The economic evaluation of malaria control technologies: The case of Nepal

Sot. Sci. Med. Vol. 34. No. 9. pp. 965-972, 1992 0277-9536/92 S5.W + 0.00 Printed in Great Britain. All rights reserved Copyright c 1992 Pergamon Press Ltd

THE ECONOMIC EVALUATION OF MALARIA CONTROL TECHNOLOGIES: THE CASE OF NEPAL

ANNE MILLS

London School of Economics and Pol&ical Science, Houghton St, London WC2 2AE and London School of Hygiene and Tropical Medicine, Keppel St, London WCIE 7HT, U.K.

Abstract-This paper illustrates the methodological issues arising from the use of economic evaluation in a developing country context, and how economic evaluation can be applied in developing countries to draw conclusions of relevance to policy-makers. The paper reports research on the cost-effectiveness of the malaria control programme in Nepal. It outlines the heirarchy of choices presented by malaria control and concentrates on the evaluation of the mix of routine strategies employed by the programme, particularly for vector control and case detection and treatment. A social perspective is taken, and emphasis placed on identifying costs falling on households, namely private expenditure on treatment and loss of days of work. Conclusions are drawn relating to the application of economic evaluation methodology to disease control programmes in developing countries.

Key words+conomic evaluation, cost-effectiveness analysis, malaria, Nepal

1. INTRODUCTION

Disease control programmes absorb a sizeable proportion of health sector expenditure in many developing countries. Indeed in some countries, of which a prime example is Nepal, disease control programmes preceded the extension of general health services to the population and were the first health programmes to reach into people’s homes. Yet the cost-effectiveness of such programmes has been surprisingly little studied.

This paper reports research on the cost-effectiveness of the malaria control programme in Nepal. The research was planned when policy-makers were reconsidering the mix of malaria control strategies, so rather than studying a particular control technique (such as spraying), a comprehensive analysis was attempted of the entire programme. The advantage of this approach was that it met decision-makers’ needs and could encompass a mix of strategies; the disad- vantage that evaluating a routine control programme is far more difficult than appraising a particular strategy. The study also attempted to take the social perspective (including private costs) frequently recommended but rarely followed in practice in developing country studies. It thus illustrates both how cost-effectiveness analysis can be applied to draw conclusions of relevance to policy-makers, and the methodological issues arising from the use of economic evaluation in a developing country context.

Address for correspondence: London School of Hygiene and Tropical Medicine, Keppel St, London WClE 7HT, U.K.

*The strategies evaluated here were those in general use in 1983-5; experiments were and are being made with other approaches.

2. MALARIA CONTROL POLICIES

Russell [l] classified the measures for prevention of malaria in individuals and for larger scale control of the disease as measures to prevent mosquitoes from feeding on man; prevent or reduce the breeding of mosquitoes by eliminating collections of water or altering the environment; measures to destroy the larvae of mosquitoes; destroy adult mosquitoes; and eliminate malaria parasites in the human host. In general, the choice of method(s) depends on the various epidemiological types of malaria and the specific locations in which it is to be applied.

Malaria endemicity varies markedly across Nepal. The southern edge of the Outer Terai (the plain bordering India) is of low receptivity to malaria; the Inner Terai and adjacent forested slopes are of mod- erate receptivity; the cultivated valleys to the North between 2000 and 4000 ft are of low receptivity; and the parts of the country above 4000ft have little or no transmission of malaria. Virtually the entire country up to 4000 ft is covered by the malaria control programme, which has been relatively successful in its control efforts.

Malaria control operational strategies consist of case detection through various mechanisms followed by blood slide examination and treatment of confirmed positive cases; and spraying with residual insecticides. These activities are common to both ‘unintegrated’ districts (population 6.1 m in 1984) where malaria control is carried out by the National Malaria Eradication Organization (NMEO) and ‘integrated’ districts (population 2.9 m in 1984) where malaria control is one of a number of services provided by the staff of the Integrated Community Health Services Development Project (ICHSDP).*

Case detection methods are characterized as active or passive. In active case detection (ACD), houses are

965

966 ANSE MILLS

visited monthly to collect blood smears from all people with a present or past history of fever or of travel to highly infected areas. In the NMEO districts, these visits are made by malaria field workers (MFW); in the ICHSDP districts, they are made by multi-purpose village health workers (VHW). Passive

case detection (PCD) describes methods where the malaria patient takes the initiative to seek treatment.

The main methods consist of detection at hospitals and health posts [PCD (H)]: detection in NMEO districts by volunteers supplied with the means to take blood smears and give presumptive treatment [PCD (V)]; and detection by specialized malaria clinics [(PCD MC)].

All slides are sent to malaria-specific laboratories. Presumptive treatment is given to all people from whom blood slides are collected and radical treatment given at home to confirmed cases. Malaria clinics are an exception to this pattern: a slide is taken and examined on the spot. and immediate radical treatment given to positive cases.

Spraying is considered to be required whenever the

village-level Annual Parasite Index (API)* minus imported cases rises above a certain level. Malathion is the insecticide of choice in most of the Outer Terai where the vector has been shown to be resistant to

DDT. DDT (or a substitute if DDT is unavailable) is used everywhere else.

3.STUDY.METHODOLOGY

The essence of cost-effectiveness analysis is the

comparison of different ways of achieving an objective. The objectives (and choices) relating to malaria control can be specified at different levelst:

(I) the objective of improving health (choice of malaria control versus other means of health improvement);

(2) the objective of malaria control (choice of vector control versus case detection and treatment and various mixes of both);

(3) the objectives of(i) vector control and (ii) case detection and treatment (choice of strategies for each);

(4) the objective of delivering a pre-determined strategy (choice of means of slide examination, choice of organizational pattern, etc.).

The importance of these levels is both conceptual and practical, the choice of effectiveness indicator being determined by the level of analysis. If the policy issue is whether to invest in malaria control or in

*All cases detected expressed per 1000 population. tThis paper includes in ‘cost-effectiveness analysis’ both

analyses which rely on measures of intermediate output (e.g. indicators of activity) and those which rely on final output measures (e.g. indicators of change in health). Purists, which those who work on developing country issues cannot afford to be, might argue that only the latter constitutes cost-effectiveness analysis.

another health programme, then the objective is at the first level, that of improving health, and the measure of health consequence used must be one that is common to many different health programmes, for instance increasing years of healthy life. If the policy issue is how best to maintain malaria control, then the objective is at the second level, and the measure of health consequence used must be one such as cases prevented which is relevant to comparisons between. for example, case detection and treatment and vector control. At the third level, effectiveness indicators can be used which are specific to particular objectives: for example, cases detected and treated for case detection and treatment strategies. At the fourth level, the desirability of malaria control and of existing control strategies is taken as given, and emphasis placed on discovering the least cost way of delivering the com- ponents of a control strategy, for instance examining a blood slide or spraying a house.

This research attempted to produce information relevant to all four levels of objective, though the results reported here relate to levels 2, 3 and 4, thus omitting the issue of whether malaria control is a cost-effective way of improving health.

Government malaria control activities were analysed in order to identify the resources used with the objectives they served and within each main objective, to allocate resources used to operational activities. Because malaria transmission and control costs differ between areas, an aggregate national analysis would have been meaningless. Five districts were therefore selected for detailed cost analysis, representing the main characteristics considered to influence the costs of malaria control. In ICHSDP districts, the cost of malaria control had to be separated from other health programmes. The costs of higher management levels were allocated to districts according to criteria which reflected the demand that districts made on these levels.

Pricing practices recommended for economic ap- praisals in developing countries were followed as far as possible (21. Resources valued in financial prices were converted to economic prices which reflected social opportunity cost. Traded goods and services were valued at world (border) prices. Valuing non- traded goods by the desirable method of separating the inputs used into labour, traded goods and non- traded goods would have been too time-consuming so the short-cut of a conversion factor was used, adjusted for the estimated foreign exchange component of each non-traded good. Economic appraisal practice in Nepal seemed not to use a savings premium or consumption weight and there were no strong reasons for choosing any particular weight; therefore no adjustments were made to efficiency prices.

Data on the costs borne by households were obtained from two surveys. A short questionnaire was completed for 3253 malaria patients in 6 districts selected by stratified random sampling (the ‘patient survey’). A much fuller questionnaire was used for

The economic evaluation of malaria control technologies

Table I. Spraying costs per capita per cycle

DDT’ Malathion Ficam

Morane Ruuandchi Ruuandehib ParsaC Morang Ruuandehi

967

Cost componenr

Per capiro fixed -district

(NME@ (N’MEO) (NMEO) (ICHSDP) (NMEe) (N’MEO)

0.38 0.28 0.36 0.20 NIA N/A ---total 2.74 2.11 2.71 N/A N/A NiA

Per capita variable 8.31 1.92 12.99 16.82 17.22 16.97 Total district fined plus variable 8.69 8.20 13.35 17.02 NIA NI’A Total per capira cost I I .05 ’ 10.03 15.70 N/A N/A N!A

‘I.5 g of a.i. per m’. b50% at I g of a.i. per m*; 50% at 2 g. ‘2g of a.i. per ml.

12 months in a case-control study in two smaller areas and 867 malaria patients and their households (and matched controls) interviewed (the ‘household survey’). Information on use of and expenditure on private sources of treatment and on days of incapacity were obtained by both surveys. The household survey was able to go into much greater depth on the disability and debility caused by malaria and its effects on household working practices and allocation of time.

Information on final output measures was not readily available from the routine information sys- tems of the NMEO and ICHSDP: the only measure available and likely to be accurate was that of ‘cases treated’. The API ought to provide information on total cases and is routinely reported but, because of differences in blood examination rates, is a poorer indicator of malaria incidence in ICHSDP than in NMEO districts. The results reported here therefore rely largely on intermediate measures of output.

4. STUDY RESULTS

In this section some of the results are presented of the analysis of the costs and cost-effectiveness of malaria control strategies in three NMEO districts, two in the Outer Terai (Morang and Rupandehi) and one in the Hills (Ilam); and in two ICHSDP districts, both in the Outer Terai (Saptari and Parsa). The major policy issues concerned the choice of insecticide and case detection method, so results relating to these are reported. Within each sub-section, costs to the government are considered first and then costs to individuals. ICHSDP costs above the district level could not be identified, so district-level costs are presented for all districts, and total costs for NMEO districts only.

4.1. Vector control strategies

Gocernment costs. In the mid 198Os, the only routine vector control method was residual spraying.

*The Ficam costs originate from a later study [3] which analyzed operational and chemical costs only; hence only variable costs can be included for Ficam in Table 1.

?A11 costs relate to 1984. Rs 16.46 = $1. JRe-plastering is anyway done regularly, so this will usually

not represent an additional cost.

Three insecticides can be compared: DDT, malathion and Ficam.* All these insecticides have been field- tested in Nepal and found to be effective in killing mosquitoes. Therefore indicators of activity, adjusted for the period of residual action, are adequate measures of output. However, comparison of alterna- tive vector control methods or of vector control and case detection and treatment requires an indicator of output such as cases prevented. Unfortunately, there is very little recent evidence in Nepal on the effectiveness of spraying in preventing cases. It appears that in some areas spraying can be highly effective, in others not effective due to vector or human behaviour or unnecessary because of local influences on transmission (for example transmission occurring in forest areas not in the sprayed, settled villages). Not infre- quently, poor application of insecticide (especially timing of spray cycles) has reduced any potential effectiveness. Table 1 therefore compares DDT, malathion and Ficam in terms of cost per person protected.

DDT costs per capita are very similar in Morang and Rupandehi, at around Rs 8 per capita per cycle (district costs only)t. Malathion costs in Rupandehi are 60% higher, and in Parsa, more than double (largely because of a higher dosage in Parsa). Ficam is over twice as expensive as DDT, and about 25% more expensive than malathion. Adjusting for the period of residual action [4] accentuates these differences. For example, the district-level cost of DDT is Rs 1.41 per person per month of protection (average of Morang and Rupandehi) whereas for malathion in Parsa it is Rs 5.67.

Costs lo indiuiduafs. Residual spraying requires little action by individuals and imposes few costs on them. They have to vacate their houses and remove foodstuffs and utensils for 3 hr, but suffer no side- effects as long as safety intructions are followed [3]. Domestic birds and animals are occasionally affected by insecticides but spraying also reduces the nuisance effect of insects and kills bed bugs. Because of the smell and residue, some housholds refuse to allow their houses to be sprayed or replaster soon after spraying.: This appears to be mainly a problem with malathion; in contrast, Ficam lacks both smell and residue.

None of these costs or consequences for households are readily quantified or valued or are of major

968 ANNE MILLS

Table 2. Dl,tnct-le\el case detectron, parasitology and treatment costs per copira. slide and per case

Cost measure

cost per cap,ta Cost per slide Cost per case

MOTXl_e’ Rupandehi’ Ilam’ saptan Parsa (NMEO) (NMEO) (NMEO) (ICHSDP) (ICHSDP)

(Rs) (Rs) (Rs) (Rs) (Rs)

2.51 3.53 8.90 0.84 1.07 13.52 14.54 25.51 14.82 10.82

1652 707 6777 965 881

“Cost of the surveillance programme whxh covers ACD and PCD except volunteers.

importance, so no attempt at quantification or valuation was made. However, they may be of some importance in influencing compliance and thus coverage rates.

4.2. Comparison of case detection and treatment strat-

egies in integrated and unintegrated districts

The integration of malaria control with other health services has been a strong policy thrust for a number of years throughout the world [5]. The realtive costs and effectiveness of vertical and integrated patterns of organization are thus of considerable interest. Table 2 shows the cost of case detection, parasitology and treatment at district level for the 5 districts, expressed per capita, per slide taken and per case detected.

There is a striking difference between the per capita costs in NMEO and ICHSDP districts. NMEO costs in Mordng are more than double and in Rupandehi, more than triple the costs in the ICHSDP districts of Saptari and Parsa. Costs per slide in the integrated

districts are close to those in the comparable NMEO Terai districts. Costs per case in Saptari and Parsa are higher than that of Rupandehi. though below that of Morang.

This cost pattern can be explained in the following way. In ICHSDP districts, case detection is only one of 8 tasks of a VHW and treatment is carried out when required, using staff time diverted from other activities. The system exhibits economies of scope but not, in relation to malaria control, of scale. The ICHSDP fixed costs (e.g. the VHW network) can be shared between many activities and since malaria work diverts the staff from other duties, marginal cost is close to average cost. Cost per capita, per slide and per case will rise as incidence rises.

In NMEO districts. staff do only malaria control and the costs are dominated by the costs of ACD, which are largely fixed and related to population size. The marginal cost of detecting and treating additional cases is very small (i.e. the NMEO enjoys substantial economies of scale) so even if incidence rises, cost per capita and per slide remain steady. Cost per case is highly sensitive to the number of cases, falling sharply as cases rise. Thus the much higher incidence in Rupandehi than Morang is reflected in a much lower cost per case. The differences stemming from incidence are accentuated in the Hill district, Ilam, by the higher costs arising from difficult terrain and a scattered population. The high costs of a

vertical organization when few cases are available to be detected is evident.

A most important consideration in this comparison of integrated and unintegrated districts is whether cases are missed. Total cases were not known and could not be investigated within the scope of this study, but it is likely that the NMEO detected a high proportion of total cases, and the ICHSDP a much lower proportion. Thus the generally lower cost of integration is to some extent offset by a less efficient detection system.

4.3. Comparison of case detection and treatment mechanisms

Government costs. The discussion above of NMEO case detection and treatment aggregated all detection mechanisms (except volunteers). To compare the relative efficiency of different mechanisms and particularly to contribute to the policy debate on active and passive methods [6], these had to be separated. Two problems arose. Firstly, many costs are joint between the case detection mechanisms so the allocation of costs is approximate. Only the volunteer programme has its own accounts. Moreover, the costs in the two ICHSDP districts were already approximations, so further disaggregation was considered inappropriate.

Secondly, the availability of different detection methods will influence each others’ effectiveness. The yield of ACD, for example, will be affected by whether a volunteer is available in the neighbour- hood. Data from the patient survey suggested that the population cannot be distinguished clearly into two groups, one which waited for an ACD worker and the other which used PCD mechanisms. The better the PCD network, the lower is likely to be the yield of ACD. Moreover, the household survey showed that the great majority of patients knew about the availability of treatment centres. This widespread aware- ness of PCD mechanisms, together with the evidence presented earlier on the relatively high cost of NMEO case detection activities, emphasizes the importance of comparing the relative costs of ACD and PCD mechanisms despite possible interaction between them.

This section therefore considers the cost of the main case detection and treatment mechanisms in NMEO districts, namely ACD, detection by volunteers [PCD (V)] and detection by malaria clinics [PCD (MC)]. Table 3 summarizes their relative cost (to the NMEO) and relative contribution to case detection.

The economic evaluation of malaria control technologies

Table 3. The relative contribution and cost’ for the NMEO of di!Terent case detection methods

Case detection method Morang Rupandehi llam

ACDb -% of total cases 56% 42% 65% -% of total case detection costs 78% 77% 76% -NMEO cost per case Rs 2059 Rs 791 Rs 6316

PCD (V) -% of total cases 19% 33% 24% -% of total case detection costs 18% 16% 24% -NMEO cost per case Rs 1518 Us 336 Rs 5523

PCD (MC) -% of total cases 2% 19% - -% of total case detection costs 2% 5% - -NMEO cost per case Rs 1048 Rs 98 -

‘District-level costs only, excluding administration and regional and national costs. blncludes all cases detected bv the MFW, including from follow-up of previous

969

infections.

The pattern in all three districts is consistent: ACD incurs the highest cost per case, with PCD (V) cheaper, and PCD (MC) cheapest. In terms of the share they absorb of total case-detection costs and the return in terms of cases detected, the pattern is consistent across all three districts that ACD absorbs a considerably higher share of total case detection costs than its share of total cases (though it still detects the majority of cases in two of the three districts).* It is apparent that the same mechanism can have widely differing costs depending on the level of use and number of cases detected. For instance both PCD (V) and PCD (MC) in Rupandehi are very low cost ways of detecting cases, yet the cost per case of PCD (V) in Morang is many times higher. This can be explained by fewer cases per volunteer and fewer (and thus less accessible) volunteers and more costly support costs per volunteer.

Costs fo individuals. A fair comparison must take account of costs other than those falling on the government, namely the opportunity cost to volunteers of spending time on malaria detection, and any difference in costs to individuals arising from the different mechanisms. These differences may stem from a propensity to pay for private sources of treatment and losses due to inability to work through illness that differ between the case detection mechanisms. Inclusion of these costs is important not only in the comparison of case detection mechanisms but also in the comparison of this strategy of malaria control with that of vector control, since more cases of malaria will occur with the former than the latter strategy, thus imposing greater costs on individuals.

(a) Costs to volunteers. Given the recent emphasis on the use of volunteers in many different health

*Ideally, an estimate would be made of the additional cost and cases detected if ACD were to be added to passive methods. Since the performance of PCD is affected by the existence of ACD, this estimate cannot be made but the incremental cost per case detected of ACD is likely to be very high.

tEven if a cost were included, the effect would be insignifi- cant: for example, 50 slides a year, at 10 min per slide, gives a total of 84 hr per year, or one day’s work.

programmes, it is important to at least consider what costs this may impose on them. The costs to volunteers of slide collection and giving presumptive treatment will depend primarily on the extent to which these activities are compatible with their main occu- pations and the amount of time required. A number of volunteers are merchants (e.g. drug-sellers), who may find volunteering increases their sales, and the duties required are not onerous (for example 35-50 slides per volunteer in the three districts in 1984). For both these reasons, no opportunity cost is attributed here to volunteers’ time.t

(b) Private expenditure. Despite the availability of free treatment, the patient and household surveys revealed that patients themselves spend sizeable sums on treatment. Moreover, these sums differed by case detection mechanism. Since the surveys also showed that expenditure patterns were very different between districts (and probably within districts), possibly reflecting differences in cash availability and access to treatment facilities and drug stores, only the results from the patient survey for Rupandehi and Morang, two of the districts for which NMEO costs are available, are reported here. (Full analysis of the patient survey is available in Ref. [7] and of the household survey in Ref. [8].)

Table 4 shows mean expenditure by case detection method for these two districts. The most marked feature is the difference in the level of costs between the districts. There are good grounds for suspecting exaggeration in Morang, though certain characteristics predominated in Morang patients (e.g. high

Table 4. Private expenditure (Rsy per malaria case in Morang and Rupandehi by case detection method

Morang Rupandehi

Mean SD n Mean SD n

ACD 113.86 177.48 I20 26.91 62.15 566 PCD (V) 80.63 202.07 80 30.93 93.88 434 PCD (MC) - - 0 53.23 67.85 404 Allb 82.65 146.08 358 31.59 73.60 2022

‘Includes expenditure on fees, drugs, laboratory examinations; special foods, sacrifice and worship, and travel.

bIncludes cases picked up through other case detection mechanisms which are not shown separately.

970 ANNE MILLS

Table 5. Mean dajs oi v.ork lost by case detection method. !&rang and Rupandehl

Rupandehi Morang Caie detrcuon method Msun SD n Mean SD n

ACD 8.8 6.8 JO7 15.3 22.6 88 PCD 0’) 9.1 7s 368 24.6 442 20 PCD (MC) 7.3 6.8 279 - - - All PCD’ 82 6.5 IO90 15.1 23.8 I55 AllO 8.3 6.6 I564 14.5 22.9 271

“Includes PCD cases which were not clnsslfied by mechanism. ‘Includes cases picked up through other case detection mechanisms

which are not shown separately

proportion of imported cases, long periods of illness, repeated episodes per person) that were associated overall in the patient survey with higher than average expenditures.

In Morang, the mean expenditure of individuals detected through ACD was more than if they were detected through PCD methods: much of this difference was due to drug purchases,* possibly reflecting a preference for self-treatment. In Rupandehi, individuals attending PCD (V) and (MC) spent higher sums than those detected through ACD. PCD (MC) attenders stood out as spending significantly more (P < 0.05) than all other types of case, perhaps because of the close geographical proximity of malaria clinics to commercial sources of treatment.

(c) Days of work lost. Individuals lose production and/or income if malaria prevents them from carrying out their normal activities. However there may be no, or a lesser loss in household income or production if other, underemployed household members replace the ill person; or no, or a lesser loss of production to society if the work that would have been done by the patient outside the household is done instead by a previously under-occupied worker. This loss should affect the choice of case detection method if use of one method rather than another leads to a longer period of incapacity.

In the patient survey, differences in days of work lost between case detection methods were relatively small in Rupandehi (see Table S), but cases attending malaria clinics seemed to have a particularly short period of incapacity, of only 7.3 days compared to 8.8 for ACD. Conclusions are difficult to draw for Morang because the figures span a very wide range and the sample size is small for several of the mechanisms. There was little difference in days of work lost between ACD and all PCD mechanisms though PCD (V) cases had a particularly long period of incapacity.

The traditional. crude approach to estimating the income or production loss due to malaria is to multiply days of work lost (or merely days of illness)

‘Information on the types of drug uas not sought. tThis makes the as&nption--which can at present be

iustified onlv on the basis of the broad similarity of the josses per day from the two surveys-that the losses from one Terai district are applicable to another.

by some measure of the average or minimum wage [9]. However, information from the household survey

showed clearly the role of the household in mediating the impact of a sick worker. The great majority of households drew on household reserves of labour, primarily of adults rather than children: only 10% of helpers in one area and 5% in the other were hired labourers. 75% and 50% of the households stated that the illness of a working household member caused no problems. 91% and 80% of households experienced no cash loss (other than hired labour and medical care costs) because of the malaria episode: the loss incurred by the balance of households primarily arose from loss of income from wage labour. Over 70% of households in both areas did not think household production would suffer because of the malaria episode.

If the financial losses that were reported are expressed per case of malaria, they represent a mean of Rs I6 and Rs 30 in the two areas. This contrasts with figures of Rs 64 and Rs I58 for these areas if the crude approach to valuation is taken of multiplying days of work lost (including an allowance for days of disability) by the local wage rate. The first estimate is likely to be an underestimate of household income and production loss because it represents immediate financial losses only and excludes the non-financial production losses of the minority of households experiencing them; the second, an over-estimate since it takes no account of the availability of spare capacity within the household or of variation in the value of labour by season.

The sample size of the household survey is too small to permit a detailed analysis of losses by case detection mechanism. Moreover, the period of work lost is influenced by factors such as the species mix of cases, the delay between start of the fever and presumptive treatment and whether or not the patient has had previous episodes of malaria, all of which vary by district [7]. Since the data in the household survey were not from Rupandehi or Morang, to estimate a cost for household production/income loss by case detection method the number of days lost from the patient survey was multiplied by the cost from the household survey, expressed per worker per day of complete disability.? It is assumed that this represents not only a household but also a social cost, since the period of peak malaria transmission coincides with the peak period of farming activity.

The costs per case of case detection and treatment including private expenditure and the value of days of uork lost are shown in Table 6. The inclusion of private costs does not change the ranking of methods. However the absolute value of private costs differs markedly between the methods, with PCD methods giving rise to higher private costs. This is particularly true in the case of PCD (MC) in Rupandehi, though a high private treatment cost is offset by a low value of days of work lost (because of the rapid treatment).

The economic evaluation of malaria control technologies 971

Table 6. Cost per case of case detection and treatment including value of costs due to private treatment and inability to work

Morang Rupandehi Case detection method (Rs) (Rs)

ACDa -government cost per case 2059 791 -private cost per caseb 81 25 -value of days of work lost 37 26 -total cost per case 2177 842

PCD (V) -government cost per case 1518 * 336 -private cost per caseb 77 29 --value of days of work lost 71 27 -total cost per case 1666 392

PCD (MC) -government cost per case 1048 98 -private cost per case N/A 51 -value of days of work lost N/A 22 -total cost per case N/A I71

“Includes all cases detected by the MFW, included from follow-up of previous infections.

bFrom Table 4. adjusted by a conversion factor.

The most uncertain variable in Table 6 is the value of days of work lost. Even if it were to be doubled, it would not change the ranking of mechanisms.

5. POLICY IMPLICATIONS

It is important to take an economic evaluation a step further, to explore the implications of policy changes. This section draws conclusions on the results in Section 4 and briefly mentions a few of the most important policy changes examined in the research (presented in detail in Ref. [8]). This discussion is structured in terms of the levels described in Section 2.

Lerel2: Choice of cector control L’S case detection and treatment

Clear-cut conclusions on the relative cost-effectiveness of vector control and case detection and treatment and the optimum mix of the two strategies was impossible in the absence of reasonable information on their effectiveness. However, the unit cost data were illuminating. The annual NMEO district cost per capita of case detection and treatment was around Rs 3. In contrast, the annual cost per capita of spraying (assuming the usual two cycles) would be around Rs I7 for DDT and Rs 30 for malathion. This difference between the two approaches suggests that considerable intensification of case detection and treatment would be possible before costs would ex- ceed those of spraying. For example, doubling the frequency of case detection would cost an additional Rs 3 per capita per annum compared to the cost per capita of introducing spraying of Rs 9 per cycle for DDT and Rs 15 for malathion.* Thus fortnightly surveillance could be cost-effective if it reduces the need for spraying.

*A substantial difference remains even if allowance is made for the increased private expenditure and work days’ loss of increased cases.

Leoel 3: Choice of means of case detection and treatment and means of vector control

ACD has been described as “having no place in long term malaria control” [6]. In Nepal, because of the recent development of PCD methods, ACD is left with fewer cases to detect but with high fixed costs. Replacing ACD by a strengthened PCD (V) system would cost an estimated 15% of the ACD cost in Ilam, 28% in Rupandehi and 56% in Morang. In addition, savings would be made in parasitology costs. The most crucial issues are whether knowledge of passive sources of treatment and concern to obtain treatment would be sufficient to maintain reasonable levels of case detection.

Conclusions on the merits of different insecticides are relatively straightforward. DDT is the most economical insecticide, followed by malathion. Of the newer generations of insecticide, those which are low in volume and weight such as Ficam have distinct operational advantages in Nepal though are still more expensive than DDT and malathion. The more expensive the insecticide used, the greater the proportion of variable costs in total costs and thus the more worthwhile become strategies such as selective coverage and focal spraying which limit the quantities used.

Level 4: Choice of ways of organizing an activity

Analysis of existing activities identified a number of changes that might improve the efficiency of the malaria control programme [8], but perhaps the most important issue is that of integration. In the districts evaluated, case detection and treatment costs were considerably lower in ICHSDP than in NMEO districts, but APIs were also lower. On the assumption of a linear relationship between per capita cost and API in the two ICHSDP districts, it can be estimated that costs of case detection and treatment in an integrated district were roughly half those of an NMEO district at an API of around 1.5. This suggests that integration might be efficient at low levels of malaria.

A crucial issue, however, is the extent to which the activities of integrated districts are adequate to prevent a major rise in transmission. It is of note that despite regular warnings by experts that case detection and treatment activities in integrated districts are very poor and many cases of malaria missed or treated late, there has not been, in those districts integrated up to 1984, any sign of a major increase in malaria. NMEO staff attribute this to their low receptivity. Since they are very similar to adjacent NMEO districts which report rather higher APIs and have more expensive control programmes, this suggests an element of overkill in NMEO strategies for these areas and the scope for a more economical programme. However the ability of the ICHSDP to contain malaria in districts of higher receptivity has’ not been tested, nor its costs in circumstances of relatively high transmission.

972 ANNE MILLS

6. METHODOLOGICAL CONCLUSIONS

The research reported here illustrates both the scope for and difficulties of applying cost-effectiveness analysis to disease control programmes in developing countries. The research was perhaps fortunate in encountering an information system that made analysis of programme costs relatively straightforward. However, in most countries and programmes, cost data tend to be the most readily available of all types of data, simply because they are required for accounting. In contrast, until recently health programmes have often not been required to prove that they are effective in terms of indicators of change in health: hence the difficulties of obtaining the information required on effectiveness from the Nepalese malaria programme.

The social perspective adopted by the research raised issues of whether the time lost by a sick person is compensated for by an increased time input by other household members, and how any time loss might be valued. The Nepal data suggest that assuming that all the period of disability is lost to the household, and that this period should be valued at the local wage, will overestimate the actual income or production loss to the household.

Indeed, the Nepal study emphasizes the importance of including consideration of the role of the household in any cost-effectiveness analysis of disease control. The household uses its resources to cope with illness, it finances preventive activities, it influences the effectiveness of government preventive activities (for instance by whether or not houses are replastered after spraying) and it affects the cost-effectiveness of case detection and treatment activities by its decisions on the use of services. Household surveys such as that employed here can help to shed light on some of these issues.

Despite the difficulties the research encountered in the application of cost-effectiveness analysis, the research also underlines the importance of this type of evaluation. Malaria control absorbs over 20% of Ministry of Health expenditure and is in regular contact with about 9 m of Nepal’s 17 m population. Policy-makers face important choices as donors provide less insecticide and as the nature of the

malaria problem changes with population move- ments, environmental change, and development of parasite resistance to insecticides and drugs. Cost- effectiveness analysis can help malaria control programmes improve their efficiency by raising pertinent questions and bringing home the resource implications, for both the government and households, of choosing different strategies.

Acknow/edgemenfs--I am most grateful for the help provided by the NMEO, particularly by Dr M. 8. Parajuli, Mr S. L. Shrestha, Mr R. G. Vaidya and the Regional and District Malaria Officers. and by the ICHSDP, particularly Dr B. B. Karki and the District Health Officers. The fieldwork for the household survey was done under contract by New Era and supervised by R. P. Shrestha and T. N. Rajbansi, and Hilary Dimond prepared the data for analysis and did preliminary analyses. The research was funded by the Overseas Development Administration. John Picard provided valuable comments on a draft.

I.

2.

3.

4.

5.

6.

REFERENCES

Russell P. F. Malaria: Basic Principles Briefl? Slated. Blackwell, Oxford, 1952. Overseas Development Administration. A Guide IO Pro- ject Appraisal in Developing Counrries. HMSO, London, 1972. Phillips M. and Mills A. The operational cost of insecticide spraying for malaria control. A case-study of Nepal. J. Trap. Med. Hyg. 94, 130-139, 1991. Fontaine R. E. House-spraying with Residual Insecri- rides with Special Reference IO Malaria Conuol. WHO/VBC/78.704, World Health Organization, 1978. Mills A. Vertical versus horizontal health programmes in Africa: idealism, pragmatism. resources and efficiency. Sot. Sri. Med. 17, 1971-1981, 1980. World Health Organization. Malaria control as part of primary health care. Technical Report Series no. 712, Geneva, 1984. Mills A. J. and Colbourne M. J. Economic study of malaria in Nepal: analysis of data collected by the NMEO and ICHSDP. Evaluation and Planning Centre, London School of Hygiene and Tropical IMedicine, 1985. Mills A. J. The application of cost-effectiveness analysis to disease control programmes in developing countries, with special reference to malaria control in Nepal. Ph.D. thesis, University of London, 1989. Mills A. J. The economics of malaria control. ‘&in- (aria-Waiting/or rhe Vaccine (Edited by Targett G.). Wiley, London, 1991.

Documents

The economic evaluation of malaria control technologies: The case of Nepal