THE COST-EFFECTIVENESS OF PERMETHRIN-TREATED BED NETS IN AN AREA OF INTENSE MALARIA TRANSMISSION IN WESTERN KENYA

ITN study.

The group-randomized ITN study was conducted in a 500 km² area of western Kenya in the contiguous sites of Asembo and Gem. Details of the study are reported elsewhere.¹⁰ The intervention was conducted from January 1997 to March 1999 in Asembo and from January 1998 to December 1999 in Gem. The primary objective of the study was to measure the effect of ITNs on all cause mortality of children 1–59 months of age. The study also estimated the effect of the intervention on all cause sick child visits (SCV) to local clinics (in Asembo only). To estimate the community effect of ITNs, epidemiologic data were linked to a base map generated using the global positioning system.⁹ Costs for the intervention were estimated from project expenditure records. Information on community costs was gathered through interviews with field staff and community members.

Resource use was categorized in accordance with three phases of the project: start-up, distribution, and insecticide treatment. In the start-up phase, meetings were held in each village to inform community members about the study rationale and design, and to provide an opportunity to ask questions. In the distribution phase, program staff distributed 45,667 polyester, 100-denier, 156-mesh bed nets, pre-treated with the target dose of 0.5 g of permethrin/m² of netting. This achieved an initial coverage ratio of 1.34 persons per ITN. The ITNs were distributed by project staff at central points in each village following a pre-arranged schedule. Each family was responsible for sending one representative to collect the nets and to receive training in how to hang them. Study staff later visited all the houses to ensure that ITNs were properly deployed. Re-treatment of nets with permethrin was scheduled every six months during the two-year intervention. Prior to each re-treatment exercise, villages were notified in advance of the day the ITNs were to be re-treated. Program staff visited each compound and re-treated the nets with the help of community members, who provided water and labor.

Methods for the economic evaluation.

The framework for the economic evaluation (Table 1) was adapted from Drummond and others.¹¹ Costs and effects for all parties affected by the intervention were included (i.e., users, community, and program providers). Both financial and economic costs were estimated. Research costs were not included. Financial costs represent funds paid for the implementation of the intervention by the project and are spread over a four-year period (1996–1999). Cost-effectiveness ratios are based on net (average) annual economic costs. It is not uncommon in ITN trials for the effectiveness data and cost estimates to be spread over different time periods.^6,12 In this study, costs covered a four-year period and effects were measured over a two-year period (two years in Asembo and two years in Gem). The method used to measure and value capital costs involved annuitizing the initial capital cost over the useful life of each asset; that is, an equivalent annual cost has been estimated that is then added to the annual recurrent estimate.¹¹ This process reflects the value-in-use of capital terms, rather than reflecting when the item was purchased. Economic costs differ from financial costs in two respects: an equivalent annual capital cost has been calculated and the value of community labor included.⁶ This was done to more accurately reflect the opportunity cost of resource use. Financial costs and economic costs are converted to U.S. dollars at the prevailing rate of exchange and deflated to 1996 values using the U.S. gross domestic product deflator. The official exchange rate at that time (1 U.S. $ = 78 Kenyan Shillings) was assumed to reflect the scarcity of foreign exchange.

For each phase of the program, costs were separated into capital and recurrent costs. The recurrent costs borne by the providers were insecticides, detergent, personnel, office supplies and services, and transport. The community recurrent costs included water, transport, and the time spent washing, drying, and re-deploying the ITNs. Capital costs, all incurred by the provider, included bed nets, vehicles, buildings, office equipment, furniture, and supplies for impregnation. The total annual cost of the ITN program was obtained by summing the provider and community costs (both recurrent and capital).

Costs for project personnel, consumables, and transport were valued using current market rates. The insecticide was valued at the duty-free purchase price; the effect of variation in insecticide price was evaluated in the sensitivity analysis. Project personnel were valued according to existing salary scales. The difference in salaries of local and expatriate staff was accepted on the basis that the opportunity cost of expatriates was higher because of easier access to international labor markets.⁶ The community members who washed and re-deployed the nets did not receive a formal wage. Their time was valued using the estimated daily wage for a farm worker (US $1.30 per day). The effect on net annual costs of varying the value of unwaged time and the number of senior staff salaries was also evaluated in the sensitivity analysis.

The costs of vehicles and equipment were annuitized over an expected life of 10 years and at a discount rate of 3%. ITNs were treated as a capital item and converted to an annual equivalent cost assuming a 5-year life span and a 3% discount rate. The effect on total costs of varying the discount rate was also tested.

Effectiveness.

Three measures of effectiveness were calculated. First, the number of all-cause child deaths prevented by ITN use was estimated.¹³ Second, life-years gained (LYG) were estimated by applying life expectancy at mean age of death of trial participants to deaths averted. A 3% discount rate was then applied to derive an estimate of discounted life-years gained (DLYG).⁵ Third, morbidity was measured in terms of the number of all-cause sick child clinic visits averted.¹⁴ The effect of ITNs on all-cause mortality was apparent only in infants 1–11 months of age, but all children less than five years of age benefited from reduced morbidity.^13,14 Since an impact on both child morbidity and mortality was observed in compounds lacking ITNs located within 300 meters of intervention villages, estimates of effectiveness with and without this community effect have been calculated. A detailed description of efficacy measures including spatial analyses and the effect of ITNs on nearby compounds lacking ITNs are reported in detail elsewhere.^9,13,14

Resource savings.

Resource savings to the health sector and households were also measured. Savings to the health sector were based on the potential reduction in health service use as measured by the difference in the absolute number of all-cause sick child clinic visits to local health centers averted by ITNs.¹⁴ The potential saving to the clinic was based on an estimate of the cost of drugs prescribed by clinic staff to sick children ≤59 months of age. These drugs, when available, are normally provided free of charge by government clinics for this age group, and non-dispensing thus provides a direct saving to the clinic. An estimate of the drug costs was obtained from a previous study conducted in similar peripheral health facilities in a neighboring province (U.S. $0.44 [1996 U.S. $] per child per visit).¹⁵ We assumed that a reduction in sick child visits would not result in a reduction in the infrastructural costs (e.g., maintenance and salary costs) of the clinic in the short term.

At the household level, saved treatment costs and the economic benefit of postponed funeral expenses were calculated. Health care expenses were assessed during a series of cross-sectional surveys¹⁶ and broken down into five categories paid by the household for treating children less than five years of age: medicines; doctor or clinic fees; laboratory (e.g., needles, syringes, and tests); traditional healer (for fees, herbs, etc.); and other items such as food and travel costs associated with obtaining health care.¹⁷ It is important to note that funeral costs can be a serious economic burden to poor families. It is therefore relevant in such contexts to estimate the economic benefit associated with delaying funeral payment. Based upon an informal survey among local residents, funeral costs were estimated to be approximately $100 per funeral. This consisted primarily of the cost of the coffin and other costs of arranging the funeral, which may include providing meals for numerous guests over a period of several days. The effect on net annual costs of lowering funeral costs by 50% and of varying the discount rate was tested in the sensitivity analysis.

Methods of analysis.

Six cost-effectiveness ratios were calculated: 1) cost per child death averted; 2) cost per LYG; 3) cost per DLYG; 4) cost per SCV averted; 5) cost per person protected; and 6) cost per ITN. In addition, net cost-effectiveness ratios were calculated by subtracting resources saved from the total economic cost divided by the relevant outcome measure. Two sets of net cost-effectiveness ratios were calculated: those that did not account for the community effect, and those that did.

The sensitivity of our results to variation in the discount rate was examined by increasing the rate to 6% and to 10%.^5,7 The effect on net annual costs of reducing the frequency of re-treating ITNs from twice a year to once a year was also determined.¹² The cost of insecticide was reduced by 20% to reflect the possibility of using cheaper insecticides in the future^5,18 The cost of insecticide was also increased by 20% to explore the effect of using duty-free prices. Since it is likely that fewer senior staff salaries would be required in a non-research setting, the effect of a 50% reduction in staff salaries has been calculated. Finally, the effect of varying the value of the time of community members who washed and redeployed the ITNs as well as a reduction in funeral costs was estimated.

Total financial costs.

The capital and recurrent financial costs of the intervention in local currency, excluding research costs, are shown in Table 2. The total financial cost of the program was U.S. $725,700. Recurrent costs were 57% of total program costs. Staff salaries represented the largest recurrent cost item at 31%. Insecticide (permethrin) represented 8% of total recurrent costs. A total of 45,667 ITNs were distributed. The annual capital cost of the program was U.S. $320,000, with ITNs being the main capital item at U.S. $230,000.

Total economic costs.

The annual economic cost of the program was U.S. $148,856 (Table 3). The key differences between Tables 2 and 3 is that for the latter, the imputed value of user labor is included and capital costs have been annuitized to provide an equivalent annual cost. Approximately 25% of annual economic costs were attributed to program start-up ($37,214), 40% to ITN distribution ($59,542), and 35% to retreatment ($52,100). The ITNs represented an average annual cost of U.S. $60,193 (40%) and insecticide represented a cost of U.S. $13,707 (9%).

Effectiveness and resource savings.

We report elsewhere that ITNs were associated with a 27% reduction in the number of sick child (less than five years old) visits to local health facilities.¹⁴ This translates to an annual reduction of 2,025 visits per year in a population of 9,375 children less than five years old, or 2,328 visits when the community effect is taken into account. Based on a protective efficacy estimate of 23%,¹³ ITNs were also predicted to reduce all-cause post-neonatal mortality by 61 deaths per year in the same population (2,005 infants 1–11 months old), or 73 deaths when additional benefits to the community are accounted for (Table 4).

The reduced number of clinic visits results in an annual saving to the health services of U.S. $891 (U.S. $ 0.44 per visit), or U.S. $1,024 when adjusting for the community effect (Table 4). Households in ITN villages spent approximately U.S. $6.5 per year less on health care for sick children less than five years old compared with those in the control villages.¹⁷ There were approximately 6,700 households with children in this age range in the ITN villages in Asembo and Gem, and 8,241 households if the communities located within 300 meters from ITN villages were also included. This represents an annual cost saving of U.S. $43,550, or $53,567 after adjusting for the community effect. Savings also accrued to households in the form of postponed funeral expenses with an average discounted saving of $77 per year. When we adjusted for the community effect, annual direct resource savings to the community amounted to $5,621. These estimates are based on 1) an average life expectancy at birth of 49 years;¹⁹ 2) a discount rate of 3%; 3) an estimated reduction in total number of post-neonatal deaths (age = 1–11 months) by 73; and 4) and an average cost per funeral of U.S. $100. This figure was U.S. $4,697 if the community effect was not taken into account (Table 4).

Cost-effectiveness.

Table 4 also shows the cost-effectiveness ratios both adjusted and unadjusted for the community effect. Approximately 62,500 adults and children were protected by the intervention at an estimated economic cost of U.S. $1.4 per person protected per year (U.S. $1.6 unadjusted). Among the children 1–11 months of age, 3,540 life-years of protection were achieved at the end of the first year at a cost of U.S. $25 per infant (U.S. $34 unadjusted). This is based on an estimate of 48.5 life-years gained at the median age of death (six months) of trial infants (48.5 life-years gained × 73 deaths averted = 3,540 life-years of protection). A total of 73 child deaths were prevented by the use of ITNs at an annual cost of U.S. $1,214 per death averted (U.S. $1,635 unadjusted). Applying a discount rate of 3%,³ each death prevented was equivalent to approximately 25 DLYG or a total of 1,825 life-years of protection. This represented an estimated cost of U.S. $49 per DLYG (U.S. $65 unadjusted). The annual net cost per SCV averted is U.S. $38 (U.S. $49 unadjusted) or U.S. $1.9 per ITN (U.S. $2.2 unadjusted).

Sensitivity analysis.

The sensitivity of net annual costs to changes in the discount rate, frequency of net impregnation, cost of insecticide and salary, or wage levels is shown in Table 5. While this analysis includes the community effect, the estimated changes in total cost are the same or within 1–3 percentage points of the estimates generated using the unadjusted figures. Increasing the discount rate applied to both capital and funeral costs from 3% to 6% or from 3% to 10% would increase the cost of the intervention by approximately 4% and 11%, respectively. The cost per child death averted would then increase from U.S. $1,214 to U.S. $1,265 or U.S. $1,354.

Reducing the frequency of bed net re-treatment to once a year also reduced the cost of the intervention by an estimated 29% with a concomitant reduction in the cost per child death averted from U.S. $1,214 to U.S. $857. A decrease in the cost of insecticide of 20% led to a reduction in the cost per child death averted from U.S. $1,214 to U.S. $1,176 (or in the case of a 20% increase in insecticide, U.S. $1,252), representing a mere 3% change in cost. If a 20% lower cost of insecticide is combined with yearly re-treatment, costs are reduced by 30% and the cost per child death averted decreases from U.S. $1,214 to U.S. $839.

The effect of valuing unwaged time at 25% above or 25% below $1.30 per day changed costs by only 0.3%. However, costs were reduced by 17% to $72,988 per annum by a 50% reduction in senior staff salaries. A 50% decrease in funeral costs was associated with a 3% reduction in costs.

In a separate analysis of effects, the discount rate applied to life-years gained varied from 3% to 6% and from 3% to 10%. This led to an increase in the net cost per discounted life-year gained from $49 to $76 and $121, respectively.