12%, and 18%, respectively, work in rural areas. This implies that urban areas count, on average, 15 times more physicians, 6 times more nurses, and 3 times more midwives and medical assistants for the countries in the dataset. This ratio is higher for radiographers, dentists, and pharmacists, who are typically employed in hospitals or in the private sector. With more than 45% of people living in rural areas worldwide, the overall distribution is highly skewed in favor of urban areas.

A number of shortcomings to the data limit the inference that can be drawn from these figures. First, the data are available for only a relatively small sample of countries, with sub-Sahara African countries very well represented but other continents heavily underrepresented, as is clear from Table 2.

Internal Geographical Imbalances

The data also suffer from a number of biases. Countries with a weak administration, ill-functioning government, or in conflict are largely missing from the data; they are also likely to have higher concentrations of health professionals in urban areas. The same applies to regions within countries: areas with weak governance are more likely to have missing data. A second bias stems from the lack of data on private sector health workers as the figures only reflect public sector health professionals. Both types of bias will lead to underestimation of health professionals in urban areas, and thus an underreporting of the problem.

Studies at the regional level paint a similar picture, confirming the general pattern and also highlighting divergences between regions. A recent study on sub-Sahara Africa, where the problem is deemed most striking due to the relative high proportion of the population living in rural areas, illustrates this. The results summarized in Figure 1 show the concentration of doctors in urban areas for 13 countries. Densities are considerably higher in urban areas, a pattern that is confirmed by other country-specific studies. In Cote d’Ivoire, for example, 70% of all doctors work in the southern, urban regions that harbor only 40% of the population, and similar disparities are seen in data from Zambia, Sudan, and Uganda.

Internal Geographical Imbalances

In Asia, the case of Thailand has been well researched. Several studies provide updated estimates of the geographic distribution of the country, illustrating that Bangkok has four times more nurses per 10 000 people than the North East, the most rural region. A similar picture emerges for Bangladesh, where 30% of nurses are located in four metropolitan districts that represent 15% of the population. An early study confirms the problem of urban–rural imbalances for Indonesia. China provides another interesting example because the majority of its nurses (98%) and doctors (67%) have been educated only up to junior college or secondary education. This provides a unique setting in which the level of education for health professionals, often believed to be a major explanatory factor for reluctance to work in rural areas, is relatively low, and because this has resulted in having more doctors than nurses. Still, urban China has more than twice as many doctors and more than three times as many nurses per capita than rural China.

Evidence for Latin America also shows similar patterns, with health workers concentrated in the capitals and more affluent areas. In Argentina, which has one of the highest numbers of health workers per person, Buenos Aires counts seven times more doctors per capita than Formosa or Misiones. In Chile, approximately 60% of public sector health professional are concentrated in the region of Santiago, which hosts only 40% of the population. In Ecuador although the capital’s province of Pichincha has 2 doctors per 1000 inhabitants, the more remote provinces of Galapagos and Orellana Esta have 0.56 and 0.43, respectively. In Guatemala, which has 0.9 doctors per 1000 inhabitants on average, 71% of health workers are concentrated in the metropolitan zone, whereas the more remote Quiche has a ratio of less than 0.1 doctors per 1000 inhabitants. A similar imbalance exists for all medical professions in this country. Nicaragua has 0.4 doctors per 1000 inhabitants on average, whereas its capital, Managua, has 1.1. In Peru, 53% of physicians, 40% of nurses, 44% of dentists, and 41% of technicians and health assistants were concentrated in the Lima Metropolitan Area, which represents close to one-third of the country’s population. Stark differences in densities across professional groups are also reported for the Dominican Republic, with 0.41 nurses per doctor in Region 1 and 3.63 in Region III. A detailed analysis of Brazil also shows substantial inequalities in numbers of health workers.

Although this article focuses on low-and middle-income countries, it is useful to remember that the same problem exists in high-income countries. In USA, for instance, 9% of physicians work in rural areas, which represent 20% of the population. The figures for Canada are very similar (9% of physicians for 24% of the population). In France, the wealthy areas of Paris city and the South have considerably more doctors than the rest of the country. In Norway, the rural and remote Northern areas have historically been underserved, and there is long tradition of policy making to try to address this.

The pattern that emerges from the above descriptive statistics is clear, but also blunt. For a subset of countries with more detailed data, a more advanced analysis and decomposition of inequalities across and within subgroups is possible – for example, according to profession or gender. A 2008 study considers the Gini coefficient, Theil T, and Theil L measures for the distribution of health workers in China, and decomposes overall inequality in between and within province inequality. The findings indicate that underlying distributions can be very different across regions. Across the measures, within province inequality accounts for between 82% and 84% of intercounty inequality. A later study draws similar conclusions for sub-Sahara Africa, calculating the Concentration Index and Gini coefficient for doctors and nurses for nine sub-Sahara African countries. Although the Concentration Index for doctors varies between 0.25 and 0.48 (for Kenya and Senegal, respectively), that for nurses varies from 0.05 to 0.54 (for Kenya and Mauritania, respectively), and generally confirms that imbalances tend to be more severe for more educated health professionals. The results for both China and Africa illustrate how aggregate figures on numbers of health workers can be misleading and may provide a highly insufficient base for policy making.

Imbalances In The Quality Of Care

With the development strategy of the past decades emphasizing increases in the supply of care, much of the debate has focused on gaps in the quantities of health-care providers. But there are also important imbalances in the quality of care that patients receive (often referred to as process quality), which is a function of the number of health workers, their performance, and the availability of complementary inputs. This chapter focus on performance differences stemming from human resources and abstracts from differences in the import of complementary inputs like the clinic’s physical condition or the availability of drugs (often referred to as structural quality).

The significance of health worker performance – or better, underperformance – is perhaps best illustrated by the results of surprise visits to health facilities in six developing countries that found 35% of health workers to be absent on average. Although the study does not set out to compare between rural and urban areas, it finds that absenteeism is generally higher in poorer areas and among higher qualified health professionals (e.g., doctors). Results from qualitative studies in Ethiopia, Rwanda, and Ghana suggest that absenteeism is higher in rural areas mainly because of poor monitoring. Other work illustrates the importance of on-the-job performance. High health-care usage rates, combined with poor health outcomes often indicate problems with quality of care. A 2007 study provides direct evidence for underperformance of doctors in rural Tanzania. Measuring what doctors know (using a vignette) and comparing this with what they do (using direct observation as well as patient recall), the authors observe a substantial know–do gap. In other words, these doctors provide lower quality care than what they could provide. Qualitative studies in Ethiopia, Rwanda, and Ghana also find indications that health worker attitudes toward patients tend to be poorer in rural areas, whereas performance problems like corruption and embezzlement seem to be higher. This is supported by studies of corruption in the health sector in Tanzania. Further underlining the importance of taking quality of care into account, other studies find that households in Tanzania bypass low-quality facilities that are nearer and increase travel time to reach facilities with better care. Also relevant are study results on medical quality in urban and rural areas in five countries that find that households in poor areas not only have more access to private facilities that provide low-quality care, but are also more likely to receive low-quality care in any facility, particularly in the private sector. The inquiry also finds that indigenous patients that come from a poorer background receive less quality care in the private sector, and infers that this is due discrimination against those patients (rather than households choosing low-quality providers). A separate study shows how workload is not the reason for poor performance among health workers in Tanzania, observing that clinicians have ample amounts of idle time. The authors conclude that scaling up the number of health workers is unlikely to raise the quality of health care. Taken together, these study results provide strong evidence for quantitative and qualitative imbalances between urban and rural areas.

Implications Of Imbalances For Health Outcomes

A number of studies have looked at the implications of quantitative imbalances. Evidence from cross-country regressions suggests that the number of health workers has a strong relationship with health outcomes. Controlling for GNI, income poverty and female adult literacy, it has been found that HRH density is strongly related with especially maternal mortality, but also infant and under-five mortality. Decomposing the effect for doctors and nurses, there is a large association for the former and absence of such an association for the latter – except for maternal mortality where nurses do seem to play a role. Other work, looking at the disease burden, also tends to find a (negative) relationship. One study of the relationship between different health worker densities and DALY’s (and DALY’s disaggregated according to three different groups) finds a strong relationship with the number of doctors in particular, whereas the association for nurses and midwives is insignificant. A similar analysis argues that countries with fewer than 2.5 health workers per 1000 population are very unlikely to achieve a desirable 80% level of coverage for skilled birth attendance and measles immunization. A further inquiry, updating these studies by making use of an extended sample of 192 (instead of 177) countries, finds an aggregate relationship between health worker density and measles immunization and birth attendance, but no longer with infant and under-five mortality. It also observes a significant association for doctors but not for nurses, concluding that that the threshold should be 2.28 rather than 2.5 health workers per 1000 population.

Although the findings from these studies are indicative, they do not provide conclusive evidence for a causal relationship, as the analysis may suffer from omitted variable bias. The number of health workers may, for instance, be correlated with government expenditures on health, donor activity, the number of clinics, the availability of equipment and medicine, or the presence of conflict, none of which are included in the analysis. Other factors, such as skill mix, negative work environment, and weak knowledge base may also be important, and their omission may further bias the estimates upwards. Conversely, the lack of a relationship between nurses and health outcomes may also be due to unobserved factors – like absenteeism among health workers – which may bias the estimates downwards. Another potential problem is that the sample suffers from selectivity. Countries with good data tend to be better organized and may have surmounted other constraints that may matter more than the number of health workers. There is, finally, also a question as to how the limitations in data comparability across countries play a role. Different countries use distinct definitions, for instance for nurses, and this introduces both measurement error and unobserved heterogeneity. More sophisticated approaches are needed if one would like to test the robustness of these findings, as recognized by the authors of some of these studies.

To address some of the shortcomings associated with using aggregate cross-country data, more recent studies focus on within country variation using subnational data. One such analysis of China concludes that the density of doctors and nurses is significant in explaining differences in infant mortality across counties in China. A similar approach to data from Brazil finds that a 1% increase in health worker density is associated with a 0.12% increase in the coverage of antenatal care on average. The papers also illustrate that there is considerable variation in the level of coverage by municipality for a given number of health workers, thereby illustrating how the analysis suffers from similar shortcomings as the ones mentioned above. As a result, they do not provide conclusive evidence on the extent to which shortages of health workers cause poor health outcomes. Although identification of causality remains a challenge, it is necessary when informing policy making, especially when providing advice on target numbers of health workers.

An analysis of data from Ghana yield evidence on a causal relationship between the number of health workers and demand and usage of health care. Making use of exogenous policy changes in the late-1980s, it is found that increasing the number of doctors and nurses to three (representing a 50% increase from the mean) would lead to a 20% increase in the predicted probability of households choosing public health care. A recent study focusing on Indonesia also provides causal evidence that relates the number of health workers and quality of health care to health outcomes. Exploiting the fact that deployment of health staff in Indonesia is based on quantitative targets per facility although not related to quality or health outcome targets, it was found that increasing the number of MDs, nurses, and midwives increases adherence to clinical protocol, which in turn leads to improved child health (measured by length). The largest gains are made by increasing the number of MDs, followed by nurses, whereas increasing the number of midwives had no effect. As the study did not include the most remote areas in Indonesia, its estimates may well be conservative.

Recent evidence for Kenya shows how absenteeism causes poorer health outcomes. Using longitudinal data for rural health clinics, it is shown how women whose first clinic visit coincides with nurse attendance are approximately 60 percentage points more likely to be tested for HIV and 13% more likely to deliver in a hospital or health center, and how this in turn affects expected HIV status. The presence of other health workers may also increase the quality of care, as shown by one study interpreting Hawthorn effects in direct observation of doctor activity as evidence that performance increases when colleagues are present.

The above-mentioned literature confirms the causal relationship between quantities of health workers and quality of care on the one side and health outcomes on the other, but does not allow clear conclusions to be drawn on the relative importance of these factors. To identify pathways through which rural health outcomes can be improved, the next step is then to return to theory to better understand why quality of care and numbers of health workers are lower in rural areas and result in lower health outcomes.

Causes Of Imbalances In Health Care: Theory And Evidence

From a theoretical perspective, there are a limited number of reasons why health outcomes in rural areas are lower. Abstracting from potential differences in disease burden, three factors play a role: poor infrastructure – including scarcity of clinics, lack of equipment, medicine, etc.; weak human resources which relates to the number of health workers, their presence and performance, as well as the combination of health worker types; and, finally, limited demand for health care, which is related to households’ information and health seeking behavior. There is currently limited understanding of the relative role of these factors, and where the binding constraints lie. This relationship can be presented in a more systematic way, by considering patient health outcomes (H) as a function of the infrastructure of the facilities in that area (K), the human resources in the facility (L), which entails number and different types of health workers (n), their presence (p) and performance (y), as well as patient household characteristics in that area (hh). It is helpful to think of this relationship in terms of a production function where the inputs are imperfect substitutes, and write health outcomes as a product of these three factors, with their power reflecting their relative weight.

Health worker inputs (L) can thus be seen as the product of three factors: the number, presence, and performance of the respective health worker categories. A next step is to consider the determinants of these respective factors. This chapter focuses on two issues central to human resources: the quantity of health workers in rural areas (n) and their performance (y). The paper refers to other work for in-depth discussion of absenteeism and issues not related to human resources, including infrastructure, availability of drugs, funding, and factors to do with demand.

Quantity Of Health Workers (N)

Ultimately, the relatively low numbers of health professionals in rural areas is rooted in the choice of health workers themselves. Job choice is typically modeled as a process of matching between job attributes and preference. Focusing on earnings and effort, in addition to other job attributes like social status, recent work adds motivation, which is especially relevant for professions where a personal mission is important, like in public service. Considering that health workers will choose to work in a rural area when they expect to derive more utility from a rural than an urban job, this framework predicts that, since earnings and amenities typically receive high weights, while differences in effort between rural and urban areas may be limited (even if weights to effort may be high), most health workers prefer an urban job. Only those with a mission that matches to working in rural areas, or those who attach a high value (weight) to living or working in a rural area, for instance because of proximity to family and friends, prefer a rural post. These predictions immediately illustrate the limited leverage that policy makers have at their disposal if they want to get more health workers into rural areas. Although in theory people can be compensated for unattractive job attributes, for most health workers, earnings will have to be very high to compensate for the disutility caused by poor amenities in rural areas.

This situation may be aggravated when taking a more dynamic perspective and consider health workers to be making a career rather than a job choice. Taking a lifetime perspective, the outcome is now determined by the discounted sum of utilities across different periods, allowing for health workers to change from rural to urban areas, with income in each period a function of human capital accumulated in the previous periods (ht-1). If an individual expects that the accumulation of human capital is slower in rural posts, for example, due to lower opportunities for formal training or because the type of experience built is not rewarded in urban jobs, she may be even less likely to choose a rural post. In a more sophisticated approach, valuations of job attributes could be allowed to vary as staff gets older. The weight attached to amenities may change and health workers may stick higher values to jobs in urban areas at certain ages, for example, at marriage age because it offers access to a larger pool of potential marriage partners, at child-bearing age because it offers access to better child care, or when children reach school going age because of the proximity of better schools, etc.

The basic predictions of the above models are supported by empirical evidence. Results from comparative qualitative research illustrate why health workers in Ethiopia, Ghana, and Rwanda generally prefer jobs in urban areas. Although rural jobs may offer extra payment and benefits, these are usually insufficient to compensate for other disadvantages. Professional isolation, limited access to training, and poor working conditions characterized by limited access to equipment and infrastructure are seen as strong drawbacks. Urban postings also provide the possibility of working in a second job in the private sector, which is usually absent in rural areas. But the reasons why rural posts are unattractive go beyond job attributes, as factors like personal isolation, the general absence of infrastructure and amenities, including the low quality of housing and absence of good schools, also play an important role. Rural postings are associated with lower career perspectives as well, as they provide less access to training, limited access to equipment and modern technology, and thinner professional networks, among others. In some rural areas, salaries are often paid with delay. However, the lack of supervision from colleagues may give more freedom in rural posts.

A growing body of quantitative work analyses health worker willingness to work in rural areas, typically studying the role of wages and other job attributes. In the absence of incentive compatible study set ups, two types of methods have been applied contingent valuation and discrete choice methods. Each of these methods have their advantages and draw backs. While contingent valuation methods find the precise reservation wage to work in rural areas, discrete choice methods focus on trade-offs between different sets of attributes. A 1998 investigation of health worker willingness to work in remote areas in Indonesia uses the first method to find that modest cash incentives can make health workers more likely to work in moderately remote areas, but that it would be prohibitively expensive for staffing of very remote areas. Health workers who grew up in remote areas are found to require lower compensation to take up a remote position. Results from a cohort study with final year health students in Ethiopia also find that expected wages affect take up of a rural post. Here, in order to get 80% of health workers in rural areas (who harbor 80% of the population), salaries would need to increase with 83% for doctors and 57% for nurses, requiring an increase in annual health expenditures of 0.9%. The study also observes substantial heterogeneity, with health professionals who grew up in more remote areas, come from a less wealthy background, or are more motivated to help the poor being more willing to work in rural areas. Assessing what other job attributes matter, the study finds that chances for promotion, access to professional training and access to schools for education of children turn out to be important. Other studies present very similar findings. A resurvey of the same Ethiopian health professionals 2 years later (when they had entered the labor market), finds that wages and other job attributes are only part of the story and that health worker characteristics like rural background and motivation play an important role, with the latter influenced by the type of school attended. An identical study with health students in Rwanda confirms these results with rural background and motivation to help the poor as important determinants of willingness to work in rural areas. Health workers who were participating in a local (church-based) bonding scheme were also more willing to work in remote areas. Another similar study in Ghana finds that doctors who grew up in a rural area, as well as those with higher motivation are more willing to work in rural areas.

Results from discrete choice studies provide further insights into the relative roles of job attributes. Focusing on doctors in Ethiopia, it has been found that doubling wages would increase the share of doctors willing to work in rural areas from 7% to more than 50%, whereas providing high-quality housing would increase it to 27% (the equivalent of a wage bonus of 46%). For nurses, doubling the salary would increase their number from 4% to 27%, whereas the nonwage attribute that is most effective in inducing take up of a rural post is the quality and availability of equipment and drugs, which would reach the same result as a salary increase of 57% for men and 69% for women. Focusing on Tanzania, another study finds that offering continuing education after a certain period of service, as well as increasing salaries and hardship allowances, would encourage health workers to work in rural areas. Decent housing and good infrastructure were also found to be important. Women were found to be less responsive to financial incentives and more concerned with factors that directly allow them to do a good job, whereas those with parents living in a remote rural area are generally less responsive to the proposed policies. When willingness to help others is a strong motivating factor, policies that improve conditions for assisting patients are effective. Analyses of similar discrete choice experiments with health students in Kenya, Thailand, and South Africa, underline that results can strongly differ between countries. Financial incentives are likely to have important effects, especially in poorer countries, but only if they are larger than a 10% salary increase as smaller raises were found to be ineffective in all three countries. Nonfinancial incentives are found to be important as well, especially access to training and career development opportunities. Improved housing and accelerated promotion were moderately effective. A study using propensity score matching also suggests that improving Clinical Officer’s access to upgrade training would not improve their retention in rural areas. A study of the situation in Liberia and Vietnam has found that although in Liberia increased pay would be the single most-powerful incentive, long-term education was the primary factor in Vietnam, and considers the differences in cost effectiveness of implementing corresponding policies. A recent study of Uganda sets out to design packages able to get medical and nursing officers in rural and remote areas using discrete choice methods. The preferred package for medical officers is a 100% increase in salary (from a current base salary of 750 000 Uganda Shilling), improvements to health facility quality, a contractual commitment to the posting for 2 years, and full tution support for continued education at the end of the contractual commitment. For nursing officers, the most preferred package contains a 122% increase in salary, improvements to health facility quality, and improved support from health facility managers. These packages would get an estimated 82% of medical officers and 90% of nursing officers in remote areas. Other studies that do not focus directly on the rural–urban choice also shed light on the importance of job attributes. Evidence for Malawi showed that graduate nurses valued high pay, as well as the provision of housing and the opportunity to upgrade their qualifications quickly. In South Africa, earning more was most attractive, whereas better facility management and equipment were next. Nurses in rural areas were more concerned about facility management.

Both quantitative and qualitative evidence indicate that there is important heterogeneity in health workers’ willingness to work in rural areas. Although the majority of health workers prefer not to work in rural areas, some do, in particular in provincial towns. Rural background in particular has been found to be strongly positively associated with willingness to work in rural areas in Indonesia, Ethiopia, Rwanda, and Thailand, among others. Higher-level health workers (e.g., doctors) are generally less willing to work in rural areas compared to lower-level ones (e.g., clinical officers or nurses), for example, in Ethiopia and Uganda. Female health workers are often less willing to work in rural areas, as shown by evidence from Congo and Ethiopia, often for security or marriage-related reasons. Younger health workers may also be more likely to take up a rural post as part of their training, although their willingness may fall rapidly when entering the labor force, as found for Ethiopia. A number of studies also observe heterogeneity in health worker motivation, with health workers who are more motivated to help the poor more likely to take up a rural post. Identical surveys among medical and nursing students in Ethiopia and Rwanda both find that helping the poor is an important explanatory factor for willingness to work in a rural job for a substantial minority of health workers. This result for intrinsic motivation is strikingly similar for the two countries, and indicates that some health workers prefer to work in rural areas because this provides for a better match between their own beliefs and the belief of the facility they work for. Recent work also finds evidence for mission matching in nonprofit organizations in Ethiopia. The higher motivation to work in rural areas in Ethiopia is also linked to the school where one was trained, with health workers trained at an NGO school more willing to work in a rural area. This suggests that either health workers get socialized into motivation, or that they self-select at an earlier stage and choose the school that matches their beliefs and motivations. Overall, this evidence underlines that certain types of health workers self-select into rural jobs.

Qualitative studies also suggests that other factors, like appreciation for a slower pace of life, may play a role, indicating that adverse selection may be important. A recent test of whether less skilled health workers – as measured by a medical knowledge test – are less likely to work in rural areas in Ethiopia finds no evidence for such adverse selection. Exploiting the existence of a lottery for allocating doctors to jobs, however, another study finds that adverse selection may occur in a different way, with lottery participants, who are not able to use their first job as a signal of ability, having flat wage profiles and higher exit rates.

Although rigorous studies using revealed (rather than stated) preference and identifying causal effects are currently absent in this area, the above evidence provides a base for an increased understanding of the labor market for health workers in low and middle-income countries. It also points already to three types of policies: those working on the demand side using wages and job attributes, those operating on the supply side focusing on certain profiles of health workers (such as those with a rural background), and policies considering matching of demand and supply and allocation of health workers to posts (or vice versa). The section Lessons for Policy Making and Ways Forward will discuss these policy options in more depth.

Performance Of Health Workers (Y)

The performance of health workers can best be understood in a classic principal agent framework, where it can be seen as a function of three factors: incentives (w), monitoring (s), and individual motivation (m). Like before, incentives are used in a broad sense, and monitoring includes both supervision and accountability to the local community; it can also include workplace, professional, and society norms regarding professional behavior.

Qualitative research suggests that performance problems of health workers may be more important in rural and remote areas taking the form of absenteeism, poor attitudes toward patients, engagement in corruption and embezzlement, or poor performance in general. Moral hazard seems mostly attributed to four factors: the perceived lack of compensation for personal and professional sacrifice; poor monitoring and enforcement; a culture of poor performance with weak norms; and lacking motivation. The public sector in general is associated with more corrupt practices, and in a number of places, a culture of corruption and free riding is deeply embedded in the public health sector.

Quantitative evidence on determinants of performance is scarcer. One study, using data for Tanzania, provides a good starting point, comparing performance of doctors in the public and private not-for-profit sectors. Like in much of the rest of Africa, these two sectors share a similar mission, often run similar health facilities, and many not-for-profit facilities also follow public sector salary scales. It was found that clinicians in the not-for-profit sector have almost exactly the same average competence as clinicians in the public sector, but their adherence to the prescribed script, an indicator of quality of care, is higher. Thus, although the not-for-profit sector hires clinicians with the same capacity as the public sector, clinicians in the not-for-profit sector perform better.

In other settings, it is more relevant to compare health workers in the public with those in the private, for-profit sector. This approach was used in a 2007 study of Delhi. It was found that, on the whole, private sector providers spend substantially more time and effort on patients. Public sector providers also do less than they know they could. This sector disparity further masks variations in the public sector, with public providers in smaller clinics and dispensaries performing substantially poorer than public providers in hospitals, who tend to do comparable to private practitioners.

Although these studies generate relevant insights into differences in performance, they do not provide guidance for ways forward. Differences in the observed average know–do gap between sectors, sometimes seen as a measure of motivation, can arise for a variety of reasons, as variations between sectors are many (including pay, type of contract, monitoring, work environment, funding available, etc). They may, in addition, arise from the different types of workers that they employ.

There is also useful evidence of how performance can change. Using the above-mentioned data for Tanzania and exploiting the presence of a Hawthorne effect, it has been shown that being observed leads to higher effort and that there exists a link between variation in doctor performance and ability and motivation. An exploration of ways forward shows that clinician performance can be improved by peer encouragement as well as token gifts. Unconditional encouragement, where doctors are asked to do more, seems at least as useful, and has at least the same long-run impact, as conditional encouragement, where doctors are incentivized to do more, although little is known about the long-term effects in either case. Assistant Medical Officers in particular are able to do much better without significant additional resources and have sufficient capacity but insufficient motivation.

Studies identifying a causal effect of incentives, monitoring, or motivation on health worker performance remain scarce, but two studies stand out, one on pay for performance, and one on community monitoring. The remainder of this section discusses each in turn. The first study reports the results of a quasi-experiment in Rwanda where part of the funding received by health facilities depended on their performance. Results from qualitative research have illustrated many performance challenges in Rwanda. Both health workers and users point toward serious problems with health workers’ attitudes toward patients, which are often characterized by impolite and rude behavior. Absenteeism and shirking are common problems, and some public facilities have ‘ghost doctors’ who are on the pay roll but do not show up for work. Especially in urban areas, absenteeism seems mostly related to having a second job, usually in the health sector and often unofficial. In this context, linking part of the funding that facilities receive to their performance may have beneficial effects. Indeed, the study finds large and significant positive effects on deliveries and preventive care visits by young children and improved quality of prenatal care, but no effects on the number of prenatal care visits or on immunization rates. It was concluded that pay for performance had the greatest effect on services with the highest payment rates that required the least provider effort. Unfortunately, the study did not investigate how health worker behavior changed. Results from qualitative research shed some light, and suggests that performance pay decreased absenteeism as well as shirking and improved the work environment, to the general satisfaction of health workers.

The second study provides evidence from a randomized intervention on community-based monitoring of public primary health-care providers in Uganda. Making use of local NGOs to encourage communities to hold their local health providers accountable, the study finds that 1 year after the intervention, child mortality had seen significant reductions and child weight had increased in treatment communities. Studying the underlying processes, evidence was found for increased monitoring from the community through existing and new channels (e.g., local councils, evaluations) and increased activity from health workers, including improved consultation of the community (suggestion boxes), better information provision (posters regarding family planning), better management of patient care (numbered waiting cards), and higher medical effort (immunizations). The study also finds a drop in absenteeism, an increased use of equipment, a reduction in patient waiting times. A follow-up paper analyses the heterogeneity in some of these treatment effects and argues that the local social context, for example, income inequality and ethnic fractionalization, plays an important role and negatively affects the community’s drive to collective action, which in turn holds back improvements of service provision.

Lessons For Policy Making And Ways Forward

Although this field would benefit from more structured research that takes special care in identifying causality, the above studies have increased the understanding and suggest five possible ways forward to address geographical imbalances in both quality and quantity of care, focusing on human resources.

A first approach concentrates on the supply side of labor for health care, given demand. A second approach starts from the demand side, focusing on how facilities can acquire the human resources they want or need, given the available supply. A third approach looks at matching health workers and jobs and how health workers get allocated to jobs. Fourth, one can look at the coordination between public and private sectors. Finally, new directions can be explored to encourage self-help in rural populations. In what follows, each of these are discussed in turn. In a final section, the authors discuss how one can go about choosing between or combining these different alternatives.

Emphasizing Labor Supply

Most human resource policies in the health sector today focus on the supply side. Starting from a needs-based approach, they concentrate on how to attain the desired number of human resources in rural and remote areas. Although this approach may be justified in settings with extremely low numbers of health workers, it generally suffers from a number of problems. A key issue that remains unclear, for instance, is what the ideal number of health workers should be. As the precise causal relationship between the number of health workers and improved health outcomes is yet to be understood, current figures represent preliminary estimates. One underlying assumption is that existing human resources are fully used, whereas evidence indicates that this is not the case. This approach also abstracts from the quality of care, assuming that existing and new health workers all provide high-quality care, in contrast to existing evidence. It also tends to undervalue the potential role of modern technology (see section How to Choose the Appropriate Approach?). An exclusive focus on the quantities of health workers supplied thus misses several important points. As a result, policies grounded solely in this approach often disappoint and are not the silver bullet they are believed to be. A classic example is the training of more health workers, which is often presented as a promising strategy to address shortages in rural areas. However, if health workers are free to choose where they work, training more professionals does not necessarily lead to more health workers in rural areas.

An important step forward with supply-based policies lies, therefore, in recognizing the heterogeneity of health worker preference. Although health workers generally prefer to work in urban areas, a clear picture is emerging of the type of health worker that is more willing and likely to take up a remote post. Having grown up in a rural area, giving more importance to helping the poor, being lower rather than higher educated (e.g., nurse vs. doctor), and possibly being young rather than middle aged (little is known about elderly), all play a role. Polices that target these workers may therefore be more effective. Evidence from Indonesia shows that much can be gained from stimulating nurses to take up rural jobs, rather than doctors, who are considerably less likely to work in rural areas. Not surprisingly, these types of policies are becoming more common. Thailand, for instance, focuses on recruiting health workers with a rural background. This can further be combined with more rural-oriented training and education, as is the case in a number of high-income countries, including the US and Norway, who have built specific institutions to provide training for rural health care.

Where increasing the number of health workers is appropriate, a detailed cost–benefit analysis is needed to assess what would be the best approach. In many cases, it may be more cost effective to increase health workers in existing facilities, rather than building new facilities. Indeed, one recent study illustrates how patients can bypass facilities to get to the ones with better services. A higher number of health workers per facility also increases monitoring and reduces professional isolation. Other evidence has also shown how increasing the number of health workers per facility improves outcomes.

Besides focusing on the number of health workers, one alternative solution is to improve the quality of care. Improved training is often raised as an important way forward. There are clear examples where the curriculum does not capture local disease realities, particularly the disease burden of the poor and those in rural areas, leaving ample room for improvement. Mozambique (nonphysician) surgeons for instance, until recently received no training in HIV/AIDS, even though it was the most common disease they treated. Over the past years, many countries have updated their curriculums, with the Malawi approach that tailors content of the training to meeting the community’s most pressing needs, often as a model. Recent examples are South Africa’s Walter Sisulu University, community-based programs at Jimma University in Ethiopia and University of Gezira in Sudan; as well as initiatives at Makere University in Uganda and the National University of Rwanda, who are making epidemiology-based curriculum revisions. Recent evidence indicates that, on the whole, health workers know what to do; the failure lies in doing it. The role of training to improve this, for example, by ameliorating attitudes or shifting norms, remains unexplored.

An issue that is receiving increased attention both for quantity of human resources and quality of care is the role of intrinsic and altruistic motivations. A 2008 study finds that health workers in Ethiopia who attended a Catholic NGO school are more willing to work in a rural area. Similarly, evidence for Tanzania shows that motivations matter for performance. Neither study, however, can distinguish whether this is an issue of selection or socialization. Are health workers’ motivations set at the time they enter the profession, or does their training and professional environment shape their motivation? In the latter case, there would be a role for training institutions shaping motivations to improve the quality of care (and possibly willingness to work in rural areas).

Demand-Side Policies

Policies focusing on the demand side try to attract more of the existing work force to and improve health care in rural facilities, taking labor supply as given. Like many other policies, human resources have seen a shift from a manpower and central planning approach to a market-based approach. In most countries, compulsory placement has been abolished. Demand-side policies that focus on human resources quantities usually start – implicitly or explicitly – from compensating differentials theory, which argues that undesired job attributes need to be offset by attractive job attributes. Although individual preference play a role for the precise level that is considered acceptable for a job attribute, there seems agreement as to whether an attribute is desirable or not and a pretty clear picture is emerging as to what health workers want in their job. The studies reviewed in the section Causes of Imbalances in Health Care: Theory and Evidence indicate that raises in rural salaries do increase health workers’ willingness to work in rural areas, but also that increasing salaries is not enough. In most low-income countries, the discrepancy in amenities between rural and urban areas may be too large to be compensated by salaries alone. Moreover, government budgets are tight and policy makers are nervous about creating precedents for salary increases among public servants, especially in highly unionized environments. Providing other benefits like housing, transportation, and especially access to training and promotion may bring some relief. Giving more certainty about future career opportunities might also help, as concerns about unsteady future postings, together with a fear for professional isolation and lack of access to training seem to prohibit health workers from planning their career and makes rural postings less attractive. It also seems to affect job satisfaction.

Alternative approaches on the demand side are to improve efficiency and to maximize the effort and quality of care provided by existing personnel. This may be done by adapting the type of contracts offered. Although evidence for the health sector remains scarce, it has been indicated that tying pay to performance can have major impacts. Studies outside the health sector also provide evidence. Qualitative research in Ghana, which implemented a pay for performance scheme, suggests that concerns that performance pay erodes intrinsic motivation and attract the ‘wrong type’ of health workers seem unwarranted. A deeper concern, namely, that linking individual pay and performance may skew health worker behavior along the dimensions by which performance is measured (which is imperfect and may be arbitrary) remains largely unaddressed. Recent approaches, like the one in Rwanda, try to address this by evaluating performance at the facility level and letting it determine the budget allocated to each facility (which the facility was free to use however it wanted).

Alternative changes in contract consist of increasing monitoring and accountability of health workers, both of which tend to be weak in remote areas. This can have two effects. First, it may affect the amount of effort and quality of care delivered. The Uganda research discussed in the section Performance of Health Workers shows how improved accountability and community monitoring can ameliorate quantity and quality of care. A second concern is whether current conditions induce adverse selection, attracting health workers with undesirable attitudes, for instance the less skilled, into rural posts. This has been investigated using test scores as an indicator for the potential quality of care, but no evidence has been found that nurses and doctors with lower test scores self-select into rural jobs. It may, of course, still be possible that health workers who are less willing to apply their skills self-select into rural posts.

Matching Health Workers And Jobs: Allocation Schemes

In most countries, the allocation of health workers to jobs happens on a voluntary basis, with health workers choosing freely what job to take. But alternative allocation mechanisms exist. One example is the use of a draft or lottery. The use of lotteries in public employment has mostly been abolished (although is still present in military draft), but remains operational in some countries, including Ethiopia where, until recently, a national lottery was used to allocate health workers to jobs. Although participation in the lottery was initially compulsory, this could no longer be enforced and an opt-out has been allowed since the early-2000s (though there is still an expectation to work for a fixed period in the public sector). Allocation by lottery has been shown to be inefficient, resulting in adverse selection with the best personnel opting out of the lottery.

Other types of compulsory placement, even if limited in time, often suffer from similar problems. An example is provided by ‘bonding schemes,’ where health workers are expected to work in a remote area for a fixed number of years, for example, 2 years, often as a way to repay their studies. Although most countries have moved away from coercive schemes, bonding schemes remain popular, and are usually organized by state or by private institutions, often religious organizations. They suffer from similar risks as other coercive schemes, including adverse selection, erosion of motivation, and low performance. Bonding approaches have been tried and tested but have not led to the success that was hoped for. This probably explains why most countries have moved away from this, or if not, have moved to a long-term contract where compulsory rural service is limited in time and compensated by access to additional training.

Policies focusing on matching would benefit from a deeper understanding of health workers’ job decision process in developing countries. A US-based analysis provides an excellent example. Having observed that the majority of health students in the US base their job choice on their internship experience, taking their first job at the facility where they did their internship, researchers developed a two-sided matching model to optimize the allocation. Although this approach may be technically demanding, much can be learned from this type of designed matching mechanism.

Combining Public And Private Sectors

The ultimate objective of health policies is to improve people’s health outcomes. A growing literature highlights the complementary role of public, private not-for-profit and private for-profit sectors to reach this objective. Studies on rural–urban imbalances in human resources often abstract from private sector activity. Perhaps the inclination of policy makers to emphasize health-care delivery through the public sector plays a role. Another reason may be that private, for-profit facilities are mostly absent in rural areas. However, not-for-profit organizations tend to be active, and, in many settings, concentrate on rural areas. The design of human resource policies that give more weight to health worker choice also require taking private sector activity into account more explicitly. Making the private sector part of the analysis also encourages bold and creative thinking about new ways to bring private health care to rural areas, moving beyond the dichotomous view that public sector’s main task is to correct the imbalances caused by the private sector (or lack thereof). Letting pharmacies and especially drugstores play a more important role is one example of how public–private cooperation can contribute. More analysis is needed that compares across sectors. Existing evidence for Tanzania shows that doctors in the private sector perform considerably better than their colleagues in the public sector. Health workers in the public and not-for-profit sectors had similar levels of knowledge, but the know–do gap was smaller among NGO workers. The know–do gap is found to be largest among public sector health workers, followed by private, for-profit professionals. Health professionals in the nonprofit sector in Ethiopia are found to be less skilled, but more motivated. These are some of the exciting findings from the scarce existing evidence. Future work in this area will generate new insights on strengths and weaknesses of the different sectors and lay the ground for creatively combined or complementary approaches. Developing a finer typology to move beyond the simple categorization of rural–urban could also bring this work forward, as it will show more clearly where private, for-profit sector activity is viable.

Encourage And Support Self-Help Among Rural Populations

Geographical imbalances in health care occur in all countries, regardless of whether they are low, middle, or high-income. This in itself suggests that they may be hard to solve. Moreover, among high-income countries, both those with more regulated labor markets (cf. Norway, France) as well as with weakly regulated labor markets (see section Encourage and Support Self-Help among Rural Populations) have imbalances, indicating that regulation of health worker labor markets might have limited impact. Policies have therefore typically focused on minimizing these imbalances, rather than eliminating them. The way forward seems to concentrate more on health outcomes, and make rural health care less dependent on the physical presence of health workers. The training of community health workers has been a tried model as a way to increase self-help by rural populations. Although there is no structured evaluation of these types of programs, existing overviews indicate that this is not the panacea it was once believed to be. Whereas the involvement of community health workers may help address needs for health care, including for infants and children, its scope remains limited. Past experience has also taught that there are many pitfalls for the implementation of such programs. A central concern is whether and how good quality of care can be guaranteed. Careful selection and training seem to be crucial. Another key for success seems to be whether the program is embedded not only in the community, but also in the health system. Initiatives that are implemented in parallel to the health system seem to be the least successful; whereas those integrated into the existing health system seem more effective. The Brazilian Family Health Program provides the largest and best known example of this approach. Although involving community health workers has potential, more structured evaluations are needed to increase the understanding of what works and why.

The renewed interest in community health work seems to lead to a new generation of community health work interventions. At least two potentially promising directions emerge. First, existing work suggests that community health care is more effective when built on existing institutions. An example is provided by a community-based approach grafted onto an existing network of women self-help groups making a substantial difference to maternal and infant health. Second, so far, little attention has been paid to how new technologies may further strengthen this approach. An impressive hands-on example is provided by the CARE foundation in India, where village health workers are equipped with a basic mini computer that can perform some basic tests, but also contains software with algorithms to support diagnosis and treatment. Moreover, the computer is connected via a mobile network to a doctor who can be consulted remotely by the village health worker; the doctor also monitors – and, if needed, intervenes – remotely. Although further work is needed to evaluate and explore these approaches, they open up promising avenues for future health care in rural areas.

How To Choose The Appropriate Approach?

None of the above approaches is a silver bullet, and in most cases the best way forward lies in a smart combination of approaches adapted to the local context and informed by past experience. There is currently limited understanding of the relative payoffs of these approaches to inform and guide tradeoffs between them. Budget constraints, often seen as a nuisance, may help focus minds and identify where returns are highest. This question seems particularly relevant in light of the rapid urbanization in developing countries. How to balance the strive for equal access to quality health care with the concern of investing in geographical areas that may soon contain even less people?

Like in other areas of policy making, there is no ‘one size fits all’ approach. Increasing overall numbers of health workers may, for instance, sound attractive where there is a general shortage and existing capacity is fully utilized, but it remains unclear whether it will improve rural health care (see discussion in the section Implications of Imbalances for Health Outcomes). And even if it does, an equally hard question is whether this the most effective way to improve health outcomes. Would the same funding bring about more changes when used in another way? The key question remains where government expenditures – and aid – are best spent.

A useful illustration of how one can make ex ante tradeoffs is provided by a study focusing on Ethiopia in the early-2000s. It describes two potential ways to increase service delivery in rural areas: building more health clinics or improving and extending the quality of health care in existing facilities. Using a simple model and applying it to household data on health-care usage in Ethiopia, the study argues that additional expenditures to improve the quality of care will most likely be more cost effective than building more clinics. The conclusion sits well with earlier reported results, which show that patients bypass ill-performing facilities, and also provides deeper meaning to the results on the Ghana study mentioned earlier. The strength of this approach seems twofold. First, by providing a simple model, one can test ex ante what would be the most effective approach. Second, designing a simple model helps to generate well-defined hypotheses that can be tested empirically and can also help select the best empirical strategy to address identification challenges (e.g., randomized control trials (RCT)).

Summary And Conclusion

This article discusses the commonly observed discrepancies in the quantities of health workers and the qualities of health care between rural and urban areas in developing countries. The key question is how to close the gap in order to improve rural health outcomes. There is little doubt that human resources matter. Studies providing causal evidence are scarce, but they confirm the importance of human resources, which affect both the quality of care and several health outcomes. This has often been interpreted as evidence that important health gains can be made by increasing the quantity of professional health personnel in rural areas. However, the understanding of the optimal number of health workers remains limited. Although a focus on numbers and shortages may be warranted in some situations, it is by no means the silver bullet it is often claimed to be. One reason is that one also observes substantial underutilization of existing human resources, both in urban and rural areas. A small but increasing number of studies have shown a real a gap between the knowledge and the practice of health workers. Quality of care thus emerges as a real concern and deserves more attention. Future work will clarify whether quantity or quality is a more important binding constraint, and under what conditions.

One key observation shows the limitations of a single focus on increasing numbers of health workers: health professionals prefer to work in urban areas. Although studies indicate that it is possible to attract more health workers to rural areas, exploiting health worker heterogeneity in preference, and making use of an appropriate mixture of supply, demand, and matching policies, the omnipresence of these imbalances in rich as well as poor countries suggests it is very unlikely that the gap between rural and urban areas can be closed. More creative approaches are needed. One way forward may lie in combining the different sectors – public, private, not-for-profit, and for-profit sectors, whose complentarity has been studied, but deserves more attention. Another way forward lies in the next generation of community health worker programs which are grafted on existing institutions, as well as applying new technologies. Undoubtedly, future work will pay more attention to comparing the cost effectiveness of different approaches. Here, RCT can help in building a better understanding, provided they are informed by theory and designed to reveal why some approaches work better than others.


  1. Anand, S., Fan, V., Zhang, J., et al. (2008). China’s human resources for health: Quantity, quality, and distribution. Lancet 372, 1774–1781.
  2. Banerjee, A., Deaton, A. and Duflo, E. (2004). Health care delivery in rural Rajasthan. Economic and Political Weekly 39(9), 944–949.
  3. Barber, S. L., Gertler, P. J. and Harimurti, P. (2007). The contribution of human resources to quality of care in Indonesia. Health Affairs 26(2), w367–w379.
  4. Basinga, P., Gertler, P. J., Binagwaho, A., et al. (2011). Effect on maternal and child health services in Rwanda of payment to primary health-care providers for performance: An impact evaluation. Lancet 377(9775), 1421–1428.
  5. Bjorkman, M. and Svensson, J. (2009). Power to the people: Evidence from a randomized experiment on community-based monitoring in Uganda. Quarterly Journal of Economics 124(2), 735–769.
  6. Brock, J. M., Leonard, K., Masatu, M. C. and Serneels, P. (2012). Health worker performance. In Soucat, A., Scheffler, R. and Ghebreyesus, T. A. (eds.) The labor market for health workers in Africa. A New Look at the Crisis. Washington, DC: World Bank.
  7. Collier, P., Dercon, S. and Mackinnon, J. (2002). Density versus quality in health care provision: Using household data to make budgetary choices in Ethiopia. World Bank Economic Review 16(3), 425–448.
  8. Das, J. and Gertler, P. J. (2007). Variations in practice quality in five low-income countries: A conceptual overview. Health Affairs (Milwood) 26, w296–w309.
  9. Das, J., Hammer, J. and Leonard, K. (2008). The quality of medical advice in low-income countries. Journal of Economic Perspective 22(2), 93–114.
  10. Dussault, G. and Franceschini, M. C. (2006). Not enough there, too many here: Understanding geographical imbalances in the distribution of the health workforce. Human Resources of Health 4, 12, doi:10.1186/1478-4491-4-12.
  11. Hanson, K. and Jack, W. (2010). Incentives could induce Ethiopian doctors and nurses to work in rural settings. Health Affairs (Millwood) 29(8), 1452–1460.
  12. Lavy, V. and Germain, J. (1995). Tradeoffs in cost, quality and accessibility in the utilization of health facilities: Insights from Ghana. In Shaw, R. P. and Ainsworth, M. (eds.) Financing health services through user fees and insurance: Lessons from sub-Saharan Africa. pp. 134–153. Washington, DC: The World Bank Publisher.
  13. Lehmann, U., Dieleman, M. and Martineau, T. (2008). Staffing remote rural areas in middle and low-income countries: A literature review of attraction and retention. BMC Health Services Research 2008(8), 19, doi:10.1186/1472-6963-8-19.
  14. Munga, M., Songstad, N. G., Blystad, A. and Mæstad, O. (2009). The decentralisation-centralisation dilemma: Recruitment and distribution of health workers in remote districts of Tanzania. BMC International Health and Human Rights 9(9), doi:10.1186/1472-698X-9-9.
  15. Serneels, P., Montalvo, J. G., Pettersson, G., et al. (2010). Who wants to work in a rural health post? The role of intrinsic motivation, rural background and faith based institutions in Rwanda and Ethiopia. Bulletin of the World Health Organization 88, 342–349.
  16. Serneels, P., Lindelow, M., Montalvo, J. G. and Barr, A. (2007). For public service or money: Understanding geographical imbalances in the health workforce. Health Policy and Planning 22, 128–138.
  17. Soucat, A., Scheffler, R. and Ghebreyesus T. A. (eds.) (2013) The labor market for health workers in Africa. A New Look at the Crisis. Washington, DC: World Bank.
  18. Sousa, A., Tandon, A., Dal Poz, M. R., Prasad, A. and Evans, D. B. (2006). Measuring the efficiency of human resources for health for attaining health outcomes across subsnational unit in Brazil. Back Ground Paper for World Health Report. Geneva: WHO.
  19. Speybroeck, N., Kinfu, Y., Dal Poz, M. R. and Evans, D. B. (2006). Reassessing the relationship between human resources for health, intervention coverage and health outcomes. Evidence and Information for Policy. Geneva: World Health Organization.
Economics of HIV/AIDS Transmission, Treatment, and Prevention
Nutrition, Health, and Economic Performance