- Email: [email protected]
Seasonal variations of household costs of illness in Burkina Faso
Soc. Sci. Med. Vol. 43, No. 3, pp. 281-290, 1996
Pergamon 027%9536(95)00374-6
Copyright © 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0277-9536/96 $15.00 + 0.00
SEASONAL VARIATIONS OF HOUSEHOLD COSTS OF ILLNESS IN BURKINA FASO R. S A U E R B O R N *l, A. N O U G T A R A 2, M. H I E N 2 and H. J. D I E S F E L D I qnstitute of Tropical Hygiene & Public Health, University of Heidelberg, INF 324, 69120 Heidelberg, Germany and 2Ministry of Health, Ouagadougou, Burkina Faso Abstract--This paper assesses the seasonal variations of the time and financial costs of illness for rural households in Burkina Faso. It is based on a multiple round survey of 566 households, which included a time allocation study. The economic parameters of households which influence health seeking behavior changed substantially between the dry and rainy seasons: revenues fell in the rainy season and were exceeded by expenditures. Household production was at its peak in the rainy season resulting in significantly higher opportunity costs of time. At the same time illness perception changed: in the rainy season, significantly fewer illness episodes were perceived, and of those, the proportion perceived as severe decreased over-proportionally. Households shifted their healer choice in the rainy season away from high cost treatment, such as the hospital and dispensary, to low cost home treatment. For all these reasons, households incurred significantly fewer costs of illness in the rainy season (27% of dry season costs). Household health care expenditures were reduced to 1/6 of dry season levels, the time costs incurred by healthy household members to tend to the sick was reduced to 1/5 and the time costs of work incapacity due to sickness fell to about 1/2 of dry season levels. The authors stress the need to carry out research in all relevant seasons when studying health seeking behavior and the household costs of illness in order to avoid serious seasonal bias. They suggest policy options to increase health care utilization in the rainy season by reducing the financial and time costs of access to health care. Finally, the authors put forward a hypothesis to be tested by future research: They argue that the cognitive (changes in illness perception) and behavioral changes (different health care seeking) reflect the high opportunity costs of time and the low availability of cash households face during the rainy season. The paper discusses the negative implication that untreated illness has on the health status of household members. Copyright © 1996 Elsevier Science Ltd
Key words--Health care costs, health expenditures, health care utilization, seasonality, agricultural production, illness perception
INTRODUCTION Seasonal rainfalls determine the rhythm of life and work of rural populations in Burkina, as in many developing countries, Agriculture is seen as the main economic activity of the household by 9 4 % of household heads. The study area has two main seasons, the rainy seasont" (san ji in Dioula), from June through November, and the dry season, from December to May. The dry season is further divided into a cold part (nene--Dccember through February) and a hot part (funteni-b~---March through May). Rainfall is concentrated in the months o f June through September, with small amounts of precipitation in May and October. Rainfalls drive the
*Author for correspondence. "['The term "rainy season" is used in the following to denote the time of agricultural activity. It is acknowledged that agricultural work starts shortly before the rains and continues for about a month after the end of the main rainfalls. 281
agricultural calendar, which shows peak activities for sowing in M a y / J u n e and harvesting in October/ November. This has been shown to have profound implications for the economy o f the household: opportunity costs of time, interest rates, wage rates, household expenditure, food prices, and more [1]. A number of authors have found that the time and financial costs involved with seeking health care are critical determinants of health care utilization [2-5]. It is therefore safe to assume that seasonal fluctuations in the economy o f the household lead to fluctuations o f the ability o f the household to allocate time and financial resources to health care. Assuming that in the peak agricultural season, the opportunity costs o f time are indeed higher and the availability of cash lower, we would expect that (i) household members spend less time on seeking care or on tending to the sick and that (ii) households spend less on health care. The latter would imply that (iii) there is a marked reduction in the use o f high cost health care alternatives, such as Western-type health services.
282
R. Sauerborn et
In addition to seasonal changes in the economic environment of the household, households experience profound seasonal changes in the burden of disease, prompting them to seek care and to incur costs. In the rainy season, the health status of the population is threatened by the combination of two factors: (i) Energy deficit: The rainy season is a time of peak physical activity with energy expenditure close to the physiological maximum [6, 7], while nutritional intake is at its lowest. Due to the resulting energy deficit, adults and children lose weight, birth weights fall, and malnutrition in children rises (for a review see [1]). (ii) High transmission of diseases: The frequency of major illnesses rises in the rainy season. Many studies in similar settings have reported an increase of the frequency and severity of major biomedically verified diseases. Increase in the incidence of disease in the rainy season has been best documented for malaria* and guinea worm disease.? Other diseases whose frequency have been reported to increase during the rainy season include childhood diarrhea [11], childhood malnutrition [12, 13], and acute respiratory tract infections [14]. This increased disease transmission can be expected to lead to substantial work incapacity in the rainy season. Based on the assumption of an increased burden of disease, we would anticipate a greater need for health care which stands in contrast to the reduced ability of households to garner resources to seek care. It is surprising that researchers have given so little attention to studying health seeking behavior during the rainy season when households face the double challenge of economic stress and increased disease burden. In a review of 16 studies assessing health care costs and utilization in developing countries, we found only two that looked at seasonal variations of these variables: Litvack and Bodart [15] analyzed health center records in rural Cameroon and found substantial variations in the number of new cases seen
*Greenwood and Picketing [8] report a more than threefold increase in the incidence of parasitemia with fever at harvest time (September/October), compared to the dry season. Their study was done in a climatic and agricultural setting in the Gambia, which is very similar to the study area. tBelcher et al. [9], reported a maximum of cases in the Danfa project villages in Southern Ghana between July and November. Cases were concentrated in the productive age groups between 15 and 54 years. Steib and Mayer [10] examined the epidemioiogy and the vectors of the guinea worm in the area of the present study. Over four years they reported annual incidence r a t e s from two of the current study villages, Toni and Kamadena, between 30/1000 and 81/1000, with a peak in the number of cases in the month of July.
al.
between the dry and rainy seasons: In the rainy season, only 60% of the number of new cases seen in the dry season sought treatment at health centers. The authors did not analyze these findings further, in fact they treated seasonality as a confounder to be controlled in their comparison of health care utilization before and after the introduction of user fees. Fabricant [16], studying the affordability and equity of primary health care costs in rural Sierra Leone, noted a similar drop (by about a third) in health care utilization in the rainy season. Neither of these two studies measured time costs. The present study therefore addresses four questions: (i) How do the household economic parameters (revenues, expenditures, time spent on household production, opportunity costs of time) change seasonally? (ii) What are the seasonal changes in perceived illness? (iii) Does health care utilization change from the dry to the rainy season, and if so in which direction? And finally, (iv) what time and financial costs do households incur in the different seasons, and is there any link with changes in the above three areas?
METHODS
A multiple round household interview survey spanning both the dry and rainy seasons was carried out using a two-stage cluster sample of n = 566 households, comprising 4820 persons. The same households were visited at six different time points: March, April (dry season) and July, September, October and November (rainy season). The household sample was drawn from a demographic surveillance area in the rural Kossi Province in Burkina Faso [17]. Using a recall period of four weeks, the same questionnaire was applied at each survey. Information was gathered on past perceived illness and healer choice as well as the financial and time costs incurred by the households of the sick individuals. Financial costs included households' outlays for user fees, drugs, lodging at a distant treatment site, and transportation. Time costs were measured in terms of the number of full days the sick and any caretakers were not able to perform their customary work (or play, in the case of children) due to illness. Time could be lost due to (i) illness inflicted incapacity, (ii) treatment associated travel and waiting time for both the sick and any caretaker. Days lost for work were converted into monetary terms as follows: economic theory equates the opportunity cost of time with the marginal cost of labor (MCL). The latter is usually approximated with the shadow wage-rate [18]. In the study area a labor market--albeit very small--does exist throughout the year. In order to obtain an estimate of the going
Seasonal variations of illness costs wage-rate, a hypothetical question was asked to the household heads: "If you were to hire somebody to work for you in the fields today, how much would you pay him per day?" In addition, a time allocation study based on the recall of "yesterday's activities" [19, 20] was carried out on the same household sample in both seasons (in March and in June 1992). With minor modifications, we used the same activity categories as the ones used by McSweeny [21] in her time allocation study of Mossi households in Burkina Faso. Household production was defined as any activity serving the households livelihood. It included agricultural production and animal husbandry, food processing and preparation, craft production, and household work such as fetching water and firewood.* The household costs of illness included three components: (i) the time cost of sickness: The opportunity cost of wages foregone by the sick person due to sickness. This includes the time costs of seeking treatment, i.e. travel and waiting time; (ii) the time cost of caretaker(s): The opportunity costs of healthy household members' time spent on treating or tending to the sick or accompanying him/her to the place of treatment. This includes the time costs of caring for or accompanying the sick to the treatment site; (iii) the financial costs of treatment: These are expenditures incurred by the household in seeking care for the sick member. They include the out-of-pocket expenditures for treatment, fees, drugs, transport, and the cost of subsistence at a distant treatment site. Total costs of illness in the household sample during the one-month recall period were calculated as follows: Let: F = total financial costs of health care in the preceding month (F CFA)t Fd = financial costs of drugs, herbs etc. (F CFA) F I = financial costs of fees (F CFA) F,r = financial costs of travel (F CFA) F~ = financial costs of subsistence (F CFA) T = total time costs in preceding month (days of forgone production) Ts = time costs of the sick person (days of forgone production) T~ = time costs of the caretaker(s) (days of forgone production) w = daily wage-rate n = number of illness episodes
*The time spent walking was included in the time spent fetching water or firewood. tF CFA--Francs Communaut6 du Franc Africain (currency of francophone West Africa). :~For a detailed description of the activity categories the reader is referred to Sauerborn [22].
283
a = age coefficient s = related to the sick individual = related to the caretaker(s) then: 1. Financial costs of illness:
F=
+ Ft + F,r + F, i
)
(1)
2. Time cost of illness:
T=i~o[(T,*a,*w)+(Tc,*ac*w)]
(2)
3. Economic cost of illness:
RESULTS
Seasonal changes in the household economy Illness related work incapacity is likely to lead to production loss. Since indirect illness costs are defined as the value of forgone production, it is essential to understand who produces how much and in which season. First, we describe household production and identify differences in the type and volume of production across gender and age groups as well as between seasons. Second, we develop a method to estimate the monetary value of a day of production lost. Household production. Household production was subdivided in four categories::~ (i) farming and animal husbandry; (ii) food processing (e.g. grinding cereal grains, producing peanut or karit6 butter); (iii) crafts and (iv) housework (cooking, fetching water and firewood). The amount of time allocated to each of these four main production activities changed significantly between the dry and the rainy seasons. The average number of hours allocated to farming increased from 0.4 to 3.8 hours per day for women and from 1.9 hours to 6.0 hours per day for men (Table 1). In the dry season, women's contribution to household production was significantly higher than that of men, while the gender differences vanished in the rainy season: men work 6.6 hours and women 6.8 hours per day. Figure 1 displays the average number of hours worked per day at different ages. The age differences in daily work time were highly significant (P < 0.00001, analysis of variance). For each five year age group between 5 and 70 years, the seasonal differences of the hours worked on productive activities were statistically significant (P < 0.00001, analysis of variance). Hence we must assume that more production is lost when a 25 year old is incapacitated by illness compared to the loss of the same amount of time in a 9 year old (or a 70 year old).
R. Sauerborn et al.
284
Table 1. Average hours per day spent on household production, by gender and season. The time allocation study is based on N = 3086 cases~for the dry season survey and N = 2805 cases in the rainy season Average time per day spent on production Dry season Household production Farming/animal husbandry
Between-season differences
Rainy season
Dry vs Women Men Men vs women Women Men Men vs women rainy season hours/day hours/day P (anova) hours/day hours/day P (anova) P (anova) hours/day hours/day Men vs women hours/day hours/day Men vs women Between-season
Farming/animal husbandry 0.4 1.9 0.0001 3.8 6.0 0.0001 0.0001 Food processing 1.4 0.1 0.0001 0.7 0.08 0.0001 0.0001 Crafts 1.I 1.3 ns 0.5 0.3 ns 0.0001 Housework 2.8 0.5 0.0001 1.8 0.2 0.0001 0.0008 Total production 5.7 3.8 0.0001 6.8 6.6 ns 0.0001 ~Timeallocation was only asked for household members above the age of 4 years. The excluded age group of under five years represented 946 individuals (19.6% of the total population). For 788 individuals (16% of the total) no time allocation was obtained.
To adjust for these differences in age-specific productivity when calculating individual opportunity costs of time, a weighting procedure was used as follows: the average number of hours/day spent on household production in the rainy period between ages 15 and 39 years* was used as the maximum work input (weight = 1.0). The proportion of this reference value that a given age group contributes to household production is the weight. The weights obtained that way are similar to the age preference assumed in estimations o f the economic burden of disease [23]. These coefficients were used in the calculation of the opportunity costs of time: the going wage-rate (w) was multiplied by the coefficient (a) corresponding to the age o f the person who incurred time costs [see equation (2) above]. Household expenditures and revenues. In the dry season, households' revenue exceeded expenditures. Towards the end o f the season (April) both revenue and expenditures fell markedly. While revenue
*Both sexes are combined in this analysis.
continued to fall--reaching its lowest point just before the harvest, expenditure climbed a g a i n - fuelled by food purchases [22]. Throughout the rainy season, until the first harvest was in (October), households remained in a negative cash balance, i.e. expenditures exceeded revenues which made it increasingly difficult to absorb health care expenditure (Fig. 2). Opportunity costs o f time. The mean of the distribution was 289 F C F A (SD = + 83 F C F A ) for the dry season (March) and 379 F C F A ( + 105 F C F A ) for the rainy season (July). The difference between both wage rates was highly significant (P < 0.00001, Anova).
Seasonal changes in illness perception The onset o f the rainy season brought two significant changes in the perception of illness, affecting household costs of illness. Firstly, fewer individuals perceived an illness in the previous month, the number of perceived illness episodes fell from 867 to 636 (Table 2). Secondly, the episodes
8.00 7.00 6.00
$.00 4.00
3.1111 ~
2.1111 1.00 0 gY
10-14Y
20-24Y
30-34Y
40-~AY
50-[~l.Y
60-64Y
Fig. I. Household production by age group and season, both genders combined.
> ffi 70 Y
Seasonal variations of illness costs
285
18,000 16,000 14,000 es
o
<
Harvesting
12,000
10,000
8000
"',,,....,....,~
""
~'.,..,
a
Expenditure/household
•
Revenue/household
°A.,
2000 0
I
Mar
Apt
II
I
Jal
II
l
l
I
Sep
Oct
Nov
Fig. 2. Synopsis of seasonal fluctuations of household expenditures and revenues.
perceived were reported as less severe. The proportion of illnesses judged by the sick as severe dropped significantly from 36% in March (dry season) to 8% in October (harvest season). Since the mean financial cost per treatment episode was significantly less ( - 47%) for an illness perceived as not severe (Anova, P < 0.05), the seasonal change in perceived illness severity had an additional effect on the financial cost o f illness. The same applied for time costs. The mean time loss was 0.7 days for an episode of mild illness and 6.1 days for a severe illness. The difference o f these means was significant (P < 0.0005).
Seasonal changes in health care utilization The average financial cost per treatment episodes* showed large variations depending on the choice of t r e a t m e n t - - f r o m 76 F C F A / t r e a t m e n t episode for home treatment to 6939 F C F A / t r e a t m e n t episode for hospital treatment. In the rainy season, households reduced the financial and time costs o f illness through two changes in health seeking behavior: firstly, much *For a detailed analysis of costs by choice of treatment in the dry season only, the reader is referred to Sauerborn et al. [5].
less use was made o f any form o f treatment: while 674 of the 1050 (64.2%) illness episodes in the dry season received some kind o f treatment (including self-treatment), only 259 of the 752 ill individuals (34.4%) were treated at all (Table 2). Secondly, households shifted their treatment choice from high-cost to low-cost alternatives. Table 3 shows that both hospital and dispensary care, as well as the time-consuming treatment of traditional healers, were chosen less frequently. H o m e treatment with its low time and financial costs was chosen in about two thirds o f illness episodes in the rainy season
Seasonal changes in the household costs of illness The combined effect of fewer perceived illnesses, o f which a smaller proportion was perceived as severe, a reduced propensity to seek treatment, a n d - - i n cases where treatment is sought at a l l - - a shift in health care choice towards low-cost treatment options led to a sharp d r o p - - f r o m 4002 to 1065 F C F A (i.e. by 7 5 % ) - - i n the average monthly household costs of illness in the rainy season. Figure 3 shows the seasonal differences in household costs o f illness for each o f the three cost components. We can draw two conclusions from this graph: (i) the mean values of all cost components were reduced
Table 2. Seasonal differences in the numbers of ill household members, perceivedillnessepisodes, number of those treated, and the number of treatment episodes Numbers Description of variable Dry season Rainy season No. of households 566 547 No. of individuals 4820 4634 No. of ill individuals (in % of study population) 867(1~1%) 636 (13.7%) Avg. no. of episodes/iUindividual 1.21 1.18 No. of illness episodes 1050 752 No. of illness episodes treated (in % of all illness episodes) 674 (64.2%) 259 (34.4%) No. of treatment episodes/illnessepisode 1.22 1.6 No. of treatment episodes 829 282
286
R. Sauerbom et al. Table 3. Seasonal differences in treatment choice and expenditure by age group and season. The cost differences between treatment choices are significant (P < 0.00001) Utilization (%) by season Treatment choice Hospital care Nurse, outpatient care Traditional healer Village health worker Home care
Average/costs/treatment F CFA (SD)
Rainy
Dry
6939 (17,261) 1432 (2560) 656 (1965) 221 (462) 76 (245)
2.1 8.5 11.7 3.5 64.9
7.4 10.1 15.4 4.6 49.0
in the rainy season compared to the dry season;* (ii) the relative size of the three components of economic costs changed from the dry to the rainy season: whereas in the dry season the relative contribution of all three components to total household costs was about the same (one third), time costs of sickness amounted to 57% in the rainy season, time costs of the caretaker to 24%. The strongest relative reduction was in financial costs: The average household financial costs of illness was reduced more than six-fold from 1250 to 197 F CFA. They contributed only 18.5% to total economic costs of illness in the rainy season (instead of 31.2% in the dry
seasons no statistically significant differences were found. DISCUSSION
Most field studies are carried out in the dry season--for logistical, transportation and convenience reasons. However, great caution is warranted when findings from the dry season are generalized. Our study shows that estimates based on extrapolations form a single survey carded out in the dry season will overstate household health care consumption and both the time and financial costs of ill health. season). Had we based our estimates of the household costs W h e n controlling for severity (Table 4), we found of illness on our March survey alone (dry season), we that households spent on average significantlylesson would have calculated the annual economic costs of episodes of severe illnessesin the rainy season than in illness to be 12 x 4002 F CFA = 46,024 F CFA. The the dry season (448 and 957 F C F A respectively, seasonally adjusted annual cost estimates, however, P < 0.005). In addition, households significantly were only 30,401 F CFA. Chambers [1] coined the reduced the amount of time healthy caretakers spent terra "seasonal bias" for such distortions arising from to tend to the severely ill:from 7 to 5.8 hours/illness the researcher's tendency to carry out studies in the episode on average (P < 0.05). W h e n the three cost convenient dry season and making linear projections components were compared for mild illnessesacross of health or economic variables that fluctuate substantially with season. The authors therefore recommend that studies assessing the household costs *It is important to note that all cost variables show a highly of illness, the demand for health care, the skewed distribution. This is due to the fact that (i) illness clusters in a relatively small number of households. Of willingness-to-pay or the economic impact of specific those, only a minority incurs any costs at all. In the dry diseases be carried out in all relevant seasons of the season survey, for example, of the 566 households, year. 69.4% (393) reported an illness in the preceding month. Sauerborn et al. [5] stressed the importance of t i m e Of these, 72% (281) households treated the illness. Of costs within total costs of illness. While this earlier these 281, only 61.9% (174) incurred any financial costs. So the financial burden of illness is borne by less than analysis was confined to a dry season survey, the a third (174 households or 30.7%) of all households [22]. findings presented in the current paper point out to
~1400
1395
1356
rJ r~ 1200 ~'-' 1000 =
~
[] Dry season
BOO ~OO
[] Rainy season
~ 4OO .~ 2OO
~
o
Time
Time
Financial
cost of illness
cost of
cost of taro
caretaker
Fig. 3. Seasonal changes of the three components of household costs of illness. Time cost of illness (due to time lost by the sick individual), time costs of the caretaker (due to time lost by household members tending to or accompanying the sick) and the out-of-pocket expenditures the household incurs to treat the sick (financial costs of care).
Seasonal variations of illness costs
Type of cost Financial costs Time costs (sick) Time costs (caretaker) *P < 0.05, **P < 0.005
287
Table 4. Costs per illnessepisode by season, controllingfor severity of illness. Costs incurred per illnessepisode (F CFA) Mild illness Severe illness Unit Dry Rainy Dry Rainy F CFA 500 (1546) 483 (1489) 957** (2632) 448** 0337) Days 1.4 ( _+8.4) 1.3 (8.4) 7.0 ( _ 16.7) 5.8 00.3) Days 0.9 ( _+4.8) 0.4 (3.1) 5.6* (18.4) 1.9" (4.9)
an even bigger relative importance of time costs in the rainy season. In spite of higher opportunity costs of time in the rainy season,* households lost less production due to the fact that fewer people perceived themselves ill, and of those, fewer assumed the sick role, i.e. stayed away from work. It is interesting to note that households reduced caretaker time substantially more than the time lost by the sick. The relative share of time costs of all rainy season costs of illness of the household grew from 68% in the dry season to 82% in the rainy season. This was due to the fact that households curbed the financial component of their illness costs even more drastically. As far as policy is concerned the most troublesome finding of our study is that households' utilization of health, especially of formal Western-type health services, was significantly lower in the rainy season. This finding is corroborated by data from Cameroon reported by Litvack and Bodart [15] who found a 40% reduction of the n u m b e r of new cases seen at health facilities in the rainy season. Fabricant [16] made a similar observation: he found that utilization rates of health services decreased significantly (by 25%) while home treatment increased. Given that in Burkina health service utilization in the dry season is already very l o w - - o n l y about 1 in 5 illness episodes are treated by professional health services--these services become almost irrelevant in the rainy season, just at a time when health needs are greatest. Obviously, health care interventions with proven, impressive efficacy cannot be expected to have any significant health impact as long as people do not use them. We support Fabricant's argument that "the drop in utilization was presumed to be due to the seasonal effect of lower ability of cash". He estimated that about l/3 of the utilization drop was attributable to poor physical access while the remainder was attributable to reduced availability of cash.t In addition to the financial costs of seeking care, the high opportunity costs of time are likely to make
*An observation confirmed by Mwabu's [18] study on the seasonal changes in the shadow price of time in Kenya. tFabricant [16] found that, in the rainy season, the average household's cash on hand was only about 1/10 of that in the dry season. :[:The economic access barriers of health care are compounded by physical barriers such as muddy or flooded roads and collapsed bridges, but these physical access barriers are not the subject of this paper.
health services economically less accessible in the rainy season. Mwabu [18] argued: the rise in the value of time in the wet season raises the costs of medical services and thus, other things being equal, the rate of their utilization falls. Households are, of course, not static entities. They can be expected to react to the economic costs of ill health, mobilize financial resources and substitute health. These "coping strategies" are discussed in depth in a separate paper [24]. Given that in the rainy season, household revenue is at its lowest and opportunity costs of time at its highest, we suggest that one way of increasing resort to health services in the rainy season would be to reduce the excessive economic access costs:~ during that period. One way to reduce the time costs to health care would be to strengthen health care capabilities within the village or even within the household. Since the obvious policy option, namely to build, staff and equip more rural health facilities, is not feasible in Burkina Faso, and in view of the failure of the "community health worker" scheme [25-27], the Ministry of Health is currently testing an alternative way of reducing both financial and time access costs by strengthening mothers' home treatment. Given that mothers provide the bulk of primary health care using traditional medicine [27, 28], the policy being developed consists of teaching mothers to incorporate carefully selected diagnostic and treatment tasks currently performed by nurses into their home treatment practices. The feasibility of mothers treating dehydration with home-made oral rehydration solution has been show by many authors [29-3 l]. Other tasks that mothers could assume--at least in the rainy season--include recognizing and treating malaria (oral chloroquine) and recognizing and treating acute respiratory infections. A study assessing the feasibility of teaching mothers to include Western type health care in their home care practices is currently under way in the study area. Another policy option currently implemented by the Burkinian Ministry of Health [32] in the study area is to reduce the time costs associated with travel by establishing pharmacies in close proximity to dispensaries so that patients no longer incur travel costs to fill prescriptions in the district capital. Whatever route is taken to reduce time costs of access to health care, research is needed to analyze the effect of these policies on household costs of illness and on health care utilization.
288
R. Sauerbomet al.
As regards policies aiming to reduce the financial costs of illness, the current study has shown that professional health services account for the largest share of households' expenditures for care, with drug expenditures ranking first. Since October 1993, Burkina's Ministry of Health has offered generic drugs through co-managed community revolving drug funds at prices below those of private pharmacies. This policy, inspired by experience elsewhere in sub-Saharan Africa (for a review see [33, 34]) has been shown to reduce the total financial cost of a health care visit and to increase health care utilization [15]. Most importantly, schemes to finance health services should address the economic burden of rainy season health care. One possibility would be to build "seasonal cross-subsidies" into existing price schedules of health services. This would mean charging lower prices for services and drugs in the rainy season and higher prices in the dry seasons, as suggested by Fabricant [16]. Another option consists in dissociating the time of payment from the time of service provision. This appears to be a logical way to take the households' fluctuating availability of cash into account (see Fig. 2). Prepayment schemes that collect premiums after the harvest season and cover basic preventive and curative care throughout the year would also share the costs between the healthy and the sick. Risk-sharing prepayment schemes have long thought to be impractical in rural sub-Saharan Africa.* However, they have recently received increasing attention [35-40]. The few documented ongoing prepayment schemes stress the need for community involvement, the use of funds--inter alia--for quality improvement of services and the need for political commitment from national policy makers to support the schemes [35]. Further research is suggested to evaluate the effect of prepayment schemes on health care utilization and on household cost of illness especially in the rainy season. We agree with Mwabu [18] that household level studies which assess the full financial and time cost of the household as well as its health seeking pattern and which capture the seasonal fluctuations are essential for the design of effective health policies. The most surprising--and probably the most controversial--finding of this study is that illness perception--both in terms of the frequency and severity of illnesses---dropped significantly from the dry to the rainy season. These changes in illness perception during times of seasonal stress were shown to be the most important factor in reducing the cost burden of the household in the rainy season. In Sierra Leone, Fabricant [16], too, observed a drop in the frequency of overall illnesses, ranging from 6-23%. This is in contrast to reports of an increase in the
*For a review see Carrin [41] and Shaw and Ainsworth [34].
frequency of biomedicaUy verified disease reviewed in the introduction, a point also made by Fabricant [16] who was surprised that "morbidity in the rainy season was not higher than in the dry season". As to the severity of disease in the rainy season, Fabricant [16] is the only author we could identify who reports a drop of the number of illnesses perceived as severe in the rainy compared to the dry season. The great majority of authors, however, report a rise in the proportion of severe diseases. An increase in monthly death-rates in the rainy season was reported for malaria in Burkina Faso [42] and in Gambia [8]; for infant deaths in Senegal [43]; and for all causes of death from Matlab, Bangladesh [44]. Three hypotheses can be forwarded to explain the apparent contradiction between a higher burden of biomedical disease in the rainy season as reported from studies in similar settings and a decrease in the burden of perceived illness in our study population: (i) It could be argued that there were indeed fewer and less severe diseases during the rainy season in the study area. However, although the present study did not measure clinical disease as defined by examination or laboratory tests, it is hard to believe that the epidemiological situation in the study population would be radically different from those described above, given the environmental and socio-economic similarities between study settings noted above. In conclusion, it is implausible to accept that people are indeed healthier during a period of reduced energy intake, peak energy expenditure, a negative energy balance, and increased transmission of major diseases [1]. (ii) Systematic under-reporting of illness episodes could have occurred during the rainy season. However, this is unlikely since hospital and dispensary records, too, showed a substantial decline of health care use in the study area. The close supervision of interviewers--one in ten interviews was chosen at random and redone by the interviewer's supervisor--makes it very unlikely that interviewer fatigue, i.e. their decreasing willingness to comply with the interview procedures, was the underlying reason. (iii) Finally, the drop in illness perception could be a strategy, if an unconscious one, to cope with simultaneous seasonal stress. A variety of factors known to reduce the sicks' propensity to seek care (such as muddy roads, unavailable transportation) are not likely to reduce illness perception per se. The authors argue that it is plausible that the high opportunity costs of time and the low availability of cash and assets in the
Seasonal variations of illness costs rainy season lead to lower illness perception. As Kroeger [45] states: . . . in peasant societies, where the demand for health services depends strongly on their availability, and where everybody's manpower is needed (particularly during the harvest period), entering the sick role is almost prohibited. In other words, people simply do not have time to be sick in the rainy season. This, in turn, leads to fewer days off work (i.e. a lower proportion of ill individuals assume the sick role), less time spent by caretakers on the sick, and less financial costs incurred for treatment. The implications of our tentative conclusion, that people indeed "ignore disease" to avoid costs in times of concurrent seasonal stress, would be serious from a public health view point: First, while "ignoring" disease may be effective in reducing costs in the short run, it carries the risk of causing disproportionately high expenditures in the future. Consider the case of inguinal hernias: they can be ignored as long as they do not incarcerate. When they do, they are b o u n d to surpass even the highest illness perception threshold, and entail all the pain and costs related to a medical emergency. Second, untreated disease may lead to negative externalities. For example, infectious diseases, if untreated, may spread to healthy households members. In their work on tuberculosis, Stanton and Clemens [46] stressed the high costs that untreated disease inflicts on households. Third, we would have to think about new ways to influence illness perception, i.e. increase people's awareness of (biomedical) diseases they may have. We readily acknowledge that our hypothesis that changes in illness perception are driven by the changing economic parameters individuals and households face, while plausible and consistent with our findings, is by no means proven to be correct by the current analysis. In order to test this hypothesis, further research is required which collects prospective and simultaneous data on: (i) illness perception-through interviews, (ii) disease--through some objective test for key diseases, such as a thick blood film for malaria or a urine test for schistosomiasis etc. and (iii) n u m b e r of cultural and socio-economic factors, notably the household's opportunity costs of time, thought to influence illness perception and health care choice.
Acknowledgements--This study was part of the "Burkina
Health Care Intervention Study" financed by the European Commission, Directorate General XII (Science, Research and Development) under contract #TS3*CT92*0078) and carried out by the Ministry of Health of Burkina Faso (as principal investigator), the Medical Faculty of Ouagadougou University, ORSTOM (Organisme de la Recherche et la Technologic pour le Developpement), Paris, and the Institute of Tropical Hygiene and Public Health of Heidelberg University. In addition, the Gesellschaft fiir Technische Zusammenarbeit (GTZ) and the Deutscher
289
Entwicklungsdienst (DED) actively supported the study. We wish to thank the entire staff of this research project for their untiring work. We would like to thank Uwe Brinkmann~, Peter Berman, Allan Hill, and Joseph Maxwell, with whom we had many stimulating discussions. We are deeply indebted to the households we interviewed, for their hospitality, and the patience with which they answered the many questions we posed. REFERENCES
1. Chambers R., Longhurst R. and Pacey A. Seasonal Dimensions to Poverty, pp. 1-7. Pinter, London, 1981. 2. De Ferranti D. Paying for Health Services in Developing Countries. World Bank Staff Working Papers No. 721, Washington, D.C., 1985. 3. Gertler P. and Van Der Gaag J. The Willingness to Pay for Medical Care. Johns Hopkins Press, Baltimore M.D. 1990. 4. Heller P. A model of the demand for medical and health services in peninsular Malaysia. Soc. Sci. Med. 16, 267, 1982. 5. Sauerborn R., Ibrangho I., Nougtara A., Borchert M., Koob E., Hien M. and Diesfeld H. J. The economic costs of illness for rural households in Burkina Faso, Trop. Med. Parasitol. 42, 219, 1995. 6. Brandt H. Work Capacity Restraints in Tropical Agriculture Development. Peter Lang, Frankfurt, 1980. 7. Bleiberg F., Brun T. A., Goihman S. and Gouba E. Duration of activities and energy expenditure of female farmers in dry and rainy seasons in Upper-Volta. Br. J. Nutr. 43, 71, 1980. 8. Greenwood B. and Picketing H. A review of the epidemiology and control of malaria in The Gambia, West Africa. Trans. R. Soc. Trop. Med. Hyg. 87($2), 3, 1993. 9. Belcher D. W., Wurapa F., Ward W. and Lourie I. Guinea worm in southern Ghana: its epidemiology, and impact on agricultural productivity. Am. J. Trop. Med. Hyg. 24, 243, 1975. 10. Steib K. and Mayer P. Epidemiology and vectors of Dracunculus medinensis in Northwest Burkina Faso, West Africa. Ann. Trop. Med. Parasitol. 82, 189, 1988. 11. Drasar B., Tomkins A. and Feachem R. The seasonal ecology of disease: Diarrheal diseases. In Seasonal Dimensions to Poverty (Edited by Chalmers R., Longhurst R. and Pacey A.), pp. 102-111. Pinter, London, 1981. 12. Chen L., Chowdhury A. and Huffman S. Seasonal dimensions of energy protein malnutrition in rural Bangladesh. Ecol. Food Nutr. $, 175, 1979. 13. Rowland M. G. M. Assessment of health programme effects with longitudinal studies. Biosoc. Sci. 10, 87, 1989. 14. Sutton R. The seasonal ecology of disease: respiratory diseases. In Seasonal dimensions to poverty (Edited by Chalmers R., Longhurst R. and Pacey A.), pp. 112-114. Pinter, London, 1981. 15. Litvack J. I. and Bodart C. User fees plus quality equals improved access to health care: results of a field experiment in Cameroon. Soc. Sci. Med. 37, 369, 1993. 16. Fabricant S. J. Community financing in Sierra Leone: affordability and equity of primary health care costs. PhD thesis, London University, 1992. 17. Sauerborn R., Nougtara A. and Diesfeld H. J. Action research on health services in Burkina Faso: concept and methodology 1988-1992. In Proceedings of the Conference on "Methodology and Relevance o f Health Systems Research" organized by the International
Children's Centre. Paris, 8-10 April 1992. 18. Mwabu G. Seasonality, the shadow price of time and effectiveness of tropical disease control programs. In
290
19.
20. 21. 22.
23. 24. 25.
26.
27.
28.
29. 30.
31,
32.
R. Sauerbornet al.
Economics, Health and Tropical Diseases (Edited by Herrin A. N. and Rosenfield P. L.), pp. 259-270, 1988. White B. Measuring time allocation, decision making and agrarian changes affecting rural women: examples from recent research in Indonesia. Paper for the Methodolgy Workshop "Research on Rural Women", IDS, 12-14 March 1982. University of Sussex, 1983. Acharya M. Time use data and the living standards measurement study. Worm Bank Working Paper 18, 1-75, 1982. McSweeny B. Collection and analysis of data on rural women's time use. Stud. Family Plan. 10, 379, 1979. Sauerborn R., Nougtara A., Benzler J., Borchert M., Diesfeld J., Hien M., Ibrango I. and Maiga M. Strategies d'Adaptation des M~nages aux Co~ts Economiques de la Maladie. Lang Verlag, Frankfurt, 1995. World Bank. Worm Development Report: Investing in Health, p. 26. World Bank, Washington, D.C., 1993. Sauerborn R., Adams E. and Hien M. Household strategies to cope with the economic costs of illness. Soc. Sci. Med. 43, 291, 1996. Nougtara A., Sauerborn R. and Diesfeld H. Assessment of MCH-services offered by professional and community health workers in the district of Solenzo, Burkina Faso. (I) Utilization of MCH-services. J. Trop. Pediat. 35(Suppl.), 2, 1989. Kabor~ B. A. Soins de Sant~ Primaires au Burkina Faso: Centre de Sant~ et de Promotion Sociale et/ou Poste de Sant6 Primaire. Thesis, University of Antwerp, 1985. Sauerborn R., Nougtara A. and Diesfeld H. Low utilization of community health workers--results from a household interview survey in Burkina Faso. Soc. Sci. Med. 29, 1163, 1989. Sauerborn R. and Nougtara A. Traditional medicine is practiced by mothers rather than by traditional healers--a study in Solenzo, Burkina Faso. Curare 15, 97, 1992. Cutting W. Can village mothers prepare oral rehydration solution? Trop. Doc. 8, 195, 1979. Chen L., Black R., Merson M., Bhatia S., Yunus M. and Chakraborty J. Village-based distribution of oral rehydration therapy packets in Bangladesh. Am. J. Trop. Med. Hyg. 29, 285, 1980. Chowdhury A., Vaughan P. and Abed F. Use and safety of home-made oral rehydration soutions: an epidemiological evaluation from Bangladesh. Int. J. Epidemiol. 17(3), 655, 1988. Ministate de la Sante. Document National sur le Renforcement des Soins de Sant~ Primaires au Burkina
33. 34.
35. 36.
37. 38. 39.
40. 41. 42.
43.
44.
45. 46.
Faso--projet de Dbmarrage de rlnitiative de Bamako. Minist6re de la Sant6, de l'Action Sociale et de la Famille, Ouagadougou, Burkina Faso, 1992. McPake B., Hanson K. and Mills A. Community financing of health care in Africa: an evaluation of the Bamako initiative. Soc. Sci. Med. 36, 1383, 1993. Shaw R. P. and Ainsworth M. Financing health services through user fees and insurance: lessons from Sub-Saharan Africa. (Draft manuscript). World Bank, 1994. Chabot J., Boal M. and Da Silva A. National community health insurance at village level: the case from Guinea Bissau. Hlth Pol. Plan. 6, 46, 1991. Baza A., Hakizimana A., Hanson K., Kwizera F. and Van Der Geest S. Health insurance and the Bamako initiative in Burundi: value for money. Int. J. HIth Plan. Mgmnt 8, 129, 1993. Arhin D. C. The health card insurance scheme in Burundi. A social asset or a non-viable structure? Soc. Sci. Med. 39, 861, 1994. Arhin D. C. Health insurance in Africa. Lancet 345, 44, 1995. Shepard D., Vian T. and Kleinau E. F. Impact of four health insurance programs in rural and urban areas of Za'ire. In Financing Health Services Through User Fees and Insurance: Lessons From Sub.Saharan Africa (Edited by Shaw R. P. and Ainsworth M.), pp. 249-286, World Bank, 1994. Abel-Smith B. and Rawal P. Employer's willingness to pay: the case for compulsory health insurance in Tanzania. Hlth Pol. Plan. 9(4), 409, 1004. Carrin G. Community financing of drugs in SubSaharan Africa. Int. J. Hlth Plan. Mgmnt 2, 125, 1987 Sauerborn R., Shepard D., Ettling M. and Etal A. Estimating the direct and indirect economic costs of malaria in a rural district of Burkina Faso. Trop. Med. Parasitol. 42, 219, 1991. Cantrelle P. Mortalit6 facteurs. In ONED, Afrique Noire, Madagascar, Comores: Ddmographie Comparde. D616gation G6n6rale de la Recherche Scientifique et Technique, Paris, 1967. Dyson T. and Crook N. Causes of seasonal fluctuations in vital events. In Seasonal Dimensions to PoverO' (Edited by Chambers R. et al.), pp. 135-141. Pinter, London 1981. Kroeger A. Response errors and other problems of health interview surveys in developing countries. WId Hlth Star. Q. 38, 15, 1985. Stanton B. and Clemens J. User fees for health care in developing countries: a case study of Bangladesh. Soc. Sci. Med. 29, 99, 1989.
Pergamon 027%9536(95)00374-6
Copyright © 1996 Elsevier Science Ltd Printed in Great Britain. All rights reserved 0277-9536/96 $15.00 + 0.00
SEASONAL VARIATIONS OF HOUSEHOLD COSTS OF ILLNESS IN BURKINA FASO R. S A U E R B O R N *l, A. N O U G T A R A 2, M. H I E N 2 and H. J. D I E S F E L D I qnstitute of Tropical Hygiene & Public Health, University of Heidelberg, INF 324, 69120 Heidelberg, Germany and 2Ministry of Health, Ouagadougou, Burkina Faso Abstract--This paper assesses the seasonal variations of the time and financial costs of illness for rural households in Burkina Faso. It is based on a multiple round survey of 566 households, which included a time allocation study. The economic parameters of households which influence health seeking behavior changed substantially between the dry and rainy seasons: revenues fell in the rainy season and were exceeded by expenditures. Household production was at its peak in the rainy season resulting in significantly higher opportunity costs of time. At the same time illness perception changed: in the rainy season, significantly fewer illness episodes were perceived, and of those, the proportion perceived as severe decreased over-proportionally. Households shifted their healer choice in the rainy season away from high cost treatment, such as the hospital and dispensary, to low cost home treatment. For all these reasons, households incurred significantly fewer costs of illness in the rainy season (27% of dry season costs). Household health care expenditures were reduced to 1/6 of dry season levels, the time costs incurred by healthy household members to tend to the sick was reduced to 1/5 and the time costs of work incapacity due to sickness fell to about 1/2 of dry season levels. The authors stress the need to carry out research in all relevant seasons when studying health seeking behavior and the household costs of illness in order to avoid serious seasonal bias. They suggest policy options to increase health care utilization in the rainy season by reducing the financial and time costs of access to health care. Finally, the authors put forward a hypothesis to be tested by future research: They argue that the cognitive (changes in illness perception) and behavioral changes (different health care seeking) reflect the high opportunity costs of time and the low availability of cash households face during the rainy season. The paper discusses the negative implication that untreated illness has on the health status of household members. Copyright © 1996 Elsevier Science Ltd
Key words--Health care costs, health expenditures, health care utilization, seasonality, agricultural production, illness perception
INTRODUCTION Seasonal rainfalls determine the rhythm of life and work of rural populations in Burkina, as in many developing countries, Agriculture is seen as the main economic activity of the household by 9 4 % of household heads. The study area has two main seasons, the rainy seasont" (san ji in Dioula), from June through November, and the dry season, from December to May. The dry season is further divided into a cold part (nene--Dccember through February) and a hot part (funteni-b~---March through May). Rainfall is concentrated in the months o f June through September, with small amounts of precipitation in May and October. Rainfalls drive the
*Author for correspondence. "['The term "rainy season" is used in the following to denote the time of agricultural activity. It is acknowledged that agricultural work starts shortly before the rains and continues for about a month after the end of the main rainfalls. 281
agricultural calendar, which shows peak activities for sowing in M a y / J u n e and harvesting in October/ November. This has been shown to have profound implications for the economy o f the household: opportunity costs of time, interest rates, wage rates, household expenditure, food prices, and more [1]. A number of authors have found that the time and financial costs involved with seeking health care are critical determinants of health care utilization [2-5]. It is therefore safe to assume that seasonal fluctuations in the economy o f the household lead to fluctuations o f the ability o f the household to allocate time and financial resources to health care. Assuming that in the peak agricultural season, the opportunity costs o f time are indeed higher and the availability of cash lower, we would expect that (i) household members spend less time on seeking care or on tending to the sick and that (ii) households spend less on health care. The latter would imply that (iii) there is a marked reduction in the use o f high cost health care alternatives, such as Western-type health services.
282
R. Sauerborn et
In addition to seasonal changes in the economic environment of the household, households experience profound seasonal changes in the burden of disease, prompting them to seek care and to incur costs. In the rainy season, the health status of the population is threatened by the combination of two factors: (i) Energy deficit: The rainy season is a time of peak physical activity with energy expenditure close to the physiological maximum [6, 7], while nutritional intake is at its lowest. Due to the resulting energy deficit, adults and children lose weight, birth weights fall, and malnutrition in children rises (for a review see [1]). (ii) High transmission of diseases: The frequency of major illnesses rises in the rainy season. Many studies in similar settings have reported an increase of the frequency and severity of major biomedically verified diseases. Increase in the incidence of disease in the rainy season has been best documented for malaria* and guinea worm disease.? Other diseases whose frequency have been reported to increase during the rainy season include childhood diarrhea [11], childhood malnutrition [12, 13], and acute respiratory tract infections [14]. This increased disease transmission can be expected to lead to substantial work incapacity in the rainy season. Based on the assumption of an increased burden of disease, we would anticipate a greater need for health care which stands in contrast to the reduced ability of households to garner resources to seek care. It is surprising that researchers have given so little attention to studying health seeking behavior during the rainy season when households face the double challenge of economic stress and increased disease burden. In a review of 16 studies assessing health care costs and utilization in developing countries, we found only two that looked at seasonal variations of these variables: Litvack and Bodart [15] analyzed health center records in rural Cameroon and found substantial variations in the number of new cases seen
*Greenwood and Picketing [8] report a more than threefold increase in the incidence of parasitemia with fever at harvest time (September/October), compared to the dry season. Their study was done in a climatic and agricultural setting in the Gambia, which is very similar to the study area. tBelcher et al. [9], reported a maximum of cases in the Danfa project villages in Southern Ghana between July and November. Cases were concentrated in the productive age groups between 15 and 54 years. Steib and Mayer [10] examined the epidemioiogy and the vectors of the guinea worm in the area of the present study. Over four years they reported annual incidence r a t e s from two of the current study villages, Toni and Kamadena, between 30/1000 and 81/1000, with a peak in the number of cases in the month of July.
al.
between the dry and rainy seasons: In the rainy season, only 60% of the number of new cases seen in the dry season sought treatment at health centers. The authors did not analyze these findings further, in fact they treated seasonality as a confounder to be controlled in their comparison of health care utilization before and after the introduction of user fees. Fabricant [16], studying the affordability and equity of primary health care costs in rural Sierra Leone, noted a similar drop (by about a third) in health care utilization in the rainy season. Neither of these two studies measured time costs. The present study therefore addresses four questions: (i) How do the household economic parameters (revenues, expenditures, time spent on household production, opportunity costs of time) change seasonally? (ii) What are the seasonal changes in perceived illness? (iii) Does health care utilization change from the dry to the rainy season, and if so in which direction? And finally, (iv) what time and financial costs do households incur in the different seasons, and is there any link with changes in the above three areas?
METHODS
A multiple round household interview survey spanning both the dry and rainy seasons was carried out using a two-stage cluster sample of n = 566 households, comprising 4820 persons. The same households were visited at six different time points: March, April (dry season) and July, September, October and November (rainy season). The household sample was drawn from a demographic surveillance area in the rural Kossi Province in Burkina Faso [17]. Using a recall period of four weeks, the same questionnaire was applied at each survey. Information was gathered on past perceived illness and healer choice as well as the financial and time costs incurred by the households of the sick individuals. Financial costs included households' outlays for user fees, drugs, lodging at a distant treatment site, and transportation. Time costs were measured in terms of the number of full days the sick and any caretakers were not able to perform their customary work (or play, in the case of children) due to illness. Time could be lost due to (i) illness inflicted incapacity, (ii) treatment associated travel and waiting time for both the sick and any caretaker. Days lost for work were converted into monetary terms as follows: economic theory equates the opportunity cost of time with the marginal cost of labor (MCL). The latter is usually approximated with the shadow wage-rate [18]. In the study area a labor market--albeit very small--does exist throughout the year. In order to obtain an estimate of the going
Seasonal variations of illness costs wage-rate, a hypothetical question was asked to the household heads: "If you were to hire somebody to work for you in the fields today, how much would you pay him per day?" In addition, a time allocation study based on the recall of "yesterday's activities" [19, 20] was carried out on the same household sample in both seasons (in March and in June 1992). With minor modifications, we used the same activity categories as the ones used by McSweeny [21] in her time allocation study of Mossi households in Burkina Faso. Household production was defined as any activity serving the households livelihood. It included agricultural production and animal husbandry, food processing and preparation, craft production, and household work such as fetching water and firewood.* The household costs of illness included three components: (i) the time cost of sickness: The opportunity cost of wages foregone by the sick person due to sickness. This includes the time costs of seeking treatment, i.e. travel and waiting time; (ii) the time cost of caretaker(s): The opportunity costs of healthy household members' time spent on treating or tending to the sick or accompanying him/her to the place of treatment. This includes the time costs of caring for or accompanying the sick to the treatment site; (iii) the financial costs of treatment: These are expenditures incurred by the household in seeking care for the sick member. They include the out-of-pocket expenditures for treatment, fees, drugs, transport, and the cost of subsistence at a distant treatment site. Total costs of illness in the household sample during the one-month recall period were calculated as follows: Let: F = total financial costs of health care in the preceding month (F CFA)t Fd = financial costs of drugs, herbs etc. (F CFA) F I = financial costs of fees (F CFA) F,r = financial costs of travel (F CFA) F~ = financial costs of subsistence (F CFA) T = total time costs in preceding month (days of forgone production) Ts = time costs of the sick person (days of forgone production) T~ = time costs of the caretaker(s) (days of forgone production) w = daily wage-rate n = number of illness episodes
*The time spent walking was included in the time spent fetching water or firewood. tF CFA--Francs Communaut6 du Franc Africain (currency of francophone West Africa). :~For a detailed description of the activity categories the reader is referred to Sauerborn [22].
283
a = age coefficient s = related to the sick individual = related to the caretaker(s) then: 1. Financial costs of illness:
F=
+ Ft + F,r + F, i
)
(1)
2. Time cost of illness:
T=i~o[(T,*a,*w)+(Tc,*ac*w)]
(2)
3. Economic cost of illness:
RESULTS
Seasonal changes in the household economy Illness related work incapacity is likely to lead to production loss. Since indirect illness costs are defined as the value of forgone production, it is essential to understand who produces how much and in which season. First, we describe household production and identify differences in the type and volume of production across gender and age groups as well as between seasons. Second, we develop a method to estimate the monetary value of a day of production lost. Household production. Household production was subdivided in four categories::~ (i) farming and animal husbandry; (ii) food processing (e.g. grinding cereal grains, producing peanut or karit6 butter); (iii) crafts and (iv) housework (cooking, fetching water and firewood). The amount of time allocated to each of these four main production activities changed significantly between the dry and the rainy seasons. The average number of hours allocated to farming increased from 0.4 to 3.8 hours per day for women and from 1.9 hours to 6.0 hours per day for men (Table 1). In the dry season, women's contribution to household production was significantly higher than that of men, while the gender differences vanished in the rainy season: men work 6.6 hours and women 6.8 hours per day. Figure 1 displays the average number of hours worked per day at different ages. The age differences in daily work time were highly significant (P < 0.00001, analysis of variance). For each five year age group between 5 and 70 years, the seasonal differences of the hours worked on productive activities were statistically significant (P < 0.00001, analysis of variance). Hence we must assume that more production is lost when a 25 year old is incapacitated by illness compared to the loss of the same amount of time in a 9 year old (or a 70 year old).
R. Sauerborn et al.
284
Table 1. Average hours per day spent on household production, by gender and season. The time allocation study is based on N = 3086 cases~for the dry season survey and N = 2805 cases in the rainy season Average time per day spent on production Dry season Household production Farming/animal husbandry
Between-season differences
Rainy season
Dry vs Women Men Men vs women Women Men Men vs women rainy season hours/day hours/day P (anova) hours/day hours/day P (anova) P (anova) hours/day hours/day Men vs women hours/day hours/day Men vs women Between-season
Farming/animal husbandry 0.4 1.9 0.0001 3.8 6.0 0.0001 0.0001 Food processing 1.4 0.1 0.0001 0.7 0.08 0.0001 0.0001 Crafts 1.I 1.3 ns 0.5 0.3 ns 0.0001 Housework 2.8 0.5 0.0001 1.8 0.2 0.0001 0.0008 Total production 5.7 3.8 0.0001 6.8 6.6 ns 0.0001 ~Timeallocation was only asked for household members above the age of 4 years. The excluded age group of under five years represented 946 individuals (19.6% of the total population). For 788 individuals (16% of the total) no time allocation was obtained.
To adjust for these differences in age-specific productivity when calculating individual opportunity costs of time, a weighting procedure was used as follows: the average number of hours/day spent on household production in the rainy period between ages 15 and 39 years* was used as the maximum work input (weight = 1.0). The proportion of this reference value that a given age group contributes to household production is the weight. The weights obtained that way are similar to the age preference assumed in estimations o f the economic burden of disease [23]. These coefficients were used in the calculation of the opportunity costs of time: the going wage-rate (w) was multiplied by the coefficient (a) corresponding to the age o f the person who incurred time costs [see equation (2) above]. Household expenditures and revenues. In the dry season, households' revenue exceeded expenditures. Towards the end o f the season (April) both revenue and expenditures fell markedly. While revenue
*Both sexes are combined in this analysis.
continued to fall--reaching its lowest point just before the harvest, expenditure climbed a g a i n - fuelled by food purchases [22]. Throughout the rainy season, until the first harvest was in (October), households remained in a negative cash balance, i.e. expenditures exceeded revenues which made it increasingly difficult to absorb health care expenditure (Fig. 2). Opportunity costs o f time. The mean of the distribution was 289 F C F A (SD = + 83 F C F A ) for the dry season (March) and 379 F C F A ( + 105 F C F A ) for the rainy season (July). The difference between both wage rates was highly significant (P < 0.00001, Anova).
Seasonal changes in illness perception The onset o f the rainy season brought two significant changes in the perception of illness, affecting household costs of illness. Firstly, fewer individuals perceived an illness in the previous month, the number of perceived illness episodes fell from 867 to 636 (Table 2). Secondly, the episodes
8.00 7.00 6.00
$.00 4.00
3.1111 ~
2.1111 1.00 0 gY
10-14Y
20-24Y
30-34Y
40-~AY
50-[~l.Y
60-64Y
Fig. I. Household production by age group and season, both genders combined.
> ffi 70 Y
Seasonal variations of illness costs
285
18,000 16,000 14,000 es
o
<
Harvesting
12,000
10,000
8000
"',,,....,....,~
""
~'.,..,
a
Expenditure/household
•
Revenue/household
°A.,
2000 0
I
Mar
Apt
II
I
Jal
II
l
l
I
Sep
Oct
Nov
Fig. 2. Synopsis of seasonal fluctuations of household expenditures and revenues.
perceived were reported as less severe. The proportion of illnesses judged by the sick as severe dropped significantly from 36% in March (dry season) to 8% in October (harvest season). Since the mean financial cost per treatment episode was significantly less ( - 47%) for an illness perceived as not severe (Anova, P < 0.05), the seasonal change in perceived illness severity had an additional effect on the financial cost o f illness. The same applied for time costs. The mean time loss was 0.7 days for an episode of mild illness and 6.1 days for a severe illness. The difference o f these means was significant (P < 0.0005).
Seasonal changes in health care utilization The average financial cost per treatment episodes* showed large variations depending on the choice of t r e a t m e n t - - f r o m 76 F C F A / t r e a t m e n t episode for home treatment to 6939 F C F A / t r e a t m e n t episode for hospital treatment. In the rainy season, households reduced the financial and time costs o f illness through two changes in health seeking behavior: firstly, much *For a detailed analysis of costs by choice of treatment in the dry season only, the reader is referred to Sauerborn et al. [5].
less use was made o f any form o f treatment: while 674 of the 1050 (64.2%) illness episodes in the dry season received some kind o f treatment (including self-treatment), only 259 of the 752 ill individuals (34.4%) were treated at all (Table 2). Secondly, households shifted their treatment choice from high-cost to low-cost alternatives. Table 3 shows that both hospital and dispensary care, as well as the time-consuming treatment of traditional healers, were chosen less frequently. H o m e treatment with its low time and financial costs was chosen in about two thirds o f illness episodes in the rainy season
Seasonal changes in the household costs of illness The combined effect of fewer perceived illnesses, o f which a smaller proportion was perceived as severe, a reduced propensity to seek treatment, a n d - - i n cases where treatment is sought at a l l - - a shift in health care choice towards low-cost treatment options led to a sharp d r o p - - f r o m 4002 to 1065 F C F A (i.e. by 7 5 % ) - - i n the average monthly household costs of illness in the rainy season. Figure 3 shows the seasonal differences in household costs o f illness for each o f the three cost components. We can draw two conclusions from this graph: (i) the mean values of all cost components were reduced
Table 2. Seasonal differences in the numbers of ill household members, perceivedillnessepisodes, number of those treated, and the number of treatment episodes Numbers Description of variable Dry season Rainy season No. of households 566 547 No. of individuals 4820 4634 No. of ill individuals (in % of study population) 867(1~1%) 636 (13.7%) Avg. no. of episodes/iUindividual 1.21 1.18 No. of illness episodes 1050 752 No. of illness episodes treated (in % of all illness episodes) 674 (64.2%) 259 (34.4%) No. of treatment episodes/illnessepisode 1.22 1.6 No. of treatment episodes 829 282
286
R. Sauerbom et al. Table 3. Seasonal differences in treatment choice and expenditure by age group and season. The cost differences between treatment choices are significant (P < 0.00001) Utilization (%) by season Treatment choice Hospital care Nurse, outpatient care Traditional healer Village health worker Home care
Average/costs/treatment F CFA (SD)
Rainy
Dry
6939 (17,261) 1432 (2560) 656 (1965) 221 (462) 76 (245)
2.1 8.5 11.7 3.5 64.9
7.4 10.1 15.4 4.6 49.0
in the rainy season compared to the dry season;* (ii) the relative size of the three components of economic costs changed from the dry to the rainy season: whereas in the dry season the relative contribution of all three components to total household costs was about the same (one third), time costs of sickness amounted to 57% in the rainy season, time costs of the caretaker to 24%. The strongest relative reduction was in financial costs: The average household financial costs of illness was reduced more than six-fold from 1250 to 197 F CFA. They contributed only 18.5% to total economic costs of illness in the rainy season (instead of 31.2% in the dry
seasons no statistically significant differences were found. DISCUSSION
Most field studies are carried out in the dry season--for logistical, transportation and convenience reasons. However, great caution is warranted when findings from the dry season are generalized. Our study shows that estimates based on extrapolations form a single survey carded out in the dry season will overstate household health care consumption and both the time and financial costs of ill health. season). Had we based our estimates of the household costs W h e n controlling for severity (Table 4), we found of illness on our March survey alone (dry season), we that households spent on average significantlylesson would have calculated the annual economic costs of episodes of severe illnessesin the rainy season than in illness to be 12 x 4002 F CFA = 46,024 F CFA. The the dry season (448 and 957 F C F A respectively, seasonally adjusted annual cost estimates, however, P < 0.005). In addition, households significantly were only 30,401 F CFA. Chambers [1] coined the reduced the amount of time healthy caretakers spent terra "seasonal bias" for such distortions arising from to tend to the severely ill:from 7 to 5.8 hours/illness the researcher's tendency to carry out studies in the episode on average (P < 0.05). W h e n the three cost convenient dry season and making linear projections components were compared for mild illnessesacross of health or economic variables that fluctuate substantially with season. The authors therefore recommend that studies assessing the household costs *It is important to note that all cost variables show a highly of illness, the demand for health care, the skewed distribution. This is due to the fact that (i) illness clusters in a relatively small number of households. Of willingness-to-pay or the economic impact of specific those, only a minority incurs any costs at all. In the dry diseases be carried out in all relevant seasons of the season survey, for example, of the 566 households, year. 69.4% (393) reported an illness in the preceding month. Sauerborn et al. [5] stressed the importance of t i m e Of these, 72% (281) households treated the illness. Of costs within total costs of illness. While this earlier these 281, only 61.9% (174) incurred any financial costs. So the financial burden of illness is borne by less than analysis was confined to a dry season survey, the a third (174 households or 30.7%) of all households [22]. findings presented in the current paper point out to
~1400
1395
1356
rJ r~ 1200 ~'-' 1000 =
~
[] Dry season
BOO ~OO
[] Rainy season
~ 4OO .~ 2OO
~
o
Time
Time
Financial
cost of illness
cost of
cost of taro
caretaker
Fig. 3. Seasonal changes of the three components of household costs of illness. Time cost of illness (due to time lost by the sick individual), time costs of the caretaker (due to time lost by household members tending to or accompanying the sick) and the out-of-pocket expenditures the household incurs to treat the sick (financial costs of care).
Seasonal variations of illness costs
Type of cost Financial costs Time costs (sick) Time costs (caretaker) *P < 0.05, **P < 0.005
287
Table 4. Costs per illnessepisode by season, controllingfor severity of illness. Costs incurred per illnessepisode (F CFA) Mild illness Severe illness Unit Dry Rainy Dry Rainy F CFA 500 (1546) 483 (1489) 957** (2632) 448** 0337) Days 1.4 ( _+8.4) 1.3 (8.4) 7.0 ( _ 16.7) 5.8 00.3) Days 0.9 ( _+4.8) 0.4 (3.1) 5.6* (18.4) 1.9" (4.9)
an even bigger relative importance of time costs in the rainy season. In spite of higher opportunity costs of time in the rainy season,* households lost less production due to the fact that fewer people perceived themselves ill, and of those, fewer assumed the sick role, i.e. stayed away from work. It is interesting to note that households reduced caretaker time substantially more than the time lost by the sick. The relative share of time costs of all rainy season costs of illness of the household grew from 68% in the dry season to 82% in the rainy season. This was due to the fact that households curbed the financial component of their illness costs even more drastically. As far as policy is concerned the most troublesome finding of our study is that households' utilization of health, especially of formal Western-type health services, was significantly lower in the rainy season. This finding is corroborated by data from Cameroon reported by Litvack and Bodart [15] who found a 40% reduction of the n u m b e r of new cases seen at health facilities in the rainy season. Fabricant [16] made a similar observation: he found that utilization rates of health services decreased significantly (by 25%) while home treatment increased. Given that in Burkina health service utilization in the dry season is already very l o w - - o n l y about 1 in 5 illness episodes are treated by professional health services--these services become almost irrelevant in the rainy season, just at a time when health needs are greatest. Obviously, health care interventions with proven, impressive efficacy cannot be expected to have any significant health impact as long as people do not use them. We support Fabricant's argument that "the drop in utilization was presumed to be due to the seasonal effect of lower ability of cash". He estimated that about l/3 of the utilization drop was attributable to poor physical access while the remainder was attributable to reduced availability of cash.t In addition to the financial costs of seeking care, the high opportunity costs of time are likely to make
*An observation confirmed by Mwabu's [18] study on the seasonal changes in the shadow price of time in Kenya. tFabricant [16] found that, in the rainy season, the average household's cash on hand was only about 1/10 of that in the dry season. :[:The economic access barriers of health care are compounded by physical barriers such as muddy or flooded roads and collapsed bridges, but these physical access barriers are not the subject of this paper.
health services economically less accessible in the rainy season. Mwabu [18] argued: the rise in the value of time in the wet season raises the costs of medical services and thus, other things being equal, the rate of their utilization falls. Households are, of course, not static entities. They can be expected to react to the economic costs of ill health, mobilize financial resources and substitute health. These "coping strategies" are discussed in depth in a separate paper [24]. Given that in the rainy season, household revenue is at its lowest and opportunity costs of time at its highest, we suggest that one way of increasing resort to health services in the rainy season would be to reduce the excessive economic access costs:~ during that period. One way to reduce the time costs to health care would be to strengthen health care capabilities within the village or even within the household. Since the obvious policy option, namely to build, staff and equip more rural health facilities, is not feasible in Burkina Faso, and in view of the failure of the "community health worker" scheme [25-27], the Ministry of Health is currently testing an alternative way of reducing both financial and time access costs by strengthening mothers' home treatment. Given that mothers provide the bulk of primary health care using traditional medicine [27, 28], the policy being developed consists of teaching mothers to incorporate carefully selected diagnostic and treatment tasks currently performed by nurses into their home treatment practices. The feasibility of mothers treating dehydration with home-made oral rehydration solution has been show by many authors [29-3 l]. Other tasks that mothers could assume--at least in the rainy season--include recognizing and treating malaria (oral chloroquine) and recognizing and treating acute respiratory infections. A study assessing the feasibility of teaching mothers to include Western type health care in their home care practices is currently under way in the study area. Another policy option currently implemented by the Burkinian Ministry of Health [32] in the study area is to reduce the time costs associated with travel by establishing pharmacies in close proximity to dispensaries so that patients no longer incur travel costs to fill prescriptions in the district capital. Whatever route is taken to reduce time costs of access to health care, research is needed to analyze the effect of these policies on household costs of illness and on health care utilization.
288
R. Sauerbomet al.
As regards policies aiming to reduce the financial costs of illness, the current study has shown that professional health services account for the largest share of households' expenditures for care, with drug expenditures ranking first. Since October 1993, Burkina's Ministry of Health has offered generic drugs through co-managed community revolving drug funds at prices below those of private pharmacies. This policy, inspired by experience elsewhere in sub-Saharan Africa (for a review see [33, 34]) has been shown to reduce the total financial cost of a health care visit and to increase health care utilization [15]. Most importantly, schemes to finance health services should address the economic burden of rainy season health care. One possibility would be to build "seasonal cross-subsidies" into existing price schedules of health services. This would mean charging lower prices for services and drugs in the rainy season and higher prices in the dry seasons, as suggested by Fabricant [16]. Another option consists in dissociating the time of payment from the time of service provision. This appears to be a logical way to take the households' fluctuating availability of cash into account (see Fig. 2). Prepayment schemes that collect premiums after the harvest season and cover basic preventive and curative care throughout the year would also share the costs between the healthy and the sick. Risk-sharing prepayment schemes have long thought to be impractical in rural sub-Saharan Africa.* However, they have recently received increasing attention [35-40]. The few documented ongoing prepayment schemes stress the need for community involvement, the use of funds--inter alia--for quality improvement of services and the need for political commitment from national policy makers to support the schemes [35]. Further research is suggested to evaluate the effect of prepayment schemes on health care utilization and on household cost of illness especially in the rainy season. We agree with Mwabu [18] that household level studies which assess the full financial and time cost of the household as well as its health seeking pattern and which capture the seasonal fluctuations are essential for the design of effective health policies. The most surprising--and probably the most controversial--finding of this study is that illness perception--both in terms of the frequency and severity of illnesses---dropped significantly from the dry to the rainy season. These changes in illness perception during times of seasonal stress were shown to be the most important factor in reducing the cost burden of the household in the rainy season. In Sierra Leone, Fabricant [16], too, observed a drop in the frequency of overall illnesses, ranging from 6-23%. This is in contrast to reports of an increase in the
*For a review see Carrin [41] and Shaw and Ainsworth [34].
frequency of biomedicaUy verified disease reviewed in the introduction, a point also made by Fabricant [16] who was surprised that "morbidity in the rainy season was not higher than in the dry season". As to the severity of disease in the rainy season, Fabricant [16] is the only author we could identify who reports a drop of the number of illnesses perceived as severe in the rainy compared to the dry season. The great majority of authors, however, report a rise in the proportion of severe diseases. An increase in monthly death-rates in the rainy season was reported for malaria in Burkina Faso [42] and in Gambia [8]; for infant deaths in Senegal [43]; and for all causes of death from Matlab, Bangladesh [44]. Three hypotheses can be forwarded to explain the apparent contradiction between a higher burden of biomedical disease in the rainy season as reported from studies in similar settings and a decrease in the burden of perceived illness in our study population: (i) It could be argued that there were indeed fewer and less severe diseases during the rainy season in the study area. However, although the present study did not measure clinical disease as defined by examination or laboratory tests, it is hard to believe that the epidemiological situation in the study population would be radically different from those described above, given the environmental and socio-economic similarities between study settings noted above. In conclusion, it is implausible to accept that people are indeed healthier during a period of reduced energy intake, peak energy expenditure, a negative energy balance, and increased transmission of major diseases [1]. (ii) Systematic under-reporting of illness episodes could have occurred during the rainy season. However, this is unlikely since hospital and dispensary records, too, showed a substantial decline of health care use in the study area. The close supervision of interviewers--one in ten interviews was chosen at random and redone by the interviewer's supervisor--makes it very unlikely that interviewer fatigue, i.e. their decreasing willingness to comply with the interview procedures, was the underlying reason. (iii) Finally, the drop in illness perception could be a strategy, if an unconscious one, to cope with simultaneous seasonal stress. A variety of factors known to reduce the sicks' propensity to seek care (such as muddy roads, unavailable transportation) are not likely to reduce illness perception per se. The authors argue that it is plausible that the high opportunity costs of time and the low availability of cash and assets in the
Seasonal variations of illness costs rainy season lead to lower illness perception. As Kroeger [45] states: . . . in peasant societies, where the demand for health services depends strongly on their availability, and where everybody's manpower is needed (particularly during the harvest period), entering the sick role is almost prohibited. In other words, people simply do not have time to be sick in the rainy season. This, in turn, leads to fewer days off work (i.e. a lower proportion of ill individuals assume the sick role), less time spent by caretakers on the sick, and less financial costs incurred for treatment. The implications of our tentative conclusion, that people indeed "ignore disease" to avoid costs in times of concurrent seasonal stress, would be serious from a public health view point: First, while "ignoring" disease may be effective in reducing costs in the short run, it carries the risk of causing disproportionately high expenditures in the future. Consider the case of inguinal hernias: they can be ignored as long as they do not incarcerate. When they do, they are b o u n d to surpass even the highest illness perception threshold, and entail all the pain and costs related to a medical emergency. Second, untreated disease may lead to negative externalities. For example, infectious diseases, if untreated, may spread to healthy households members. In their work on tuberculosis, Stanton and Clemens [46] stressed the high costs that untreated disease inflicts on households. Third, we would have to think about new ways to influence illness perception, i.e. increase people's awareness of (biomedical) diseases they may have. We readily acknowledge that our hypothesis that changes in illness perception are driven by the changing economic parameters individuals and households face, while plausible and consistent with our findings, is by no means proven to be correct by the current analysis. In order to test this hypothesis, further research is required which collects prospective and simultaneous data on: (i) illness perception-through interviews, (ii) disease--through some objective test for key diseases, such as a thick blood film for malaria or a urine test for schistosomiasis etc. and (iii) n u m b e r of cultural and socio-economic factors, notably the household's opportunity costs of time, thought to influence illness perception and health care choice.
Acknowledgements--This study was part of the "Burkina
Health Care Intervention Study" financed by the European Commission, Directorate General XII (Science, Research and Development) under contract #TS3*CT92*0078) and carried out by the Ministry of Health of Burkina Faso (as principal investigator), the Medical Faculty of Ouagadougou University, ORSTOM (Organisme de la Recherche et la Technologic pour le Developpement), Paris, and the Institute of Tropical Hygiene and Public Health of Heidelberg University. In addition, the Gesellschaft fiir Technische Zusammenarbeit (GTZ) and the Deutscher
289
Entwicklungsdienst (DED) actively supported the study. We wish to thank the entire staff of this research project for their untiring work. We would like to thank Uwe Brinkmann~, Peter Berman, Allan Hill, and Joseph Maxwell, with whom we had many stimulating discussions. We are deeply indebted to the households we interviewed, for their hospitality, and the patience with which they answered the many questions we posed. REFERENCES
1. Chambers R., Longhurst R. and Pacey A. Seasonal Dimensions to Poverty, pp. 1-7. Pinter, London, 1981. 2. De Ferranti D. Paying for Health Services in Developing Countries. World Bank Staff Working Papers No. 721, Washington, D.C., 1985. 3. Gertler P. and Van Der Gaag J. The Willingness to Pay for Medical Care. Johns Hopkins Press, Baltimore M.D. 1990. 4. Heller P. A model of the demand for medical and health services in peninsular Malaysia. Soc. Sci. Med. 16, 267, 1982. 5. Sauerborn R., Ibrangho I., Nougtara A., Borchert M., Koob E., Hien M. and Diesfeld H. J. The economic costs of illness for rural households in Burkina Faso, Trop. Med. Parasitol. 42, 219, 1995. 6. Brandt H. Work Capacity Restraints in Tropical Agriculture Development. Peter Lang, Frankfurt, 1980. 7. Bleiberg F., Brun T. A., Goihman S. and Gouba E. Duration of activities and energy expenditure of female farmers in dry and rainy seasons in Upper-Volta. Br. J. Nutr. 43, 71, 1980. 8. Greenwood B. and Picketing H. A review of the epidemiology and control of malaria in The Gambia, West Africa. Trans. R. Soc. Trop. Med. Hyg. 87($2), 3, 1993. 9. Belcher D. W., Wurapa F., Ward W. and Lourie I. Guinea worm in southern Ghana: its epidemiology, and impact on agricultural productivity. Am. J. Trop. Med. Hyg. 24, 243, 1975. 10. Steib K. and Mayer P. Epidemiology and vectors of Dracunculus medinensis in Northwest Burkina Faso, West Africa. Ann. Trop. Med. Parasitol. 82, 189, 1988. 11. Drasar B., Tomkins A. and Feachem R. The seasonal ecology of disease: Diarrheal diseases. In Seasonal Dimensions to Poverty (Edited by Chalmers R., Longhurst R. and Pacey A.), pp. 102-111. Pinter, London, 1981. 12. Chen L., Chowdhury A. and Huffman S. Seasonal dimensions of energy protein malnutrition in rural Bangladesh. Ecol. Food Nutr. $, 175, 1979. 13. Rowland M. G. M. Assessment of health programme effects with longitudinal studies. Biosoc. Sci. 10, 87, 1989. 14. Sutton R. The seasonal ecology of disease: respiratory diseases. In Seasonal dimensions to poverty (Edited by Chalmers R., Longhurst R. and Pacey A.), pp. 112-114. Pinter, London, 1981. 15. Litvack J. I. and Bodart C. User fees plus quality equals improved access to health care: results of a field experiment in Cameroon. Soc. Sci. Med. 37, 369, 1993. 16. Fabricant S. J. Community financing in Sierra Leone: affordability and equity of primary health care costs. PhD thesis, London University, 1992. 17. Sauerborn R., Nougtara A. and Diesfeld H. J. Action research on health services in Burkina Faso: concept and methodology 1988-1992. In Proceedings of the Conference on "Methodology and Relevance o f Health Systems Research" organized by the International
Children's Centre. Paris, 8-10 April 1992. 18. Mwabu G. Seasonality, the shadow price of time and effectiveness of tropical disease control programs. In
290
19.
20. 21. 22.
23. 24. 25.
26.
27.
28.
29. 30.
31,
32.
R. Sauerbornet al.
Economics, Health and Tropical Diseases (Edited by Herrin A. N. and Rosenfield P. L.), pp. 259-270, 1988. White B. Measuring time allocation, decision making and agrarian changes affecting rural women: examples from recent research in Indonesia. Paper for the Methodolgy Workshop "Research on Rural Women", IDS, 12-14 March 1982. University of Sussex, 1983. Acharya M. Time use data and the living standards measurement study. Worm Bank Working Paper 18, 1-75, 1982. McSweeny B. Collection and analysis of data on rural women's time use. Stud. Family Plan. 10, 379, 1979. Sauerborn R., Nougtara A., Benzler J., Borchert M., Diesfeld J., Hien M., Ibrango I. and Maiga M. Strategies d'Adaptation des M~nages aux Co~ts Economiques de la Maladie. Lang Verlag, Frankfurt, 1995. World Bank. Worm Development Report: Investing in Health, p. 26. World Bank, Washington, D.C., 1993. Sauerborn R., Adams E. and Hien M. Household strategies to cope with the economic costs of illness. Soc. Sci. Med. 43, 291, 1996. Nougtara A., Sauerborn R. and Diesfeld H. Assessment of MCH-services offered by professional and community health workers in the district of Solenzo, Burkina Faso. (I) Utilization of MCH-services. J. Trop. Pediat. 35(Suppl.), 2, 1989. Kabor~ B. A. Soins de Sant~ Primaires au Burkina Faso: Centre de Sant~ et de Promotion Sociale et/ou Poste de Sant6 Primaire. Thesis, University of Antwerp, 1985. Sauerborn R., Nougtara A. and Diesfeld H. Low utilization of community health workers--results from a household interview survey in Burkina Faso. Soc. Sci. Med. 29, 1163, 1989. Sauerborn R. and Nougtara A. Traditional medicine is practiced by mothers rather than by traditional healers--a study in Solenzo, Burkina Faso. Curare 15, 97, 1992. Cutting W. Can village mothers prepare oral rehydration solution? Trop. Doc. 8, 195, 1979. Chen L., Black R., Merson M., Bhatia S., Yunus M. and Chakraborty J. Village-based distribution of oral rehydration therapy packets in Bangladesh. Am. J. Trop. Med. Hyg. 29, 285, 1980. Chowdhury A., Vaughan P. and Abed F. Use and safety of home-made oral rehydration soutions: an epidemiological evaluation from Bangladesh. Int. J. Epidemiol. 17(3), 655, 1988. Ministate de la Sante. Document National sur le Renforcement des Soins de Sant~ Primaires au Burkina
33. 34.
35. 36.
37. 38. 39.
40. 41. 42.
43.
44.
45. 46.
Faso--projet de Dbmarrage de rlnitiative de Bamako. Minist6re de la Sant6, de l'Action Sociale et de la Famille, Ouagadougou, Burkina Faso, 1992. McPake B., Hanson K. and Mills A. Community financing of health care in Africa: an evaluation of the Bamako initiative. Soc. Sci. Med. 36, 1383, 1993. Shaw R. P. and Ainsworth M. Financing health services through user fees and insurance: lessons from Sub-Saharan Africa. (Draft manuscript). World Bank, 1994. Chabot J., Boal M. and Da Silva A. National community health insurance at village level: the case from Guinea Bissau. Hlth Pol. Plan. 6, 46, 1991. Baza A., Hakizimana A., Hanson K., Kwizera F. and Van Der Geest S. Health insurance and the Bamako initiative in Burundi: value for money. Int. J. HIth Plan. Mgmnt 8, 129, 1993. Arhin D. C. The health card insurance scheme in Burundi. A social asset or a non-viable structure? Soc. Sci. Med. 39, 861, 1994. Arhin D. C. Health insurance in Africa. Lancet 345, 44, 1995. Shepard D., Vian T. and Kleinau E. F. Impact of four health insurance programs in rural and urban areas of Za'ire. In Financing Health Services Through User Fees and Insurance: Lessons From Sub.Saharan Africa (Edited by Shaw R. P. and Ainsworth M.), pp. 249-286, World Bank, 1994. Abel-Smith B. and Rawal P. Employer's willingness to pay: the case for compulsory health insurance in Tanzania. Hlth Pol. Plan. 9(4), 409, 1004. Carrin G. Community financing of drugs in SubSaharan Africa. Int. J. Hlth Plan. Mgmnt 2, 125, 1987 Sauerborn R., Shepard D., Ettling M. and Etal A. Estimating the direct and indirect economic costs of malaria in a rural district of Burkina Faso. Trop. Med. Parasitol. 42, 219, 1991. Cantrelle P. Mortalit6 facteurs. In ONED, Afrique Noire, Madagascar, Comores: Ddmographie Comparde. D616gation G6n6rale de la Recherche Scientifique et Technique, Paris, 1967. Dyson T. and Crook N. Causes of seasonal fluctuations in vital events. In Seasonal Dimensions to PoverO' (Edited by Chambers R. et al.), pp. 135-141. Pinter, London 1981. Kroeger A. Response errors and other problems of health interview surveys in developing countries. WId Hlth Star. Q. 38, 15, 1985. Stanton B. and Clemens J. User fees for health care in developing countries: a case study of Bangladesh. Soc. Sci. Med. 29, 99, 1989.