This study showed that despite an innovative program that combined a flat fee for care for most people with an exemption system for indigents, there continued to be significant financial barriers in accessing health care in Karuzi. Even though access to health care was readily available for 70% of overall population, about 30% sought care elsewhere or stayed home. Why was this so?
1. Significant overall poverty
A key feature was that the national poverty line in rural areas was evaluated at 1031.73 BIF/person/week [
24], so that 88.0% (83.5-91.6) of the population in Karuzi lived below that threshold at the time of the survey, confirming a context of widespread generalized poverty. Cash available at the household level was limited and it was difficult to mobilise quickly in the face of unexpected expenses for an episode of illness.
Although most households in the province owned a piece of land, it was just enough for subsistence, not for profit. Assessments carried out at the national level revealed that the average size of land per household was less than 0.5 ha [
25], barely enough to feed household members, let alone generate income. The situation became worse during the crisis that affected the country during the past decade when many households had to sell their land [
25]. "Landless" was the main indigence criterion reported in the survey.
2. The exemption system: mixed results
The exemption system was found to protect only 8.6% (6.3-11.0) of the population. This was lower than the 15% of households identified by the Hcoms but larger than in other provinces in Burundi [
3] or in other countries where classical exemption mechanisms were in place [
26‐
28]. Exemption schemes, even those endorsed by policy and legislation, are rarely fully effective. Schemes aimed at targeting the poor with exemptions often miss the intended beneficiaries, those in greatest need [
29].
MSF studies have shown that classical exemption mechanisms in Mali, Sierra Leone or Haiti covered less than 2% of the population" [
30]. And Ridde [
31] noticed that, although exemption of payment for indigents was one of the core principles of the Bamako Initiative, in many settings, exemptions had not been implemented or had not been able to protect very many people in the community.
However, results similar to those of our study were found with equity funds in Cambodia [
29,
32]. Their design was close to the exemption system developed in Karuzi and included the existence of donor funding, the presence of a driving agent, clear separation of roles, and appropriate identification techniques. These studies reached conclusions similar to ours: while these mechanisms are superior to traditional waiver systems in terms of health services utilisation by the target group, studies reveal remaining barriers to access and indebtedness prevention.
However, existing studies tend to focus on hospital and health structure data while failing to provide information on non-users of the system. Our study therefore brings additional data on general population access. These data are crucial to be considered for national policy choices.
Our results showed that cardholders had better financial access to health care than non-cardholders. It was also encouraging that almost all cardholders benefited from free health care in public HCs.
Although most of the cardholders were in a worse socio-economic situation than the rest of the population, the exemption system had problems properly covering the poor as shown by 56% of the households who had at least one indigent criterion but did not get cards. One reason for this low coverage could be the time lag between the date of identification and the survey. This would reinforce the idea that exemption systems do not capture properly the dynamic dimension of poverty. At the same time, 5% of households did get cards despite not having met the criteria. As well, although Hcoms had identified 15% of households as indigent, our survey found only 8.6% of households in possession of a card. This difference could be explained by the following:
▪ Around 400 cards (3.5% of all the cards distributed) were never distributed to the identified households because the householders were absent during the distribution meetings and the Hcoms did not follow up.
▪ Initially, some health staff members were reluctant to accept the exemption system and confiscated cards when they considered that beneficiaries did not qualify as indigent. This situation improved with time thanks to the reporting of the Hcoms.
▪ Cards lost were often not renewed by the Hcoms.
These difficulties should have been handled by the Hcoms with updates to the list every two months. However, this would have implied constant and time-consuming re-assessment. Experience in this project showed that maintaining an accurate list was a complicated and imperfect process, very demanding both financially and in terms of human resources. Members of Hcoms were not paid and this may have affected their motivation.
Furthermore, the performance of the system was hampered by the difficulty in defining and interpreting indigence criteria in a context of generalised poverty. The feasibility and accuracy of distinguishing the poor from the non-poor to determine eligibility for exemptions, is fraught with problems [
28,
33]. The notion of indigence is complex and covers both poverty and social exclusion [
34]. Criteria might have been subjectively and arbitrarily interpreted even though the members of the health committees had been trained to recognise eligible households. "A major difficulty is to identify very poor people in a population in which the poverty is rife" [
33].
The exemption system was demanding in terms of other human and financial resources. In addition to the Hcom members' time, human resource inputs included significant time of two full-time staff members, financed by MSF. Significant financial costs arose from MSF's financial support to HCs for compensation of revenue lost due to exemptions, considered a key condition for fee exemption schemes to be pro-poor [
12]. This raises the question of whether such resources should be provided to a system which benefits a relatively small number of people, particularly in a context of poverty. Other authors argued that "A universal free healthcare approach is justified in all situations with widespread misery or when time does not allow individual assessment schemes to be implemented [...]. Alternatively, identifying people living in poverty (by proxy means testing) and targeting benefits to them could be more attractive than a universal approach if the proportion of poor people in the society is not overwhelming..." [
10]
3. Access to health care
Despite these difficulties there were some positive elements to the project. In 2003, for all of Burundi, the level of access to health care in areas where cost-recovery was implemented was 58% of the population [
3]. In the Karuzi project, access to health care was much better, as 70% of the population went for a consultation to the closest HC and received full medication. Other encouraging results were that consultations in the informal sector were few (6.4%) as compared to other African contexts [
30].
Although mechanisms set up in the project to improve access proved to have done so compared to other areas in Burundi where cost recovery schemes were implemented, they still revealed important limitations in their potential to increase the use of health services for the population. Almost 15% still had financial barriers to access the closest HC and 10% of the population remained excluded from health care mainly for financial reasons. These findings were in line with other surveys revealing that even low fees constituted an obstacle to patients' access in contexts of widespread poverty [
30,
35]. MSF's experience has also revealed that targeting strategies compared poorly to general exemption or those based on large categories (like women and children), such as those implemented in Burundi in 2006 (national free care policy for under-5s and pregnant women). In addition, in contexts where health fees were totally abolished for all patients, evidence has shown an increase in the use of health services and specific benefits for the poorest households [
35].
Limitations
There were a number of limitations to the study. MSF was well known in the province especially for the identification of indigents. Although the surveyors - MSF employees - clarified that they were not in charge of identifying indigents in the community, respondents might have answered questions in a way to maximise their benefit, for instance, by overestimating their expenses in hope of being included on the indigent lists. This factor is likely to have been limited given the very low level of expenses reported. If present, this bias would lead to an underestimation of real poverty levels rather than an overestimation, as expenses were used to assess the socio economic status of households. As well, confronted by a western medical organisation, the respondents might have underreported the use of traditional medicines or of the informal sector.
An additional concern may be the 90.7-day recall period for last illness episode, which may have reduced the accuracy of details recalled. However, potential recall bias was reduced since the illness of inquiry was the most recent during the recall period.