Cost of malaria control in Sri Lanka.

Research Cost of malaria control in Sri Lanka F. Konradsen, 1 P. Steele, 2 D. Perera, 3 W. van der Hoek, 4 P.H. Amerasinghe, 5 & F.P. Amerasinghe 6 The study provides estimates of the cost of various malaria control measures in an area of North-Central Province of Sri Lanka where the disease is endemic. We assumed that each measure was equally effective. In these terms, impregnating privately purchased bednets with insecticide was estimated to cost Rs 48 (US$ 0.87) per individual protected per year, less than half the cost of spraying houses with residual insecticides. Larviciding of vector breeding sites and especially the elimination of breeding habitats by flushing streams through seasonal release of water from upstream reservoirs was estimated to be cheaper than other preventive measures (Rs 27 (US$ 0.49) and Rs 13 (US$ 0.24) per individual protected, respectively). Inclusion of both operational and capital costs of treatment indicates that the most cost-effective intervention for the government was a centrally located hospital with a relatively large catchment area (Rs 71 (US$ 1.29) per malaria case treated). Mobile clinics (Rs 153 (US$ 2.78) per malaria case treated) and a village treatment centre (Rs 112 (US$ 2.04)) per malaria case treated) were more expensive options for the government, but were considerably cheaper for households than the traditional hospital facilities. This information can guide health planners and government decision-makers in choosing the most appropriate combination of curative and preventivemeasures to control malaria. However, the option that is cheapest for the governmentmay not be so for the householders, and further studies are needed to estimate the effectiveness of the various preventive measures. Voir page 308 le re sume en francËais. En la pa gina 309 figura un resumen en espanÄ ol. Introduction A large proportion of Sri Lanka's health budget is spent on malaria control. The most recent year for which a central figure is available is 1989, prior to the devolution of the health budgets to the provinces. At that time 11.6% of the total budget was spent on all forms of antimalaria activities nationwide, including both malarious and nonmalarious parts of the country (1). In 1992, it was estimated that 36% of the total health expenditure in a malaria-endemic district of the country was used to combat the disease (2). The cost to the government of antimalarial activities relates mainly to vector control, e.g. spraying houses with insecticides and larviciding vector breeding habitats, supporting diagnostic and treatment facilities, and maintaining an information system. The bulk of malaria control expenditure is currently used on importing insecticides. However, since the implementation of a selective spraying programme in 1994 the total amount of insecticides used has declined significantly (1). Discussion of the most cost-effective use of the budget allocation for antimalaria activities in Sri Lanka requires comparison of the cost-effectiveness of a range of preventive and curative interventions. This comparison needs to be based on standardized cost estimate procedures and comparable units of effective outputs. Furthermore, to be able to discuss the most suitable interventions, we need to include the cost to households and not only the cost estimates borne by government agencies. In Sri Lanka, the transmission of malaria exhibits considerable seasonal and annual changes, the population is highly knowledgeable about the disease, the majority of the cases are caused by Plasmodium vivax, and the health facilities provide relatively good coverage. All of these make it very difficult to compare health intervention strategies in Sri Lanka with cost-effectiveness studies carried out in Africa. To the best of our knowledge the only studies of the cost of malaria control to the government in Sri Lanka are those by Attanayake (2) and Graves et al. (1). The present study includes a number of additional interventions not dealt with by these two previous studies, such as water manage- 1 Environmental Health Biologist, International Irrigation Management Institute, PO Box 2075, Colombo, Sri Lanka. Requests for reprints should be sent to this author. 2 Economist, Development Options (Private) Limited, Colombo, Sri Lanka. 3 Regional Malaria Officer, Anti-Malaria Campaign, Anuradhapura, Sri Lanka. 4 Epidemiologist, Research Leader, Health & Environment Programme, International Irrigation Management Institute, Colombo, Sri Lanka. 5 Parasitologist and Head, Department of Molecular Biology and Bio- Technology, Faculty of Science, University of Peradeniya, Peradeniya, Sri Lanka. 6 Professor of Applied Zoology, Department of Zoology, Faculty of Science, University of Peradeniya, Peradeniya, Sri Lanka. Reprint No. 5803 301Bulletin of the World Health Organization, 1999, 77 (4) # World Health Organization 1999 ment for vector control and a village-level treatment centre. The cost estimates were generated from data for a rural area in the malaria-endemic part of Sri Lanka, where ongoing research activities made it possible to estimate the cost to the government and households of both traditional and novel malaria control inter- ventions. Results from epidemiological studies and surveys of treatment-seeking behaviour undertaken among the study population were used to assess the most appropriate interventions for the area. Materials and methods Study area The study was carried out in Anuradhapura District, in theNorth-Central Province of Sri Lanka, in an area covered by seven neighbouring villages, consisting of 512 households and a total population of 2575 (average household size, 5 persons). The office of the Anti-Malaria Campaign (AMC) in Anuradhapura is headed by a regional malaria officer (RMO) who undertakes entomological surveillance and special surveys and provides advice to the divisional director of health services, the public health inspectors, and other personnel responsible for malaria control activities within the district. The majority of the households are engaged in agriculture, primarily subsistence farming. In 1995, a survey in the study area showed that 48% of malaria patients first sought treatment at the government hospital in Kekirawa, approximately 25 km from the cluster of study villages; 30% used the government managed mobile clinics; and the remain- ing 22% private or other Western-style facilities (3). In fact, a number of patients attended both the mobile clinics and hospital. The community had a good understanding of the transmission route of malaria and the drugs required for its treatment, and placed great importance on positive blood-slide confirmation before treatment. Data collection The cost estimates and the epidemiological back- ground information were collected for the period 1 October 1994 to 30 September 1995. A review of the health statistics for theKekirawa division over the period 1979 to 1995 shows that the level of malaria transmission in 1994±95 was average (4). Data on salaries and travel allowances, staff input, cost of fuel, chemicals and other supplies, and the amount of chemicals required for each interven- tion were provided by various branches of the Ministry ofHealth, including the AMC. In addition to weekends and national holidays, two days per month were added to the cost of labour input to cover the expenses related to sick leave and personal holidays. This additional cost was averaged over a month and added to the daily salary rates. Chemicals are currently purchased by the AMC and transported to Anuradhapura town for storage. The cost of transport and storage was added to the import price of the chemicals. The cost of chemicals was based on the standard formulations used by the AMC. The cost of capital investment was derived from actual expenditures recorded by the Ministry of Health; full depreciation was assumed after 7 years for vehicles and 20 years for buildings. Maintenance expenses were also included. Although the data needed for analysis of the different malaria control interventions have been taken to be representative for the whole area covered by the regional office of the AMC in Anuradhapura District, they have only been exemplified and validated for the study area. On the basis of data collected as part of a previous survey in the study area, we were able to estimate the cost to the households of seeking treatment at the various facilities (3). The number of blood slides examined and cases treated by the government mobile clinic and at one of the main facilities (Kekirawa hospital) servicing the study area were obtained from the AMC and the registrar of the malaria unit of the hospital, respectively. Cost estimates for the village-based treatment centre were obtained from the reports of ongoing research activities in the study area. In order to compare different preventive and curative interventions, we devised common units of measurement. For curative interventions two units were used: the cost per blood slide examined by a microscopist, and cost per blood slide found to be malaria positive and treated. For preventive inter- ventions, the unit was taken to be the cost to protect a person for a year or for a season. For all interventions, both operating and fixed costs to the government and households were calculated. Some interventions required a vehicle when the activities were carried out or for supervisory purposes, adding considerably to the capital costs. The fixed cost of the hospital was based on current replacement value and was annual- ized based on the approach described by Phillips et al. (5), while the fixed cost of the treatment centre was taken to be the rent of the building. Residual spraying A spraying team from the health authority in Kekirawa town visited the study area routinely and treated all sprayable surfaces of 25 dwellings per day. The records showed that more houses than this were visited per day but due to householders' absences and refusals only 25 houses were treated. The team included four spray machine operators and one field assistant, who were all paid a salary and travel allowance. The cost of the spraying operations was based on the inputs needed for a ``walking unit'' (actually making use of a bicycle), with a vehicle only being used for supervisory purposes. The operational and capital costs of the supervision were based on the assumption that a public health inspector with a driver could supervise four walking units per day. Recently the insecticide had been changed from Research 302 Bulletin of the World Health Organization, 1999, 77 (4) malathion to lambda-cyhalothrin or fenitrothion, and it was assumed that themosquito vectors had still not developed resistance to these newly introduced chemicals. Fenitrothion was sprayed three times a year and lambda-cyhalothrin twice a year. Feni- trothion costs Rs 404 (US$ 7.35) per kg and approximately 0.50 kg was required for each house. Lambda-cyhalothrin costs Rs 3870 (US$ 70.36) per kg, but only 0.06 kg was required for each house. In areas where malathion was still used it was sprayed three times a year (0.90 kg per house) at a cost of Rs 153 (US$ 2.78) per kg. Household spraying with malathion had a very high acceptance level in the study area, with more than 90% of the houses being treated (3). This acceptance was expected to remain high following the introduction of the newer insecticides, since they stain the walls less and are nearly odourless. The annual cost of protective clothing and of other equipment for spraying was Rs 2380 (US$ 43.27). The cost of routine check-ups for the spray operators, using the cholinesterase test, was Rs 2500 (US$ 45.45) per person with check-ups being carried out every 2.5 months. Bednet impregnation The government health services supported a mos- quito bednet impregnation programme, in which a public health inspector, a field assistant, a driver and a labourer visited an area to dip the nets in a solu- tion of permethrin. This insecticide costs Rs 1120 (US$ 20.36) per litre, which is sufficient for about 20 bednets. The procedure was to impregnate approximately 100 bednets or about 50 households (assuming two double bednets per household) per day. These limits arose because villagers brought their nets to the location where the impregnation was to take place. Also, the efficiency of the impregnation programme depended on the presence of a large proportion of bednet users among the community. At the time of the study, 7% of the households in the study area owned at least one privately purchased mosquito net (cost, ca. Rs 600 (US$ 10.91) for a double net). Larviciding The primary vector of malaria in Sri Lanka, Anopheles culicifacies, mainly breeds in streams and stream-bed pools. It has therefore been the practice of the AMC to apply larvicides to streams running close to human settlements during the dry period when the water levels in the rivers and streams are falling. One stream (the Yan Oya) that crosses the study area was found to be the only breeding site forA. culicifacies during the period mid-June to early September (6). To apply the larvi- cides, a walking unit consisting of a team of spray machine operators and a field assistant from the health department in Kekirawa visited the area and sprayed pools along an average of 1 kmof stream-bed per day. A public health inspector, with a driver, supervised the walking unit every fourth working day. Assuming that mosquito breeding in the stream could have an impact on the households living within 1.5 km of the stream, it would take one day to protect an average of 56 households along its banks within the study area. The larvicide used was temephos (cost, Rs 753 (US$ 13.69) per litre), with ca. 0.4 litres being consumed over a 1-km stretch. To effectively protect this area for the 10-week peak stream-bed breeding period, approximately four site visits were required. Water management for larvae control TheYanOya stream is used as a conveyance canal for irrigation water from an upstream to a downstream reservoir, a distance of approximately 25 km. As part of the ongoing research in the area, management of the water releases was an effective means of eliminating the breeding of A. culicifacies during the period mid-June to early September (7). The water management procedure for mosquito control relied on the release from the upstream reservoir of 22 000 m 3 water every tenth day over a 2-day period. This was expected to reduce formation of pools in the stream and flush existing pools, thereby disturb- ing the breeding of the larvae present in the stream at the beginning of the tenth day of water release. Thus the flush was required about six times during the 10-week peak breeding season. Since the water released from upstream is captured downstream, the only water lost to the system is the increased conveyance loss due to the periodic releases into the dry stream. This was estimated at 10% by a water balance study (Interna- tional Irrigation Management Institute, unpublished data, 1998). The estimated value of water in paddy production in the general study area is Rs 3 (US$ 0.06) per m 3 (8), based on the standardized gross value of rice produced divided by the quantity of water diverted to the field. Thus the value of the water lost through conveyance was estimated to be Rs 6600 (US$ 120.00) per control intervention. A farmer under the supervision of an Irrigation Department official regulated the gate at the upstream reservoir. However the cost of this management input is negligible when it is divided by all the households protected (Rs 0.40 per household per intervention). As a result of this flushing, an estimated 600 house- holds would be protected along this 25-km stretch of the Yan Oya stream. Diagnosis and treatment at the government hospital, Kekirawa The government hospital in Kekirawa is one of several centres for diagnosis and treatment of malaria within the Kekirawa division (population ca. 45 000). A microscopist and a field assistant attended to patients all days of the week from 09:00 to 16:00, except on Sundays and national holidays. Thus the service was available for about 290 days a year. Positive cases were treated with a standard course of chloroquine (CQ) and primaquine (PQ), and for suspected resistant cases pyrimethamine + sulfadoxine was prescribed (at a cost of approximately Rs 60 (US$ 1.09) per case treated). Cost of malaria control in Sri Lanka 303Bulletin of the World Health Organization, 1999, 77 (4) However, for our analysis each positive case of malaria was assumed to have been given only CQ and PQ at an average dose of six CQ tablets (a total of Rs 4 (US$ 0.07)) and six PQ tablets (a total of Rs 2 (US$ 0.04)) per case. A new lancet and glass slide were used for each blood sample at a total cost of Rs 4 (US$ 0.07) per patient. During the 1994±95 study period, 10 769 blood slides were examined, 2607 of which were positive (slide positivity = 24%); this corresponds, on average, to 37 blood slides per day. In addition to staff and other operating costs, the fixed costs of the hospital were calculated at a standard rate of Rs 8070 (US$ 146.72) perm 2 . Themalaria unit at Kekirawa hospital occupies ca. 28m 2 . This corresponds to a total capital cost of Rs 225 960 (US$ 4108.36), which annualized over 20 years using a real discount rate of 6% and assuming 290 working days a year is equivalent to Rs 68 (US$ 1.24) per day. The capital cost of a microscope (Rs 267 000) was annualized over 15 years, which is equivalent to Rs 94 (US$ 1.71) per day. Maintenance costs of the hospital were assumed to be 10%of capital costs at Rs 7 (US$ 0.13) per day, and have been included in the operating cost figures. Mobile clinics for the diagnosis and treatment of malaria During the main transmission periods, the govern- ment health service operated specialized mobile clinics. Over the one year study period the study area was visited 20 times, mainly during the months when malaria transmission was high. Patients received the same service at these clinics as they did at the government facility, but the distance they had to travel was shorter, with a maximum distance of 7 km andwith ca. 60% of the population living within 2 km of these clinics. Since themobile clinic visited the area on fixed days of the month, no special promotional efforts were carried out to support the service. The clinic staff included amicroscopist, field assistant and driver, all of whom received wages and a travel allowance. As with the hospital, it was assumed that on average each positive case received six CQ tablets and six PQ tablets. During the 1994±95 study period, 1009 blood-slides were examined and 275 positive cases identified during the 20 visits. This corresponds to an average daily total of 50 blood slides and an average of 14 positives per day. A fixed cost for the type of vehicle used by the mobile team was calculated at Rs 1.3 million (US$ 23 636.36). Spread over 7 years, annualized using a discount rate of 6% and assuming 290 days of use per year, this amounts to Rs 803 (US$ 14.6) per day. Maintenance costs were calculated at Rs 1500 (US$ 27.27) per month, and assuming 290 days of use per year amounts to Rs 62 (US$ 1.13) per day. These maintenance expenses were included in the operating cost figures. Village level diagnosis and treatment centre A diagnosis and treatment centre was established in May 1996 in a rented room in the local post office in a village located in the centre of the study area. Secondary school graduates from the study area received a total of 16 days' training at the Universities of Colombo and Peradeniya in the management of a village health facility, mainly focusing on blood slide examination and record keeping. This was followed by on-the-job training by a university graduate located in the study area and byAMC staff. Following the establishment of the centre the existing staff trained new assistants. The trained assistants were paid a daily salary slightly below the level of a field assistant in the government service, but received no field allowance. The centre was staffed by two assistants every day during its opening hours from 08:00 to 16:00. For our analysis, the start-up expenses and the initial promotional work (i.e. training, initial supervision, poster campaign, and community meet- ings) were not included in the cost estimate. The diagnostic procedures and the drug regimen at the village level centre were equivalent to those at the government facility in Kekirawa and the mobile clinics. To maintain the quality of work, two half-day supervisory visits were made every month by staff of the government health service. After the village treatment centre was opened, the mobile clinics no longer visited the study area. A survey carried out on 90% of the households in the study area showed that 74% of the patients with suspected malaria first attended the village treatment centre, with the government hospital in Kekirawa being the second choice (International Irrigation Management Insti- tute, unpublished data, 1998). To assess the cost per blood slide examined and per case treated for the same period as that used for the government hospital and the mobile clinics, we used detailed epidemio- logical data collected in one of the seven study villages (9). In this village over the period October 1994 to September 1995 each person had an average of 0.8 episodes of malaria. Had the village treatment centre been up and running during this period, and with 74% of the patients seeking treatment at it, 1524 malaria patients would have been diagnosed and treated among the population of 2575. The centre is open 305 days a year and would therefore have seen five positive cases per day. We assumed that the village treatment centre had a slide positivity rate similar to that of the government hospital (24%), equivalent to 20 blood films per day. The fixed cost of the building occupied by the village centre was estimated to be Rs 20 (US$ 0.36) per day, based on a rent of Rs 500 (US$ 9.09) per month. Results Prevention The residual spraying programme was the most expensive preventive intervention, with fenitrothion being considerably more so than malathion (Table 1). The insecticidewas by far themost costly component of the spraying programme, amounting to 75%of the Research 304 Bulletin of the World Health Organization, 1999, 77 (4) total operating expenditure. Impregnated bednets were considerably cheaper for the government (ca. US$ 0.87 per individual protected per year), less than one half the cost of protecting an individual through residual spraying. If we assume that three and not two bednets are needed per household, the cost to the government per individual protected per year increases to US$ 1.27, which is still substantially less than the cost of the residual spraying programme. However, to the government cost of impreg- nating bednets should be added the cost to the households of purchasing the bednets. If it is assumed that two bednets are required per house- hold, the cost would be Rs 50 (US$ 0.91) or Rs 70 (US$ 1.27) per individual protected per year, depending on whether the life span of the bednet was estimated at 6 years or 4 years. Nevertheless, the total cost to the government and households combined made impregnated bednets more cost- effective than the residual spraying programme.Only if the estimates are based on use of three double bednets per household and the nets last 4 years does the combined government and household cost of the impregnation programme make this intervention slightly more expensive per individual protected (at US$ 2.90) than that of the residual spraying programme. The two cheapest preventive options were larviciding and water management, but it is unclear whether the protection they afforded was compar- able to the above approaches, and estimates of their effectiveness were difficult to make. Also, they were appropriate for use only for approximately the 3- month period when the main vector A. culicifacies breeds in the streams close to human habitation. However, by interrupting the majority of the breeding early in the transmission season, vector abundance in the succeeding months would be reduced. In streams where water can be managed using existing structures, the use of controlled flushing would provide protection at around half the cost of the use of chemical larvicides. Treatment The village treatment centre had a smaller catchment area than Kekirawa hospital, and therefore examined fewer blood slides per day. The mobile clinic had the same catchment area as the village treatment centre, but was functional only during the high transmission season, while the village treatment centre was open all year around. This meant that more blood slides were taken per day when themobile clinic was in operation compared with the village treatment centre. When themobile clinic was not in operation as amalaria unit it was assumed that the vehicle and staff were employed elsewhere and, except for the microsco- pist, were not necessarily involved in malaria control activities. Thus, based on operating cost, the village treatment centre and mobile clinic were more expensive than the hospital (Table 2). Table 3 shows the cost estimates to the government for curative interventions, including both operational and capital costs, as well as house- hold expenses related to seeking treatment. In the household survey carried out in 1995, a total of 85 households reported a recent malaria episode for which they sought treatment at the hospital. These households incurred a median cost of Rs 55 Table 1. Operating and capital costs to the government of various antimalaria preventive interventions Intervention Operating No. of Operating cost No. of Annual Annual Total annual cost per day households per household interventions operating capital cost per (in Rs) protected per intervention per year cost per cost per individual per day day needed household household protected (in Rs) for protection (in Rs) (in Rs) (in Rs) Spraying Fenitrothion 6 412 25 256 3 768 (13.96) a 24 (0.44) 158 (2.87) Lambda- 7 409 25 296 2 592 (10.70) 16 (0.29) 122 (2.21) cyhalothrin Malathion 4 613 25 185 3 555 (10.09) 24 (0.44) 116 (2.11) Bednet impregnation 5 244 50 105 2 210 (3.82) 32 (0.58) 48 (0.87) (permethrin) Larviciding b 1 652 56 30 4 120 (2.18) 16 (0.29) 27 (0.49) (temephos) Water 6 600 600 11 6 66 (1.20) 0 (0) 13 (0.24) management b a Figures in parentheses are US$ (US$ 1 = Rs 55). b Interventions with only seasonal applicability. Cost of malaria control in Sri Lanka 305Bulletin of the World Health Organization, 1999, 77 (4) (interquartile range, Rs 35±78) on transport and on foodwhile the patient and the accompanying persons were awaiting diagnosis and treatment. The 54 households that sought treatment at the mobile clinic spent much less (median, Rs 0; interquartile range, Rs 0±10). The survey of the household costs was conducted before the village treatment centre was established, but it is assumed that households attending the centre would spent the same amount as those attending the mobile clinic. Inclusion of the household costs alters the findings, with the village treatment centre being the cheapest option and the mobile clinic themost expensive at Rs 153 (US$ 2.78) per positive case treated, followed by the hospital at Rs 126 (US$ 2.11) for the same output. However, with possible errors in the estimates, it is probable that, after including the cost to the households, the three curative interventions become comparable in cost. A focus on the centralized treatment facility with a large catchment area is a cheap option for the government but shifts a large part of the expense to the households. Estimation of the cost of the three curative services depends largely on the average number of cases registered at the facilities. Therefore, to produce a rough sensitivity test for the estimates generated in Table 3, we increased and decreased by 25% the average number of positive cases treated at the three facilities in the calculation. Based on the revised case load at the three facilities, the total cost to the government was estimated to be US$ 1.04±1.72 per positive case treated at the hospital, US$ 2.25± 3.62 per case treated at mobile clinics, and US$ 1.65± 2.63 per case at the village treatment centre. This shows that, even with the most expensive estimate for the government hospital and the least expensive for the two other treatment alternatives, the hospital was still, from a government perspective, cheaper or approximately the same cost as the two alternatives. An important item for both preventive and curative interventions was the capital cost of the vehicle. In tests of the sensitivity of the final estimation for the preventive interventions, the cost per individual protected per year increased by a maximum of 7.5% for the bednet impregnation programme when the lifetime of the car was reduced from 7 years to 4 years. For all other preventive interventions the cost increased by less than 2.5%. The total cost estimated per positive case treated for the mobile clinic increased by 19% when the lifetime of the vehicle was reduced to 4 years, making this curative service even less competitive. The cost estimate for the village treatment centre changed only marginally, and that for the government hospital remained the same when the lifetime of the car was reduced. We were unable to obtain any reliable estimates of input wastage for items such as insecticides and drugs. For the residual spraying Table 2. Operating costs to the government of different curative interventions Intervention Operating cost Average Average Operating cost per Operating per day number of number of blood-film cost per (in Rs) blood-films positive blood- examined positive per day films (in Rs) blood-film per day (in Rs) Hospital 483 (8.78) a 37 9 13 (0.24) 53 (0.98) Mobile clinic 1 242 (22.58) 50 14 25 (0.45) 89 (1.61) Village level 443 (8.05) 20 5 22 (0.40) 89 (1.61) treatment centre a Figures in parentheses are US$ (US$ 1 = Rs 55). Table 3. Operating and capital costs to government and households of various treatment interventions Intervention Operating cost Capital cost Capital cost Cost to Cost to Total cost per per positive per day per positive government household per positive case for case (in Rs) (in Rs) case per positive case positive case both government (in Rs) (in Rs) (in Rs) and household (in Rs) Hospital 53 162 (2.95) a 18 (0.32) 71 (1.29) 55 b (1.00) 126 (2.11) Mobile clinic 89 897 (16.3) 64 (1.16) 153 (2.78) 0 153 (2.78) Village level 89 114 (2.07) 23 (0.42) 112 (2.04) 0 112 (2.04) treatment centre a Figures in parentheses are US$ (US$ 1 = Rs 55). b Median direct cost for the patient from the study area and, possibly, a guardian to seek treatment at the government hospital. Research 306 Bulletin of the World Health Organization, 1999, 77 (4) activities especially, this may have important implica- tions; for example, if insecticide wastage were 25%, the costs of spraying would increase by 20%. Discussion Seen from the government perspective, a centrally located hospital with a relatively large catchment area is the most cost-effective way of treating malaria patients, while the widespread use of mobile clinics and village facilities is an expensive control strategy. For larval control, the use of designated water management strategies should be explored since this option is far cheaper than the use of chemically based larvicides. For the reduction of bites by adult mosquitos, a programme of impregnating privately purchased mosquito bednets would provide protec- tion to the households at approximately half the price of a residual spraying programme. The residual spraying programme has become more expensive in Sri Lanka because the mosquito vectors have developed resistance to the insecticides and each of the newly introduced insecticides is more costly than its predecessor. In general, the spraying operations in Sri Lanka are not very costly due to the low capital investments in, for example, vehicles. However, this may not be a realistic option in other developing countries with a less developed infrastructure and a low population density. The assessment of the number of patients seeking treatment at any one of the three facilities in the present analysis was a best estimation and is likely to vary considerably over time. For example, the number of patients visiting the government hospital may vary according to the number of mobile clinics operating in its catchment area and influence the average number of slides taken per day. We therefore assessed the sensitivity of the final cost estimation for the curative interventions by changing the average number of slides taken per day. The results showed that even with a substantial reduction in the number of positive cases treated at the government hospital it was still a cost-effective option from a government perspective compared with the two alternatives. The cost per case treated by the mobile clinics (Rs 153, US$ 2.78) is well below the Rs 500 (US$ 9.09) reported previously for four districts in Sri Lanka (1). Our analysis indicates that the cost- effectiveness of the mobile clinics is largely influ- enced by the life span of the vehicle, whichmay partly explain this difference in cost estimates. However, the main reason for the lower cost per case treated in our study is probably because attendance at the mobile clinics was higher than that reported by Graves et al. (1). Hencemobile clinics may be a viable option if they are operated rationally, targeting areas of high transmission, and functioning only when malaria is most prevalent. Having mobile clinics as part of the control strategy may also provide some flexibility in countries such as Sri Lanka, where the disease pattern is highly unstable. The capital and manpower inputs needed to upgrade a large number of peripheral hospitals with blood-filming units would be significant and could result in a very high cost per case treated in years of low malaria incidence. However, to be able to make use of the mobile clinics in a rational manner the health information system needs to be good enough to identify areas of high transmission quickly. An alternative strategy would be to establish village-level treatment facilities combined with the services of central hospital facilities. The estimates that we have given in this article show that, from a government perspective, village- level facilities will probably never be as cost-effective as central facilities. However, inclusion of the cost to the households in the assessment indicates that village facilities provide considerable benefits to the community both by reducing direct expenditure and the time involved in seeking treatment (the oppor- tunity cost of malaria to the households in the study area has been discussed in more detail elsewhere (10)). In addition, the village treatment centre may encourage people to seek treatment earlier, and thereby reduce the overall transmission of malaria. Assuming that accord can be reached between government health services and the relevant labour unions, a village treatment centre would provide an ideal opportunity to increase the community's involvement in diagnosing and treating malaria, with the people from the villages managing the centres under the supervision of experienced government microscopists. To determine the best mix of curative interventions, in addition to the health information system discussed above, an improved accounting system is also needed to facilitate decision-making. The cost-effectiveness comparison between the residual spraying programme and the impregna- tion of bednets is based on the assumption that all the interventions are equally effective. This of course is debatable. It is assumed here that all individuals make use of bednets all year round, which may not be realistic. On the other hand, insecticide-impregnated bednets are very important in reducing the incidence of disease (11). Also, impregnated bednets have an impact on transmission both through the applied insecticide as well as by reducing the number of mosquito bites experienced by people using them, and they may therefore provide more effective protection than residual spraying. In addition, even though people may not use the bednets all year round they are more likely to use them during the periods with the most mosquitos. The newly introduced insecticides in the Sri Lankan residual spraying programme are still likely to be effective, but the comparison with malathion may no longer be valid owing to the development of resistance among some of the potential vectors (12). The insecticide spraying figures assume no wastage but studies should be carried out to quantify the true levels involved. The effectiveness of the residual spraying activities also depends on community collaboration. This was not a problem in the study area but may be Cost of malaria control in Sri Lanka 307Bulletin of the World Health Organization, 1999, 77 (4) an issue elsewhere in Sri Lanka (13). However, the conclusion still holds that a bednet impregnation programme is a very cost-effective option for the government to pursue. New ways of implementing the programme should be explored, including training community representatives to be responsible for impregnating the nets, supervised by government staff. The cost of a vehicle for use in the impregnation programme is a significant expense but increased community involvement could reduce this, in addition to lowering the input of government staff. The cost to the household of purchasing bednets is high and probably prohibitive for a large segment of the population under the current socioeconomic conditions in Sri Lanka. Previous findings in the study area showed a significant positive correlation between household income and bednet use (3). A sudden change in focus from a residual spraying programme to bednet impregnation could therefore leave the poorest segment of the population unprotected.Ways of increasing bednet use therefore need to be explored. If a high coverage could be achieved, the cost-effectiveness of the impregnation operation would also be improved by increasing the number of bednets per day that could be treated by a team. Some of the proposals put forward by Evans (14) and Brinkmann & Brinkmann (15) on involving the private sector, nongovernmental organizations, and integrating the bednet impregnation activities into the routine health services are also of relevance in Sri Lanka. In parts of Sri Lanka or elsewherewhere natural streams are used as conveyance canals for irrigation water it is very important that the responsible government departments explore the possibilities for integrating water management objectives and malaria control activities. Managing water in local streams will save the health department significant sums of money and provide a low cost means of reducing the vector abundance for part of the year. It may be worth while also to examine the cost- effectiveness of other environmental management- based options. The present study was carried out in a small area but we believe that its findings are valid for other parts of the dry zone of Sri Lanka. Furthermore, the small area approach made it possible to collect detailed information on inputs and their costs. The most important advantage was the collection of reliable malaria incidence data, which is often impossible for larger areas. The study included preventive and curative interventions that were actually taking place in the study area. Water management was included as an environmental control measure that has been worked out in detail for the area and is ready for implementation. Other preventive and curative interventions could have been included, such as the use of community health volunteers and the biological control of mosquito larvae; however, these were neither practised nor tested at the time of the study and estimates would have been speculative. n Acknowledgements We are most grateful for the support provided by Dr W.P. Fernando, Director, AMC, and the public health inspector in Kekirawa, Mr D. Weerasinghe. This study was supported by the Danish Interna- tional Development Agency (DANIDA) and the Government of Japan. We also wish to acknowledge the constructive comments provided by the two anonymous reviewers. Re sume Couà t de la lutte antipaludique aÁ Sri Lanka Cette e tude donne une estimation du couà t de diverses mesures de lutte antipaludique mises en úuvre dans une zone situe e dans le centre-nord du Sri Lanka, ouÁ la maladie est endeÂmique. Nous avons suppose que chaque mesure avait la meÃme efficacite . Nous avons estime aÁ 48 roupies sri lankaises (ReSL) Ð soit US $0,87 Ð par personne et par an le couà t de l'impre gnation des moustiquaires achete es aÁ titre prive , soit moins de la moitie du couà t du traitement d'un logement par un insecticide aÁ effet reÂmanent. La destruction des larves de vecteurs par e pandage de larvicides, mais surtout par inondation des gõÃtes larvaires en proce dant aÁ des laà chers d'eau saisonniers depuis les retenues situe es en amont, s'est re ve le e plus e cono- mique que les autres mesures de pre vention (respective- ment ReSL 27 ou US $0,49 et ReSL 13 ou US $0,24 par personne prote ge e). Si l'on tient compte des de penses de fonctionnement et de l'investissement que repre sente le traitement des malades, on constate que pour les pouvoirs publics, c'est un hoà pital central desservant une zone relativement e tendue qui est le plus e conomique (ReSL 71 ou US $1,29 par malade traite ). Les dispensaires mobiles (ReSL 153 ou US $2,78 par malade traite ) et les centres antipaludiques de village (ReSL 112 ou US $2,04 par malade traite ) repre sentent les options les plus couà teuses pour les pouvoirs publics, mais pour les me nages, leurs services sont beaucoup moins chers que ceux des hoà pitaux traditionnels. Ces re sultats peuvent aider les pouvoirs publics aÁ choisir la combinaison la plus approprie e de mesures pre ventives et curatives pour combattre le paludisme. Il reste que l'option la plus e conomique pour le secteur public n'est pas forceÂment la moins couà teuse pour les me nages, et d'autres travaux seront ne cessaires pour e valuer l'efficacite des diverses mesures de pre vention. Research 308 Bulletin of the World Health Organization, 1999, 77 (4) Resumen Costo de la lucha contra el paludismo en Sri Lanka En el presente estudio se calculan los costos de diversas medidas de control del paludismo en una zona endeÂmica del centro norte de Sri Lanka. Partimos de la hipo tesis de que todas las medidas eran igualmente eficaces. Sobre esa base, el costo de impregnar con insecticida mosquiteros adquiridos por particulares se calculo en Rs 48 (US$ 0,87) por persona protegida por anÄ o, menos de la mitad del costo de pulverizar las casas con insecticidas residuales. Se estimo que la aplicacio n de larvicidas en los criaderos de vectores, y especialmente la eliminacio n de dichos criaderos mediante la liberacio n estacional de agua a partir de embalses situados aguas arriba, resultaba ma s barata que otras medidas preventivas (Rs 27 (US$ 0,49) y Rs 13 (US$ 0,24) por persona protegida, respectivamente). Considerando los gastos tanto de funcionamiento como de capital, la intervencio n ma s eficaz en relacio n con los costos para el Estado resulto ser la proporcionada por un hospital central con un a rea de captacio n relativamente grande (Rs 71 (US$ 1,29) por caso de paludismo tratado). Los consultorios mo viles (Rs 153 (US$ 2,78) por caso de paludismo tratado) y los centros de tratamiento de aldea (Rs 112 (US$ 2,04) por caso de paludismo tratado) fueron opciones ma s costosas para el Estado, pero considerablemente ma s baratas para las familias que los servicios hospitalarios tradicionales. Esta informacio n puede orientar a los planificadores sanitarios y los encargados de adoptar las decisiones a nivel estatal a la hora de elegir la combinacio n ma s apropiada de medidas curativas y preventivas para combatir el paludismo. Sin embargo, la opcio n ma s barata para el Estado puede no serlo para los residentes, y se necesitan au n ma s estudios para calcular la eficacia de las diversas medidas preventivas. References 1. Graves PM, Fernando D, Attanayake N. Intensified malaria control programme in Sri Lanka with emphasis on primary care approach. Consultants report, New Health and Family Planning Project 1995. Washington, DC, IDA/World Bank, 1995. 2. Attanayake N. Cost effectiveness of anti-malaria activities in Sri Lanka. Ph.D. Thesis. London, London School of Hygiene and Tropical Medicine, 1994. 3. Konradsen F et al. Household responses to malaria and their cost: a study from rural Sri Lanka. Transactions of the Royal Society of tropical medicine and hygiene, 1997, 91 (2): 127±130. 4. van der HoekW et al. Correlation between rainfall and malaria in the dry zone of Sri Lanka. Annals of tropical medicine and parasitology, 1997, 91 (8): 945±949. 5. Philips P, Mills A , Dye C. Guidelines for cost-effectiveness analysis of vector control, PEEM Guidelines Series 3. Geneva, World Health Organization, 1993. 6. Amerasinghe FP et al. Anopheline (diptera: Culicidae) breeding in a traditional tank-based village ecosystem in north central Sri Lanka. Journal of medical entomology, 1997, 34 (3): 290±297. 7. Konradsen F et al. Anopheles culicifacies breeding in Sri Lanka and options for control through water management. Acta tropica, 1990, 71: 131±138. 8. Molden D et al. Indicators for comparing performance of irrigated agricultural systems. Colombo, International Water Management Institute, 1998 (Research Report No. 20). 9. van der Hoek et al. Risk factors for malaria: a microepidemio- logical study in a village in Sri Lanka. Transactions of the Royal Society of tropical medicine and hygiene, 1998, 92 (3): 265±269. 10. Konradsen F et al. Measuring the economic cost of malaria to households in Sri Lanka. American Journal of tropicalmedicine and hygiene, 1997, 56 (6): 656±660. 11. D'Alessandro U et al. A comparison of the efficacy of insecticide treated and untreated bed-nets in preventing malaria in Gambian children. Transactions of the Royal Society of tropical medicine and hygiene, 1995, 89 (6): 596±598. 12. Herath PRJ, Joshi GP. Pesticide selection pressure on Anopheles subpictus in Sri Lanka: comparison with two other Sri Lankan anophelines. Transactions of the Royal Society of tropical medicine and hygiene, 1989, 83 (4): 565±567. 13. Ault SK. Anthropological aspects of malaria control planning in Sri Lanka. Medical anthropology, 1983, 7: 28±49. 14. Evans DB, Azene G, Kirigia J. Should governments subsidize the use of insecticide-impregnated mosquito nets in Africa? Implications of a cost-effectiveness analysis. Health policy and planning, 1997, 12 (2): 107±114. 15. Brinkmann U, Brinkmann A. Economic aspects of the use of impregnated mosquito nets for malaria control. Bulletin of the World Health Organization, 1995, 73 (5): 651±658. Cost of malaria control in Sri Lanka 309Bulletin of the World Health Organization, 1999, 77 (4)