Are user fees a barrier to health care in poor countries?

The 1987 Bamako declaration promoted user or consultation fees for health care as a means to raise revenue and improve the quality of services. However, user fees may pose a barrier to access, and hence the key Sustainable Development Goal of Universal Health Coverage (UHC), for the global poor who typically have a high elasticity of demand for health services. The evidence has been mixed though on the impact of adding or removing user fees. A Cochrane review found that utilisation of services typically declined significantly with the introduction of fees and that quality was often found to improve with fees, but they also questioned the reliability of these studies due to a “high risk of bias”. Indeed, the evidence can be conflicting as to the effect of user fees on health service utilisation. Consider the following two studies from two similar countries: Malawi and Zambia.

The first looks at the effect the introduction and removal user fees had on health centre outpatient attendances, new diagnoses of malaria, and HIV diagnoses in a rural district of Malawi (which I should declare I authored!) Of 13 centres in Neno district, four were operated by the Christian Health Association of Malawi, of which one has always charged user fees and three introduced them in July 2013. The other centres were operated by the Ministry for Health and an NGO, Partners In Health, and did not charge fees. In July 2015, one centre removed user fees. These changes in charging status created a neat natural experiment. A plot of outpatient attendances shows what happened:

Figure 3

Even without modelling it is clear what happened – attendances dropped with the introduction of user fees and increased when they were removed. Similar changes were seen in new malaria and HIV diagnoses Further analysis also suggested patients weren’t moving between centres to avoid fees.

The second study, published this week, looks at a 2006 policy to remove user fees for publicly-funded health care facilities in rural districts across Zambia. The policy was instigated by the Zambian president as a step towards UHC, but was implemented haphazardly with funding not being completely in place and districts choosing to distribute the funding they received in different ways. Using data from a repeated cross-sectional health survey, the corresponding plot of the effects of the policy is:


Evident from this and reinforced by their synthetic control analysis, the policy did little to change the proportion of people seeking health care. The key impact of the policy was to reduce out of pocket expenditure as it seems people switched from using private providers to public providers. So why do the results of these studies, with seemingly similar ‘treatments’ in similar poor rural populations, differ so much?

In an earlier study of the Zambian policy it was found that outpatient attendances recorded in routine data – the same data used in the Malawi study above – there were large increases in use of public facilities when user fees were removed. The new study adds evidence though that this increase was a result of people switching from private to public providers. In Neno district, Malawi there are no private providers – only those in the study. Nevertheless, private providers also charge, so health care use in the face of fees was markedly higher in Zambia than Neno, Malawi. Perhaps there are relevant differences then in the populations under study.

Zambia, even in 2006, was much wealthier than Malawi in 2013. GDP per capita in comparable dollars was $1,030 in 2006 Zambia and $333 in 2013 Malawi. And Neno district is among the poorest in Malawi. The Malawi study population may be significantly poorer then than that in Zambia, and so have yet more elastic demand. Then again, Zambia is one of the most unequal countries in the world, its wealth generated from a boom in the copper price and other commodities. Its Gini coefficient is 57.5 as compared to Malawi’s 43.9. Thus, one may expect rural Zambians to perhaps be comparable to those in Malawi despite GDP differences. Unfortunately, there aren’t further statistics in the paper to compare the samples – and indeed no information on the relative prices of the user fees. And further, the Zambia paper does look at the poorest 50% of people separately and finds little difference in the treatment effect although there does appear to be large levels of heterogeneity in the estimated treatment effects between districts.

Differences in conclusions may also results from differences in data. For example, the Zambia study looked at changes in the reported use of formal health care among people who had an illness recently, whereas the Malawian study looked at outpatient attendances and diagnoses. Perhaps a difference could arise here such as reporting biases in the survey data.

It is not clear why results in the Zambian study differ from those in the Malawian one and indeed many others. It certainly shows the difficulty we have understanding the effect even small charges can have on access to care even as the quality of the evidence improves.


Sam Watson’s journal round-up for 9th May 2016

Every Monday our authors provide a round-up of some of the most recently published peer reviewed articles from the field. We don’t cover everything, or even what’s most important – just a few papers that have interested the author. Visit our Resources page for links to more journals or follow the HealthEconBot. If you’d like to write one of our weekly journal round-ups, get in touch.

The association between income and life expectancy in the United States, 2001-2014. JAMA [PubMedPublished 26th April 2016

The relationship between wealth, income, and health has been examined in many hundreds of studies. The relationship is complex and still not well understood. Individual material circumstances affect patterns of consumption and health behaviours; relative socioeconomic position modifies these effects through psychosocial stressors and social patterns of behaviour; environmental and other neighbourhood factors also have further influences. So, what sets this study apart? It uses data from 1.4 billion tax records from citizens of the United States (p-values are still reported however!). Adults aged 40-76 from 1999 to 2014 are included. The authors find a gap in life expectancy of 15 years for men and 10 years for women between the top and bottom 1% of the income distribution. This gap increased over time with gains to life expectancy in the top quintile of the income distribution larger than those in the bottom between 2001-2014. There were also significant differences within income quintiles between high and low income areas, which were also correlated with health behaviors. This study provides a fascinating and comprehensive overview of income and health in the US. However, it does not really further our understanding of how these various relationships work, despite some bold claims.

The effects of exposure to better neighborhoods on children: new evidence from the Moving to Opportunity experiment. American Economic Review [RePEcPublished April 2016

The previous study in this round-up showed that there was a wide variation in changes to life expectancy over time for individuals in the bottom income quintile by geographic area. Between 2001 and 2014 changes to life expectancy varied between an increase of four years to a decrease of two years. But, these correlations don’t really provide enough information to formulate policy; is it the individuals in the neighbourhood or the neighbourhood itself? One way of investigating this would be to randomly provide the opportunity to move from a high poverty to a low poverty area and investigate the outcomes. In the 1990s the Moving to Opportunity (MtO) experiment provided vouchers to families to move to better neighbourhoods. Previous investigations of the effects of these vouchers found that families who moved had better physical and mental health, subjective well being, and safety but did not find evidence of improvements to employment or earnings. This present study examines the outcomes for those who were children at the time of the MtO program. Circumstances in early life, even as early as birth, are hypothesised to be important to later life outcomes. People who were part of families who were assigned vouchers and who were aged under 13 at the time of the program had greater earnings in their 20s and were more likely to attend college. These effects were not replicated in those aged 13-18 at the time of the MtO program.

Responding to risk: circumcision, information, and HIV prevention. The Review of Economics and Statistics Published May 2016

Understanding how people respond to new information on risk is important to formulate effective health policy. If individuals perceive their risk of an adverse health outcome to be low they may engage in riskier behaviours and conversely for those with perceived high risk. The net effect of greater health education is therefore ambiguous. This study investigates the effect of the perception of the risk of contracting HIV on the practice of safe sex among males in Malawi. Male circumcision can reduce the risk of transmission of HIV by up to 60%, thus circumcised males are at lower risk than their uncircumcised counterparts. The study included both circumcised and uncircumcised males and provided information about HIV transmission and circumcision to a random set of the participants in the study. The study found that the low risk (circumcised) males did not engage in more risky sexual behaviour while the higher risk (uncircumcised) males both reduced the frequency of sex and increased their use of condoms. Extrapolating from these results, this study may suggest that an increase in information has an overall positive effect by reducing risk.