OBJECTIVE To examine household out-of-pocket expenditure on health care, particularly malaria treatment, in rural Burkina Faso.
METHOD Comprehensive analysis of out-of-pocket expenditure on health care through a descriptive analysis and a second, multivariate analysis using the Tobit model with emphasis on malaria, based on 800 urban and rural households in Nouna health district.
RESULTS Households will spend less on malaria, either in or outside the health facility, if given the choice to do so, because they feel confident to self-treat malaria. Seeking health care from a qualified health worker incurs more out-of-pocket expenditure than self-treatment and traditional healers, and if necessary, households sell off assets to offset the expenditure. More than 80% of household out-of-pocket expenditure is allocated to drugs.
CONCLUSION This has policy implications for malaria control and the Roll Back Malaria Initiative. Communities need to be educated on the risks of malaria complications and the potential risk of inappropriate diagnosis and treatment. Drug or health services pricing policy needs to create an incentive to use the health services. In the fight against malaria, building alliances between households, traditional healers and health workers is essential.
Like most African countries, Burkina Faso introduced user fees as a mode of financing government health services within the framework of the Bamako Initiative. This was in response to the severe problems in financing health services in most of sub-Saharan Africa. Government health budgets declined in real terms in response to macroeconomic problems at the time while demand for health services increased, partly because of population growth and successful social mobilization. The Bamako Initiative was announced at a meeting of African Ministers of Health, sponsored by the World Health Organization (WHO) and United Nations Children Fund (UNICEF) in 1987. Its goal is `universal accessibility to Primary Health Care (PHC). The attainment of this goal would be enhanced through a substantial decentralization of health decision making to the district level, community level management of PHC, user financing under community control and a realistic national drug policy and provision of basic essential drugs, leading to a self-sustaining PHC with emphasis on promoting the health of women and children (WHO 1988).'
User charges in government health services were considered feasible because when `free' government services were rationed or of poor quality, people paid substantial amounts for health care in the private sector, often for inappropriate treatment or medicines. User fees were intended to redirect private expenditure towards more effective government health services and appropriate treatment, as retained fees would ensure regular drug availability and better quality.
The impact of user fees is forcing some governments to rethink or to abolish the regime altogether. Uganda, for example, abolished them in March 2001 (Wendo 2001), although the experience so far has been that health facilities have been overwhelmed and shortage of drugs is now frequent (Kikonyogo 2001). Only in district, regional and national referral hospitals are patients given the option of paying and being served quickly or queuing for free services. The FY 2001 Foreign Operations Appropriations Bill allows the US to oppose any IMF, World Bank or regional development bank loan that calls for user fees or service charges from poor people for primary education or primary health care (US Congress 2000).
One of the effects of user fees, which has not attracted much attention, is the multiplier effect it has in worsening the disease burden of some illnesses, either by causing delays in treatment leading to complications, or by inappropriate self-diagnosis and treatment, which may increase drug resistance. Malaria is an excellent example: It can be treated in just 3 days, yet kills millions every year. While many factors are responsible, delays in diagnosis and treatment are the main reason for its high mortality and complications such as cerebral malaria, severe anaemia, jaundice, and renal failure. An estimated 80% of malaria-related deaths are caused by cerebral malaria (Kakkilaya 2001), hence early treatment is essential to successful control of malaria. Unfortunately patients postpone seeking care; partly because of the fear of paying user fees (Foster 1991) until such complications begin to manifest (Ebisawa et al. 1980). Secondly, to economize, patients may buy insufficient quantities of the drug, or share one dose between two or more people if more than one family member is sick (Foster 1991). This increasingly renders some largely affordable and effective drugs such as chloroquine (CQ) ineffective by building resistance. As a result some countries have had to abandon CQ as the first-line drug in favour of more expensive alternatives. Malawi, for example, replaced CQ with sulphadoxine-pyrimethamine (SP) in 1993 (Zoguereh & Delmont 2000), and Kenya is in the process to do the same (Shretta et al. 2000).
In this paper, we examine the burden of out-of-pocket expenditures on households in Nouna, Burkina Faso. To assess the impact of the likelihood of delaying treatment and possible drug resistance, we analyse the burden of out-of-pocket expenditure by disease or health condition, in relation to its components and to the strategies employed to offset the expenses. We describe the factors for differential burden among households with particular emphasis on malaria.
Data and sample
The data is based on a panel survey of 800 households in the district of Nouna, north-western Burkina Faso, conducted under the auspices of the Nouna Health Research Center as part of an ongoing project to evaluate health care interventions. The Demographic Surveillance System (DSS) in the area provided the sampling frame. Households were sampled in a two-stage cluster procedure, with each household having the same probability of being selected (Levy & Lemeshow 1999). At the first stage, we selected seven clusters in urban Nouna and 20 clusters in the 41 rural villages; at stage two, respondent households were selected in each cluster. The sample proportions of rural and urban households reflect their respective fractions in the DSS: We selected 480 of 4630 households in the 41 villages, and 320 of 2802 households in Nouna 320 (62% rural, 38% urban).
The survey questionnaire comprised of socioeconomic and morbidity modules. To capture seasonal variation, the morbidity module is used four times a year and the socioeconomic module twice a year. This paper draws on data collected in the first two survey rounds conducted during the dry (June) and rainy season (September) in 2000. Each adult responded to questions pertaining to him/herself and appropriate proxies were identified for children and individuals who could not answer for themselves. If a household member was absent, interviewers made three more attempts to see the adult in question.
The perception of illness rather than disease was used to analyse out-of-pocket expenditure in Burkina Faso for two reasons: first, it is the perception that determines whether an individual seeks self-treatment or any other type of health care provider (Coreil 1983). Secondly, the resulting expenditure depends to some extent on the perceived type and severity of illness. To allow the possibility of future comparison with other similar studies, the illnesses were translated and mapped according to the Global Burden of Disease Classification (Murray & Lopez 1996). We use the term `disease' to denote a labelled diagnostic category, not as one that has been clinically determined.
Aggregate incomes were measured based upon local prices and quantities of agricultural products (animal and crops), regular cash incomes (salaries and pensions in some cases) and cash transfers, as the population is predominantly farmers who produce for home consumption and a little surplus for sale. Incomes were measured in two steps: first by summing up the value of sold animal products, crop products, cash earnings and transfers to the households; secondly, by subtracting expenditure on seeds, bought animals and cash transfers out of the households. If households bought animals and did not sell them in the reference period, this was considered savings and excluded. If the sale of crop products was bigger than the harvest, then the value of the sold crops replaced the value of the harvest. The value of animals, crops and material assets (plough, carts) was computed separately, similar to a procedure used to estimate incomes of rural communities in Sierra Leone (Fabricant et al. 1999), where a high correlation with wealth proxies, such as people's ranking of the rich individuals and families in the community, was found.
We also present estimates based on the Tobit model for examining out-of-pocket expenditure on health care. We used the probit estimation to identify the role of malaria in treatment choice and the ordered probit estimation to determine the likelihood of an illness being severe. Not all out-of-pocket expenditure incurred was observed over the entire study period. Some episodes had not ended and for these, complete information was only available on independent variables (age, sex, and income) but not for the dependent variable (out-of-pocket expenditure on health care). We did know the minimum amount of out-of-pocket spent on health care. The Tobit model, the adjusted Tobit model and sample selection models (Scott Long 1997) apply to continuous but limited dependent variables such as out-of-pocket expenditure on health care used in econometric literature.
Tobit analysis is done separately for treatment within and outside health facilities using the following variables as independent: whether an illness is malaria or not, age, sex, value of material assets, value of agricultural produce, value of animals, payment arrangement, urban/rural and household size. Household income, which is the sum of individual incomes belonging to the household, did not affect the out-of-pocket expenditure and therefore was dropped from the analysis and replaced by individual components. All monetary values are in Burkina Faso currency CFA (1 USD=550 CFA, March 2001). Severity of the illness as perceived by the patients was considered endogenous and not used as a dependent variable.
The role of malaria on treatment choice and severity is noteworthy. Malaria does not determine treatment choice using the probit estimation method. But people in the urban area and those with a high income (determined by the value of animals) were more likely to seek care from the health facilities, indicating ease of geographical and monetary access. Using the ordered probit estimation, malaria was less severe than other illnesses (Z=–4.75, P < 0.0001) and illnesses reported by people in urban areas were less severe (Z=–4.27), indicating again the problem of geographical access to health facilities for timely treatment. Nouna town is served by the district hospital while the rural areas have few first-line health facilities.
Out-of-pocket expenditure by illness
We examined the out-of-pocket expenditure according to illness and expenditure components (Table 1). Hospitalization is expenditure for a hospital bed and services but excluding drugs; transport includes all forms of transport; stay is expenditure on upkeep while seeking care; consultation is consultation fee, and others refers to any other expenditure including laboratory examinations.
Table 1. Components of household out-of-pocket expenditure by disease, Nouna, Burkina Faso
Ranking of illness and out-of-pocket expenditure on health care provides insights into how differently the morbidity and economic burden attributable to different illnesses impact on the population. For example, a highly ranked illness caused by morbidity may be the one with the greatest out-of-pocket expenditure, which is the case with malaria (Table 1). Relative rankings show differential impact; oral conditions, for example, ranks 8th as a cause of morbidity, but 13th as a cause of out-of-pocket expenditure, implying that the population with this illness experienced a significantly lower burden because of out-of-pocket expenditure relative to morbidity.
Comparing malaria and other illnesses, Figure 1 shows the components of out-of-pocket expenditure on health care. Clearly drugs make up the greatest proportion: 90% in case of malaria, against 4% on transport, 3% on consultation, none on stay while seeking treatment and 1% on hospitalization. For other illnesses, 84% of the expenditure is on drugs, 3% on transport, 4% on consultation, 4% for stay while seeking treatment and 3% on hospitalization.
Out-of-pocket expenditure and reason by treatment choice
We also examined out-of-pocket expenditure by treatment choice, specifically to understand how out-of-pocket expenditure varies with treatment choice and illness. A number of reasons were given for seeking self-treatment, trained health worker or traditional healers, most importantly lack of money, competence and severity of illness. These have policy implications for the control of malaria.
Figure 2 shows the reasons for choosing self-treatment, a trained health worker or a traditional healer and illustrates four points. On aggregate, as the illness becomes severe, patients choose to go to the health worker (see severity category) regardless of the illness in question. About 42.7% of malaria patients treat themselves because they feel competent to handle the situation, against 29.8% of patients with other illnesses. Regardless of the illness in question, a substantial number of patients chose self-treatment (81.2% for malaria, 64.3% for all other illnesses) – indicating a high preference for it, especially in cases of malaria. Traditional healers are rarely consulted for malaria treatment (0.9%) compared with other illnesses (17%). Much more out-of-pocket money was spent on treatment from health workers (748 540 CFA) than on self-treatment (373 340 CFA) or traditional healers (119 260 CFA).
Household strategies for mitigating the out-of-pocket expenditure burden
The strategies that households in Nouna use to mitigate out-of-pocket expenditure were pre-coded as (1) sold personal assets, (2) received free treatment, (3) received money as a gift, (4) borrowed money, (5) used cash, liquid savings and (6) worked for the money. But only four strategies were used, namely selling of personal assets, free treatment, money as a gift, and borrowing money. Table 2 shows that no households used cash or liquid savings. As expenditure on health rises, households sell assets to meet the cost. This observation is consistent for all illnesses but less so for malaria. More than three-quarters (78.9%) of households borrow money to pay for health care. This is slightly more for malaria. Free treatment and donations of money are very rare.
Table 2. Strategies for mitigating out-of-pocket expenditure on health in Nouna, Burkina Faso
Factors for differential out-of-pocket expenditure on health
We examined the factors for differential out-of-pocket expenditure using the Tobit estimation model separately for in and outside the health facility. The model estimates in Table 3 show differences for self-treatment and health facility treatment of malaria compared with all other illnesses, expenditure source or payment arrangement and value of material assets. Differences are observed for treatment in health facility only for age, sex and whether a household is urban or rural. There are significant differences in self-treatment only for payment arrangement and the value of animals.
Table 3. Model results of determinants of differential out-of-pocket expenditure on health, Nouna, Burkina Faso
It matters less whether treatment is sought from the health facility or self-treatment: less out-of-pocket money (significant at 1%) was spent on malaria compared with all other illnesses, 4503 CFA less for self-treatment and 8615 CFA less for the health facility. There are two possible explanations to this: first, the perception that malaria has been with the community for so long that it is perceived not as a threat but as an illness that can be treated with some tablets of CQ. This is corroborated by Figure 2, which shows that a large proportion of individuals chose self-treatment for malaria because they thought they were competent. Secondly, the value of material assets owned is significant at 5% for both self-treatment and treatment in the health facility. As this is an indicator of income, one would expect that the higher the income, the more an individual household is likely to pay.
For self-treatment most people first borrow to offset expenditure, and then, as bills increase, resort to selling their assets (Table 2). The possible explanation for this is that households fear low bargaining power in the period of need and therefore prefer to borrow and offset the debt later, but as costs rise, their credit worthiness decreases, or nobody may be able to lend such an amount of money.
We found a significant difference in out-of-pocket expenditure for self-treatment associated with the value of animals: the greater their value, the less out-of-pocket money a household is likely to spend. At first this looks odd as animals are a proxy of wealth. But when looking at the culture of the community and economic activity, two issues become clear: animals, especially donkeys, are used for farming and it is essentially difficult to sell donkeys in the event of illness. Cattle are a way of life for some sections on the population, and they would rather die than sell their cattle.
People spent more out-of-pocket on treatment from the health worker than on self-treatment and traditional healers (not shown in the table). Using the probit analysis, the difference was significant at 1%. Treatment elsewhere may have been incomplete, and to complete the course may have required more expenditure. Health workers may be more expensive, incurring cost both for consultation and transport.
Being in the urban area determines how much a household pays for health care in the health facility rather than on self-treatment, possibly because of differences in income levels between urban and rural areas or geographical access to health facilities.
Finally, age and sex are significant regarding choosing treatment in health facilities vs. self-treatment. The older a person is, the more out-of-pocket she or he is likely to pay for health care. Sauerborn et al. (1996) found age to be one of the factors for differential out-of-pocket expenditure on health in the area. Households spend more on males than females. This is, however, only true for health facilities, not self-treatment.
What implications do these results have for the global aim to Roll Back Malaria (RBM)? RBM comprises strategies to reduce access barriers to prompt and appropriate treatment, to sector-wide approaches and financing, to monitoring drug resistance and to improving quality of care at home. Our results have policy implications for these proposed strategies.
Less out-of-pocket is spent on malaria than other illnesses, yet it is the leading cause of morbidity and mortality in the population (Würtwein et al. 2000) regardless of treatment choice. Malaria may not be perceived as a threat any more (Sommerfeld et al. 2001), and people may consider themselves capable of diagnosing and treating malaria, and therefore prefer self-treatment to visiting public health facilities (Figure 2). This observation is worrying, and may have far-reaching implications for both policy and the future disease burden. In their study of perceived risk and vulnerability Sommerfeld et al. (2001) found that malaria ranked 10th in terms of being perceived as a threat, yet was the illness the community was most vulnerable to. People may not be able to diagnose and appropriately treat malaria. Studies elsewhere have shown that is the case; for example, only 20.1% of the mothers/guardians in Kibaha district of Tanzania knew the correct paediatric dose regime of CQ syrup, the most common medication (Nsimba et al. 1999). The policy concern is that communities need to be made aware of the risks of complicated malaria and the potential danger of inappropriate self-diagnosis and treatment.
Drugs make up about 90% of the total out-of-pocket expenditure for malaria and 84% for other illnesses, hence the likelihood of under dosing and consequent development of resistance is high (Shretta et al. 2000). One policy proposal would be to provide essential drugs free for those illnesses with the greatest burden, and to subsidize expenses for other treatment components for illnesses causing a lower burden.
Treatment choice and corresponding expenditure differences show that consistently more out-of-pocket expenditure was incurred in health facilities in comparison with self-treatment and traditional healers. About 98% of the respondents who cited lack of money as a reason for treatment choice chose either self-treatment or traditional healer. Other studies reveal a similar pattern, for example in Morocco, where those who visited public health facilities paid six times more than self-treaters (Hotchkiss & Gordillo 1999). The government could reduce the price of services, including drugs, to the level of other options, followed by mass information campaigns. The success of the Expanded Program on Immunization has largely depended on the availability of vaccines and awareness campaigns. The other option would be to make drugs for health conditions with the highest burden freely available.
A small proportion chose traditional healers, who are part of most communities in Africa, and whose role is likely to increase as economic conditions worsen. Arguably they can only treat a small portion of the illnesses in question successfully. It is not clear whether surveys capture all visits to traditional healers. Health policy should recognize and aim to integrate them in strategies to improve the health of the population.
Most households borrow to offset the out-of-pocket expenditure burden. This is in line with the theory that it is better to borrow than to liquidate ones assets as assets lose value when you are in need. Sauerborn (1994) found that `loans were perceived as buffers between the time of need for cash and the time when households saw a possibility of paying back. Selling an animal under pressure would lead to a bad price. So it was more advantageous, even for the wealthy, to take a loan and take some time to sell the necessary number of animals at a time when prices would be favourable. The same was true for cereal crops'. However, as more out-of-pocket expenditure is incurred, it becomes increasingly difficult to secure a loan and people may have to resort to accepting a poor price.
Differential resource allocation within households has been reported in a number of studies (Chen et al. 1981) and these have largely shaped policy development and targeting of intervention policies (Sen 1984; Das Gupta 1987; Haddad & Reardon 1993). These studies have mainly investigated gender and age in allocation of resources, termed as gender and age bias, respectively. The results show that males are allocated considerably higher out-of-pocket expenditure on health than females while seeking treatment in the health facilities. This contrasts with studies in western Africa (Haddad & Reardon 1993) which did not find any gender difference but conforms to studies conducted in South Asia (Chen et al. 1981; Sen 1984; Das Gupta 1987) which found discrimination against females. A possible explanation is that the methods used in the analysis in the two studies referred to in West Africa could not detect the gender bias. Haddad used stratified outlay equivalent analysis on the International Food Policy Research Institute (IFPRI) data set in Burkina Faso, and could not find any differences in expenditure between boys and girls. Sauerborn et al. (1996) used analysis of variance and found that although the average health care expenditures for sick women was half that of sick men (480 CFA vs. 1160 CFA), the difference was not significant (P=0.06). In both cases above, one would suspect that there was no optimal use of the available data. In this paper, we attempted to make the best use of all available data by Tobit analysis, and did not find gender bias in out-of-pocket expenditure on health care, either.
As expected, age is statistically significant (Caldwell et al. 1983), but only for health facility treatment, not self-treatment (Sauerborn et al. 1996). Caldwell et al. found under representation of the young and old among patients, which they attribute to parent perceptions of childhood illnesses. Sauerborn et al. report differential health expenditure biased towards 11–59 year-olds based on case studies in age groups of 0–10, 11–59, and 60+ years.
Households will spend as little as possible on malaria because they feel confident to treat it themselves. Hence policies aimed at fighting the spread of malaria need to focus on at least three issues: educating population on the risks associated with malaria complications and the associated risk of inappropriate diagnosis and treatment; pricing of drugs and services to create an incentive to use the health services; and building alliances between households, traditional healers and medical health workers.
We thank Dong Hengjin of the Department of Tropical Hygiene and Public Health, Germany, Chutima Suraratdecha of International Health Policy Programme, Health Systems Research Institute, Thailand, and Hsu Ke of the Global Program on Evidence for Health Policy, World Health Organization, for their constructive comments on earlier drafts.