SEARCH

SEARCH BY CITATION

Keywords:

  • household cost;
  • malaria;
  • disease burden;
  • indirect cost

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Summary Short-run economic consequences of ‘malaria’ on households were examined in a household survey in Matale, a malaria-endemic district of Sri Lanka. On average a household incurred a total cost of Rs 318 (US$ 7) per patient who fully recovered from ‘malaria’. 24% of this was direct cost, 44% indirect cost for the patient and 32% indirect cost for the household. Direct costs were greater for those seeking treatment in the private sector. Notably a large proportion of direct costs was spent on complementary goods such as vitamins and foods considered to be nutritional. Indirect cost was measured and valued on the basis of output/ income losses incurred at the household level rather than using a general indicator such as average wage rate. Loss of output and wages accounted for the highest proportion of the indirect cost of the patients as well as the households. Relative to children, more young adults and middle-aged people had ‘malaria’ which also caused greater economic loss in these age groups. Women tended to care for patients rather than substitute their labour to cover productive work lost due to illness. We compare the methods used by other researchers for valuing indirect cost, demonstrating the significant impact that methods of measurement and valuation can have on the estimation of indirect cost, and justify the recommendation for methodological research in this area.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This paper is a response to recently renewed interest in the economic consequences of malaria ( Shepard et al. 1991 ; Mills 1993, 1994; Ettling et al. 1994 ; Asenso-Okyere & Dzator 1997), an area where little work has been undertaken. It is quite clear that malaria places an enormous economic burden on the households of economically vulnerable people living in endemic regions in developing countries, both in terms of costs of securing treatment as well as loss of output and wages. The value of such information to policy makers and managers of control programmes is that researchers have claimed that greater importance needs to be given to control of this over other diseases, largely because of potential improvements in the productive capacity of the economy and therefore general economic development ( Shepard et al. 1991 ).

To date, most studies have imputed average wage rates from the local formal economy ( Asenso-Okyere & Dzator 1997; Konradsen et al. 1997 ) or estimated average household production in an area ( Sauerborn et al. 1991 ; Shepard et al. 1991 ). However, as most patients live within a range of informal economic networks and may themselves be working or not, tools for the measurement and valuation of cost (particularly indirect cost) should be applied with care. At present few researchers have questioned either the methods or assumptions used to provide estimates of the costs of malaria and none have examined the implications of alternative approaches to valuation. The nature of informal and subsistence economies is particularly important to recognize as it means that members of households may or may not enter into production, which itself may or may not be transformed into economic development. The implication of this is that existing figures for economic loss are likely to overemphasize the economic (as opposed to social or otherwise) burden and therefore provide unreliable information to decision-makers with respect to prioritization of resources. To address these issues, it is essential to explore alternative methodological approaches.

This aim of this paper is to measure and value the direct and indirect costs of perceived malaria morbidity at household level. For indirect cost, we use an output-related approach by relating loss of time with loss of output at an individual and household level, and compare the results with other valuation techniques used in the malaria literature. The first part of the paper describes specific features of the methods used in the household survey in Sri Lanka 1. The results show the predominant pattern of attendance of patients at different health care providers, and detail the range of costs of treatment, including direct payments, number of days lost to production and nonproduction as well as the value of lost production. The discussion focuses on the novelty and validity of this household-based output method for the valuation of indirect cost in comparison with other approaches and the implications for decision-makers, planners and researchers.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Matale district in the Central Province of Sri Lanka was chosen because it had the highest annual parasite incidence rate (89.5) reported in 1991. Among the cases detected by the Anti-Malaria Campaign, 72.4% were P. vivax and 27.3% were P. falciparum. From about 212 village-level administrative units (Grama Niladhari–GN Divisions) in the malarious zone in the district, 54 were selected as the sample for the household survey using the probability proportional to size method ( Bennett et al. 1991 ); 20 households were selected randomly from each GN division, making the total sample 1080. Following the results of a pilot survey, the revised questionnaire 2 was used in early 1993 3.

Every attempt was made to gain full co-operation from the households in which the questionnaire was administered. A letter was sent to each household, one week in advance, explaining the objectives of the survey. The principal investigator or a field supervisor visited each village a few days before the survey to see the village administrative officer or, if not available, to visit several villagers for a brief discussion about the survey. When the research team visited the village, they were mostly helped by villagers to locate and introduce the team to the households selected. Each household was visited by a pair of researchers, one to ask questions and the other to record the information.

During the interview, each respondent was asked to think about the previous month and whether or not they, or any one in their household had had ‘malaria’. As blood-testing facilities were only available in some areas, this survey was based on people's perceptions 4. Those that replied ‘yes’ were then asked a series of questions about: their knowledge of and attitudes towards malaria; length of time spent ill and the extent to which their daily activities had been affected; sources and types of treatment when ill; direct costs of receiving treatment; extent of economic consequences from any time not working; and socio-economic and demographic status 5. Based on this information, both direct and indirect costs were measured and valued for fully recovered patients 6.

The direct out-of-pocket costs of receiving treatment, such as travel cost, drugs, medical tests and consultation fee, were recorded for both formal (Western and Ayurvedic) and informal (ritual and self-treatment) sources of treatment. As the pilot study had shown that households incurred other expenses as a direct result of having ‘malaria’, data on all other direct costs (termed ‘complementary cost of treatment’ (CC)) were also collected, and included cost of vitamins, nutritional food, special foods and drinks 7.

The measurement of indirect cost was primarily based on an output-related approach ( Goldschmidt-Clermont 1987). Productive work was broadly defined as involvement in any economic activity with the potential to add to the disposable income (in kind or cash) to the household. Therefore, whenever illness had adversely affected the productive work of any household member, it was measured either in terms of output units or person days. This method excluded time loss of economically inactive patients such as schoolchildren, pre-school children, job seekers and people with learning difficulties. Furthermore, those engaged in housework were also excluded, except on the occasions when they covered the productive work of the household. Both completely and partially disabled days of the patients were taken into account in this exercise 8.

The valuation of indirect cost was based on actual loss of income attributable to illness. Respondents were asked to state whether there was any form of income loss, and if so how much. This could either have been as a result of direct monetary loss, e.g. loss of daily wage of a casual labourer, or as a reduction in farm income or production such as due to being unable to harvest tobacco leaves at the proper time (and hence receiving a lower price) or the destruction of crops by wild animals as fields were left unattended. When respondents provided information about changes in agricultural production (for commercial or subsistence purposes), the market prices of those products in the respective subdistrict were used to calculate the money value. When other household members attempted to cover the loss of productive work of patients or the household members who cared for the patient, the average wage rates (of males and females separately) of the respective subdistrict was used as the opportunity cost 9. This was because, in the event of household members being unable to cover that work, local labour would have to be hired.

Collecting data on income is often difficult, therefore even though a standard form was designed, a series of indirect questions related to income were asked of all except those with fixed salaries. The intention was to make the interview as informal and relaxed as possible. Questions related to income were asked at different stages of the interview in a haphazard manner in order not to worry the respondent about intentions to collect income data 10. For instance, after some time in the interview, a question would be asked rather informally about the previous season's harvest (using, for example, observations about the types of crops grown around them). Then, after some general discussions about the harvest, investigators moved on to more detailed discussions about difficulties in agriculture, with an emphasis on cost. In this way more appropriate information on net income was collected, and after the interview the investigators filled in the details of such information in the questionnaire.

The methods of analysis included descriptive statistics showing the mean and variance in the direct and indirect costs of ‘malaria’ morbidity. Variations in mean costs are also analysed by sources of care. Finally, an attempt is made to compare the effects of different methods of valuing indirect costs by applying similar criteria used by other researchers and re-estimating costs. Exact comparisons are not always possible as, for example, data was not available to give a weight to monthly wages to indicate the demand for labour ( Konradsen et al. 1997 ) or seasonal variations in production ( Sauerborn et al. 1991 ). However, comparisons were possible using Konradsen's classification of age, and assumptions about which members of the population are economically active, as well as Sauerborn's calculations of adult's time spent caring for sick children. Such assumptions were used in association with locally collected data ( Sauerborn et al. 1991 ; Konradsen et al. 1997 ).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The response rate for the household survey was 96% with 5219 respondents in 1038 households; 344 belonging to 270 households stated that they had had ‘malaria’ and fully recovered during the previous month 11. This subsample of households was analysed and the results are presented in the following sections.

Socio-economic background

The majority of patients were adult males with a large proportion (44%) in the 21–45 age group. In this predominantly Sinhala-Buddhist agricultural area, adults tended to be mainly engaged in either farming 12 (35%) or household activities (32%). Excepting pre-school children, 96% had received or were receiving primary education and 53% secondary education. The average monthly income of households was Rs. 4074 (SD = Rs. 3469) 13, with monthly per capita income at Rs. 820 (SD = Rs. 592).

Health-seeking behaviour

On average, a patient made 1.35 visits to a formal source of treatment. Public medical centres were the most commonly used treatment source mainly due to the availability of blood testing facilities at many of them, and as the final resort for severe illness ( Attanayake 1994). However, almost 36% of visits were made to private clinics ( Table 1). The reasons for this tendency include the lack of services after working hours and long waiting times at public medical centres as well as better attention at the private clinics of public doctors. In fact, 22% of patient visits were made to the private clinics of public doctors. Self-treatment was common among all patients, with 97% involved in it before or after formal treatment. Panadol, Disprin/Aspirin and a herbal mixture called paspanguwa were the most popular drugs for self-treatment.

Table 1.  Percentage of visits and average direct cost of receiving treatment by source of treatment and type of cost (Rs. in 1993)* Thumbnail image of

Blood smears were taken at 60% of the patient visits to public sources and 9% in private facilities. At the public medical centres blood smears were subjected to examination only where such facilities were available, otherwise they were sent to the laboratory at the district malaria office for examination. Through the blood tests conducted at both public and private medical centres, 47% of all patients (or 90% of patients given blood tests) were confirmed as malaria patients.

Direct cost

On average, a household incurred Rs. 75 for treatment per patient ( Table 1) which included special food (Rs. 28); treatment cost (Rs. 23); travel cost to the patient (Rs. 9); travel cost to the accompanying person/s (Rs. 8); nutritional food (Rs. 6); blood tests (Rs. 1); and purchasing vitamins (cts. 50 – one rupee = 100 cents). The highest cost per visit (Rs. 91) was for the patients who attended purely private clinics, which happens in 13% of visits. The next highest cost per visit was for the private clinics of public doctors (Rs. 79) with 23% of visits followed by inpatient care at public hospitals (Rs. 55) in 8% of visits 14. Cost of self-treatment and ritual treatment per patient was Rs. 4 and Rs. 6, respectively.

Table 2 demonstrates that 46% of patients sought care only from the public facilities as outpatients, and that 25% sought care from private sources only. Only 11% of patients used a combination of public and private services. This table also shows that, for the same patients, the cost of receiving treatment was Rs. 37 (SD 112) per patient and the cost of complementary goods/services (CC) was Rs. 38 (SD 40). Patients who had received treatment only from private western sources had the highest average cost of treatment (Rs. 86) followed by the combination of public inpatient or outpatient care with private western treatment (Rs. 73). The variations in average cost are a function of the source of care in that there were substantial differences in travel cost to private clinics located far away from the patient's residence as well as in length of stay as inpatients.

Table 2.  Average direct cost of treatment by the combination of sources of treatment (Rs. in 1993) Thumbnail image of

Indirect cost for patients

The number of completely disabled days lost was four per patient (SD = 2.9). The total number of partially disabled days lost was 5.3 per patient (SD = 2.8). The distribution of completely disabled days was more lepto kurtic than partially disabled days (14.01 compared to 3.15), and the former was more positively skewed than the latter (3.04 compared to 1.49). This indicates more uniformity in the period of complete disability than partial disability. For instance, 27% of patients were completely disabled for 3 and 20% for 2 days. Table 3 shows that 39% (= 133/344) of patients were ‘economically active’15. The remainder were either school/ pre-school children, job seekers, elderly or engaged in housework. However, of these 133 people only 59 patients (45%) were affected in terms of lost income, and these were spread through salary earners, agricultural production and businesses. The majority (= 33) lost wages due to absence from a paid job, at an average rate of Rs. 502 (Rs. 124 per economically active patient). However, output losses caused the largest economic burden as eight patients lost, on average, Rs. 2613 (Rs. 157 per economically active patient). Among other losses were 15 people who had to hire labour at an average rate of Rs. 530 (Rs. 60 per economically active patient). Finally, five patients suffered from business losses with an average of Rs. 610 (Rs. 23 per economically active patient). Therefore the value of time lost from productive work was Rs. 364 per economically active patient.

Table 3.  Indirect cost per economically active patient who had recovered fully from ‘malaria’ by occupation/source of income (Rs. in 1993) Thumbnail image of

Indirect cost to household members

Table 4 depicts the mean indirect cost borne by household members other than the patients as Rs. 102 per fully recovered patient. This principally comprised loss of output (Rs. 24), loss of wages (Rs. 20) due to attending to the patient, loss of time and, hence, loss of income due to accompanying the patient for treatment (Rs. 15) and cost of incentives (Rs. 12) for the neighbours, friends, relatives, etc., who helped the household members. Relatives, friends, etc., had helped to cover almost 11% of household heads' unattended work. The other losses (Rs. 31) primarily referred to the opportunity cost of labour substitution among the household members to cover the unattended productive work of the patient, and the household members who cared for the patient. However, the labour substitution within the household was not so high as only 19% of the economically active patients' unattended productive work was shared by other household members. Whilst labour substitution within the household or hiring labour was not possible for 37%, hired labour and/or help from relatives and/or friends was given in 21% of instances. The former is partly affected by the patient's occupation, for example a housewife is unable to take over a husband's skilled job whilst he is ill.

Table 4.  Indirect cost borne by household members, other than the patient, per fully recovered patient (Rs. in 1993) Thumbnail image of

Total cost for the recovery from illness

Table 5 shows the average cost of treatment from an episode of ‘malaria’. This is a combination of the direct and indirect costs borne by patients as well as the costs of malaria treatment borne by public sector providers ( Attanayake 1994). The average cost of treatment for ‘malaria’ was Rs. 375, of which Rs. 80 (21%) was borne by patients directly, Rs. 95 (25%) by the public sector 16, and Rs. 200 (53%) by the patients indirectly. Therefore on average patients bore 74% of the economic costs of illness. The variation of cost burden is also interesting, and Table 5 shows that the indirect cost for patients and families was extremely high for those patients who attended both private and public sources (Rs. 521). Most of these patients were severely ill for a long period and turned to public sources for inpatient treatment after failing to recover from treatment provided in the private sector.

Table 5.  Average costs of treatment of an episode of ‘malaria’ by source of treatment (Rs. in 1993) Thumbnail image of

Comparison of the effect of output-related method with other methods

Values given to time loss in this study on the basis of the output-related method were compared with the application of methods used in a number of other studies to the results of this study and are presented in Table 6. Using the method outlined in this paper, whilst total indirect cost stands at Rs. 83 549 (i.e. Rs. 48 455 ( Table 3) + Rs. 35 094 ( Table 4)), the average indirect cost per patient is Rs. 243. The conventional method of average wage rate is used in four studies ( Jayawardane 1993; Sawyer 1993; Asenso-Okyere & Dzator 1997; Konradsen et al. 1997 ). Whilst two studies have used daily output per adult ( Shepard et al. 1991 ; Sauerborn et al. 1991 ), the other two have used average income per day ( Sharma et al. 1990 ; Ettling et al. 1994 ). First, using the different approaches the total value of the lost days varies from Rs. 16 608 ( Sauerborn et al. 1991 ) to Rs. 321 156 (Sharma et al.), while the value for output-related method stands at Rs. 83 549. Only on two occasions do values come close to the output-related method i.e. Rs. 83 088 for the average wage rate method but only for fully disabled days ( Asenso-Okyere & Dzator 1997) and Rs. 84837 for the average income per day method ( Ettling et al. 1994 ). Similar variations can be observed for the estimate of average cost per fully recovered patient as it ranges from Rs. 48 ( Sauerborn et al. 1991 ) to Rs. 934 ( Sharma et al. 1990 ), while the value for the output-related method stands at Rs. 243. The two methods that gave similar totals also had roughly equal values for the averages. Obviously these differences in total as well as average figures are simply due to the differences in the costing methods adapted from the selected studies, which are explained in the footnotes of Table 6 and examined in the Discussion.

Table 6.  Comparison of alternative methods for valuing loss of time due to malaria (Rs in 1993) Thumbnail image of

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

This study estimated both direct and indirect cost of illness at the individual and household levels. The estimated cost indicates that indirect cost accounted for 76% of the household cost per fully recovered patient, with the remaining 24% covering both treatment and complementary costs. The study has also shown the importance of private sector provision for treating ‘malaria’ at the household level.

These two facts alone suggest that both donors and policy makers need to pay special attention to private as well as public providers of services when formulating control strategies and measures. It also highlights the importance of considering a variety of viewpoints when conducting economic evaluations of health services. To exclude the patients' viewpoint may encourage shifting of costs onto a population already bearing significant financial burdens from the disease. This study also highlights the need to consider both direct and indirect costs borne by patients. When economic evaluations do consider patient costs, they often only value direct out-of-pocket expenses ( Walker & Fox-Rushby 2000). This study shows that to do so in this case would have underestimated household costs by 300%. The implication for policy makers and researchers is that assumptions of negligible indirect costs may encourage the provision of socially nonviable interventions.

Our findings fit in to a mixed set of results on the household costs of ‘malaria’. Ettling et al. (1994) and Sauerborn et al. (1991) found indirect costs to be less than 40% of the average cost per patient, whereas Asenso-Okyere and Dzator (1997) found that direct costs amounted to only 21% in Ghana. However, we recommend caution in interpreting results across such different contexts as we have found a range of methodological differences between studies. A selection of these issues are presented and discussed below.

First, there are notable differences in the inclusion and exclusion criteria bounding these studies. A key difference is whether confirmed cases of malaria are the basis for responses or not. Some ( Kaewsonthi 1988; Ettling et al. 1991 ; Mills 1993, 1994; Konradsen et al. 1997 ) focused on those with results from blood slides, which therefore ignore all those people who do not attend official treatment sources. Others used clinical judgement based on reports of symptoms ( Asenso-Okyere & Dzator 1997). As this study was based on personal perceptions, there is clearly the possibility that some patients do not have clinically defined malaria, and hence may have different (or similar) household costs from malaria patients. However, the main advantage is that a more complete picture of the total burden of the disease is given, as it focuses on all patients in the study area rather than only those who had attended clinics with blood testing facilities 17. It is also the case that, in endemic areas, experience of the disease helps people recognize clinical symptoms at the onset of infection ( Nur 1993). In Sri Lanka, where the adult literacy rate is around 90% and the population has been exposed to an island-wide malaria control programme for over half a century, there is no reason to suppose a large difference and indeed this population has demonstrated a good understanding of the symptoms of malaria ( Attanayake 1994). This is also supported by Jayawardane (1993) who found that, as settlers became familiar with malaria episodes, confidence in their own ability to use malaria tablets increased their sense of control. One of the principal impacts of setting the inclusion criteria as broad as this is that the costs of accessing private care can be obtained, with their corresponding high costs of Rs. 121 for private clinics compared to Rs. 51 for public clinics.

Other factors that may explain differences in published results are related to the population included. First, there was a high proportion of patients belonging to young and middle age groups, similar to the case in Nepal ( Mills 1994), and more adult males were affected than others by the disease. But, this is not similar to many other countries and especially those within Africa ( Shepard et al. 1991 ). Secondly, this study, whilst primarily confined to the household boundaries, did encompass support given by friends and relatives. Such support included provision of treatment and refreshments, patient care as well as cover for productive work, all of which affect the range of costs calculated.

Perhaps the sharpest differences this paper has been able to highlight are the alternative approaches taken by researchers in the measurement and valuation of indirect costs. We adopted an output-related approach, which influenced both the choices of what to measure and how to value it. We first consider issues of measurement 18. Measurement of indirect costs was confined to economically active patients, which excluded time loss by those engaged in housework, schooling and job seeking. This has been a common feature of other studies as well ( Asenso-Okyere & Dzator 1997) and is based on the premise that economic loss is limited to impact on production (and hence Gross National Product (GNP)). However, the exclusion of domestic work in measuring and valuing indirect cost should be treated with caution. As GNP has long been criticised as an inadequate measure of welfare and is criticised by those seeking to widen economic loss to reductions in welfare and national development ( Todaro 1992), this exclusion would indeed underestimate the indirect cost. Under the situations where women play the leading role in caring for patients, particularly for children, and indeed in the case of illness primarily affecting women, exclusion of domestic work from the indirect cost estimates could certainly lead to gender-biased policies.

The measurement of time loss was also affected by the number of days lost as well as the type of day lost (i.e. whether fully or partially disabled). In Matale, although only 10% of adults and 13% of children were severely ill with ‘malaria’, on average a patient was found to lose four fully disabled and 5.3 partially disabled days during an episode of ‘malaria’19. The number of completely disabled days was higher than that reported for Nepal ( Mills 1993), Malawi ( Ettling et al. 1994 ), and Burkina Faso ( Shepard et al. 1991 ). We also estimated the number of partially disabled days and found that it was 25% higher than the number of fully disabled days.

There are two broad choices concerning how to apply values to the time loss: to apply some kind of average price to the whole (or part) of the sample, or to price each person's time loss according to their own cost of lost production. The most common approach has been to adopt a single price and apply it to the whole sample. Thus, Kaewsonthi (1988), Sawyer (1993), Jayawardane (1993) and Asenso-Okyere and Dzator (1997) applied the average wage rate of the study area and Sharma et al. (1990) used mean daily income to value 100% of all time loss measured. Others have adopted different forms of adjustment to this process. For example, Sauerborn et al. (1991) made some adjustments to average product per person on the basis of seasonal variations and Shepard et al. (1991) based their estimates on ‘calculated output per day’. Konradsen et al. (1997) , on the other hand, weighted wage rates to reflect the demand for labour 20. A crucial question to pose of each of these studies is how far an average wage (whether or not weighted) can be considered to reflect the real output loss of a sick farmer. This study has shown that not only was the output of 80% of farmers not affected, but that when a farmer did give less attention to the crop, it could result in a loss very much higher than the average wage times the number of days lost. One can question the relevance of applying local subsistence wages to time loss of those engaged in other occupations. However, the combination of our approach to measurement and valuation led to a small sample (= 59) of people whose time was valued, and within this a few observations had significant leverage on the mean value. But perhaps future research should consider the cause of particular distribution rather than imposing untested assumptions.

Mills (1994) took a different approach, and chose to value any change in hired wage labour and wage losses. However, this will underestimate the impact on farmers. Therefore this study related time losses to actual output/income losses at individual and household levels, instead of using general indicators related to output or income in order to access more reliable values, based on the theory that value is represented by economic production. A clear indication of the difference this can make is seen by re-estimating costs using the assumptions of Ettling et al. (1994) about the age at which people become economically active, which would generate a much higher cost per patient than identified in this study because in Matale almost 50% of the patients between 15 and 20 years were engaged in studies.

The choice of methods to measure and value change is potentially an important explanatory factor of the variation in household costs reported to date. To judge the impact of method on results, this study re-estimated household costs in Matale using the range of assumptions currently represented in the literature. The lowest cost per patient (Rs. 48) was calculated by applying daily output per adult ( Sauerborn et al. 1991 ). The average wage rate method used by Jayawardane (1993) also resulted in lower values because total loss from infected household heads was divided by the total number of patients in the whole sample. However, the average loss per infected household head was high (Rs. 817). This contrasts with the high figures found using Sawyer's approach ( Sawyer 1993) which simply applied the average cost of household heads to all households irrespective of whether the household head was a patient or not, seemingly leading to a substantial over estimate of costs.

Whilst methods appear to have an important impact on results, it is also the case that research has been undertaken in a variety of different countries. In the absence of a carefully designed international study examining cross-cultural similarities and differences, our observations are best considered as hypotheses for future examination. Do these results indicate a different sensitivity to malaria or illness in Sri Lanka? Earlier work ( Caldwell et al. 1989 ) suggested that households in Sri Lanka give priority to steps leading to an early recovery for the patient and less emphasis to covering the productive work lost of the patient or the household member who cared for the patient. This contrasts with Nur (1993) and Bonilla and Rodriguez (1993) who found that women and children, or women alone, bore a relatively high proportion of indirect cost as labour substitutes. Differences can also be seen through the contrasting findings of Jayawardane (1993) in Sri Lanka and Nur's (1993) research in Sudan with respect to the diversity of interactions amongst family members and neighbours and the impact this is likely to have on the type and range of costs borne when a person becomes ill.

This study highlights the importance of including both direct and indirect costs in assessing short-run economic loss from malaria and in studies of cost-effectiveness. It also shows the need for evaluators and funders of programmes to pay special attention to the private sector in formulating control strategies and measures. Viable malaria control programmes need to take into account the whole range of costs faced by households, and current estimates of the impact of malaria on the productive capacity of populations seem to bear little relation to reality, and therefore introduce significant bias. Some of this ‘bias’ could be seem to stem from alternative philosophies of value. However, it would seem that different methods are even used from the same points of thinking. Whilst we are aware that time constraints are likely to mean that few researchers will adopt the detailed approach to costing taken in this study, it is clear that there are significant variations in estimates of cost which have yet to be examined systematically. Our more detailed study has at least indicated the importance that methods have in estimation. Therefore we recommend that researchers either consider the possibility of adopting similar methods, and/or at least clarify the impact that different assumptions of quantity and price can have on their calculations of direct and indirect cost.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Research upon which this paper is based received support from the UNDP/World Bank/WHO Special Programme for Research and Training in Tropical Diseases. We also thank two anonymous reviewers for their careful commentary.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Asenso-Okyere WK & Dzator JA (1997) Household cost of seeking malaria care, a retrospective study of two districts in Ghana. Social Science and Medicine 45, 659 667.DOI: 10.1016/s0277-9536(96)00383-8
  • Attanayake N (1994) Cost-effectiveness of anti-malaria activities in Sri Lanka. PhD Thesis,
  • Bennett S, Wood T, Liyanage WM, Smith DL (1991) A simplified general method for cluster sample surveys of health in developing countries. World Health Statistics Quarterly 44, 98 106.
  • Bonilla E & Rodriguez A (1993) Determining malaria effects in rural Colombia. Social Science and Medicine 37, 1109 1114.
  • Caldwell J, Gajanayake I, Caldwell P, Peoros I (1989) Sensitization to illness and the risk of death: an experience for Sri Lanka's approach to good health for all. Social Science and Medicine 28, 365 379.
  • Ettling MB, Thimasarn K, Shepard DS, Krachaiklin S (1991) Economic analysis of several types of malaria clinics in Thailand. Bulletin of World Health Organization 68, 467 476.
  • Ettling MB, McFarland DA, Schultz LJ, Chitsulo L (1994) Economic impact of malaria in Malawian households. Tropical Medicine and Parasitology 45, 74 79.
  • Goldschmidt-Clermont L (1987) Economic evaluation of unpaid household work: Africa, Asia, Latin America and Oceania. Women, Work and Development, no. 14. International Labour Organisation, Geneva.
  • Jayawardane R (1993) Illness perception: social cost and coping-strategies of malaria cases. Social Science and Medicine 37, 1169 1176.
  • Kaewsonthi S (1988) Internal and External Cost of Malaria Surveillance in Thailand, II Final Report. TDR project no. 850059 . World Health Organization, Geneva.
  • Konradsen F, Van Der Hoek W, Amarasinghe PH, Amarasinghe FP (1997) Measuring the economic cost of malaria to households in Sri Lanka. American Journal of Tropical Medicine and Hygiene 56, 656 660.
  • Mills A (1993) The household cost of malaria in Nepal. Tropical Medicine and Parasitology 44, 9 13.
  • Mills A (1994) The economic consequences of malaria for households: a case-study in Nepal. Health Policy 29, 209 227.
  • Nur ETM (1993) The impact of malaria on labour use and efficiency in the Sudan. Social Science and Medicine 37, 1115 1119.
  • Sauerborn R, Shepard DS, Ettling MB, Brinkmann U, Nougtara A, Diesfeld HJ (1991) Estimating the direct and indirect economic costs of malaria in a rural district of Burkina Faso. Tropical Medicine and Parasitology 42, S1/219 S1/223.
  • Sawyer D (1993) Economic and social consequences of malaria in new colonisation projects in Brazil. Social Science and Medicine 37, 1131 1136.
  • Sharma RC, Malviya VS, Bhati PG (1990) Economic loss due to malaria in Kheda district, Gujarat. Indian Journal of Malariology 27, 149 155.
  • Shepard DS, Ettling MB, Brinkmann U, Sauerborn R (1991) The economic cost of malaria in Africa. Tropical Medicine and Parasitology 42, S1/199 S1/203.
  • Todaro MP (1992) Economics for a Developing World. 3rd edn. Longman, London.
  • Walker D & Fox-Rushby J (2000) Economic Evaluation of Parasitic Diseases. A Critique of the Internal and External Validity of published studies. Tropical Medicine and International Health 5, 237 249.
Footnotes
  1. The aim of the overall economic study was to estimate the cost-effectiveness of anti-malaria activities in the selected district (Attanayake 1994).

  2. A copy of the questionnaire is available in Sinhalese and English on request.

  3. Matale district is served by two monsoons called south-west from May to September and north-east from November to March. Each monsoon follows a peek of malaria occurrence, which normally take place during June/July and December/January. The household survey was initiated just after the north-east monsoon.

  4. Hence malaria is placed in quotation marks because it incorporates both confirmed as well as presumptive cases of malaria. Therefore it is possible that ‘malaria’ includes some people who did not have the biological state of malaria.

  5. Those that answered ‘no’ were only asked questions about the household level practices and costs of preventing malaria in addition to the socio-economic and demographic questions. Questions on preventive care were asked from those that answered ‘yes’ as well.

  6. This excluded patients who were currently ill and who had not recovered at the point of the survey. Their inclusion would artificially lower the costs incurred per fully recovered patient and create difficulties in comparison with other studies. Full recovery was defined as the ability to restart the engagement in his/her main and subsidiary occupations or the absence of any physical or mental disability. In partial recovery, patient was able to engage in some of those activities.

  7. E.g. King coconuts are not drunk for their ‘nutritional’ value during an episode of malaria but in accordance with local beliefs that it is effective in altering the ‘heating’ of the body, so precipitating an early recovery from illness.

  8. Complete disability period was defined as the number of days in which the patient had to avoid the engagement in his/her main and subsidiary occupations due to physical and/or mental disability. Partial disability was the inability to engage in some of those activities.

  9. Prevailing wage rates of all main occupations were collected during discussions and interviews. Since these rates were almost the same for many parts of each subdistrict, the most frequently stated rate (mode) in each subdistrict was taken as the indicator of central tendency. Sharing work among other household members was largely confined to manual agricultural work. Therefore the average wage rate of casual manual labourers was taken as the opportunity cost of the time spent by the other household members in sharing work.

  10. Such an approach was essential because there seemed to be reluctance among the respondents to express their actual income. As the pilot survey showed, some tended to underestimate their income assuming that the survey results might adversely affect their receiving of government income support. Some others, who were not income support receivers, seemed to think that the survey results might be used in the future to determine income support, so they tended to underestimate income. High-income receivers attempted to hide their income due to fears of getting caught by income tax.

  11. Full recovery was defined as the ability to restart the engagement in his/her main and subsidiary occupations or the absence of any physical and mental disability.

  12. Slash and burn method is largely used for farming in the jungles. Farmers have to spend nights in the huts of their farms to protect their cultivation from wild animals. This may be a prime reason for the high incidence rate among middle age males.

  13. In 1993 average value of the official exchange rate was US$= Rs. 48.25.

  14. In Matale district inpatient care was not available in the private sector.

  15. Throughout the paper the term ‘economically active’ is used to indicate the engagement of a household member in any nondomestic activity leading to generation of income either in kind or cash.

  16. Whilst 37% of patients sought care from the private sector, costs from the viewpoint of private sector provision are shown as zero because patients had paid for the services. This is to avoid double counting (and partly because it is difficult to access such information from the private sector).

  17. However, although blood-testing facilities were limited, nearly half of the patients had had their diagnosis, confirmed by blood tests at public medical centres. No statistically significant differences (at 95% confidence) were found between the average costs of treatment of confirmed and unconfirmed cases. Hence the inclusion of the unconfirmed cases could not be considered to have biased the overall estimate of average cost (Attanayake 1994).

  18. We distinguish measurement by focusing on the physical quantities of resources, in this case the length of time and quality of time available for productive activity. Valuation is then concerned with alternative approaches to pricing this time.

  19. Respondents were asked to make a judgement about the level of illness on the basis of their exposure to the disease.

  20. The weight was based on the number of days worked in each week (which was very low in off peak agricultural periods).