Improving the cost-effectiveness of IRS with climate informed health surveillance systems

Quantifying the burden of malaria remains a major challenge as many infections may be asymptomatic, inadequate diagnosis and reporting and the fact that a majority of febrile cases do not reach the formal health system make estimation imprecise. Recent estimates espouse the figures of 300–600 million infections and 1–3 million deaths per year [1]. Less contentious is the fact that Africa bears the brunt of the malaria burden, with estimates suggesting greater than 60% of the worlds clinical cases and more than 90% of the worlds malaria deaths [2].

Africa's population is estimated to pass 1 billion people by 2010 [3]. The current goal of the Global Malaria Programme and the Roll Back Malaria Partnership is to "halve the malaria burden by 2010" by focussing on treatment, prevention and epidemic response[4]. Africa is not yet on track to achieve this goal [5].

It is estimated that 615 million Africans live in endemic regions, where those most at risk are children under 5 and pregnant women, non-pregnant women and adult males are protected, to varying degrees, by acquired immunity. There are a further estimated 125 million people at risk from epidemic malaria in sub Saharan Africa [6]. By definition in epidemic prone areas, transmission intensity (usually indicated by the entomological inoculation rate, or EIR) is insufficient for the population to develop acquired immunity; therefore this entire population of 125 million people is at risk of severe morbidity and mortality from malaria [7]. People living in malaria endemic regions acquire immunity to malaria through natural exposure to malaria parasites. This naturally acquired immunity is protective against parasites and clinical disease but it only results after continued exposure from multiple infections over time. The transmission intensity influences the course of development of clinical and parasitic immunity. In areas of intense transmission, young children bear the burden of malaria, but as they grow older they build up acquired immunity. In areas of low endemicity both children and adults suffer disease and high parasitaemia because exposure is less [8].

Pregnant women are more susceptible than non-pregnant women to malaria because of a combination of immunological and hormonal changes associated with pregnancy, and with the unique ability of certain variants of infected erythrocyts to sequester in the placenta [9]. Proportionally more pregnant women are at risk in epidemic settings because in high transmission areas, multi-gravidae are partially protected from adverse effects and only primigravidae are indisputably at greater risk of infection than non-pregnant women [10]. The clinical features of malaria infection during pregnancy also vary by epidemiologic setting. Severe disease (such as cerebral malaria and respiratory distress syndrome) is a more common characteristic in low and unstable malaria transmission areas, whereas pregnant women in high transmission areas rarely experience cerebral malaria but instead have more frequent exposure to malaria leading to severe anaemia [11]. In all areas, malaria in pregnancy is responsible for low birth weight (LBW) deliveries [12].

In epidemic prone areas malaria control interventions aim to protect the entire population for mortality, whereas in endemic areas, interventions only afford protection to vulnerable sub-population groups such as young children and pregnant women from mortality and the general population from morbidity.

Approximately 20% of sub-Saharan Africa's population, under five's and pregnant women, living in endemic malaria areas are subject to malaria as a major life threatening disease (Africa Malaria Report: WHO-AFRO 2006). This is roughly equivalent in numbers to the estimated 125 million people at risk of severe disease and death from malaria epidemics [7]. However, although the population at risk is roughly equivalent in endemic and non endemic areas, in the former the risk is constant from one year to the next whereas in epidemic prone areas the risk may only be high in certain years. Hence the temporal element of risk needs to be considered. Thus endemic and epidemic malaria require different approaches to control and prevention. Endemic malaria requires ongoing measures whereas successful control of epidemic malaria relies on measures being applied in the right place at the right time, this is even more important when resources are scarce [13].

The global malaria strategy calls for sustained commitment to address malaria epidemics [14], but there has been less focus on epidemics relative to other strategic areas of intervention such as scaling up of ITN use.

Recently there has been a renewed emphasis on malaria vector control with indoor residual spraying (IRS), especially under the US Government's Presidential Malaria Initiative (PMI). PMI is supporting enhanced malaria control in fifteen African countries, many of which are or will be using IRS as part of a package of malaria control. A growing body of evidence shows that IRS and ITNs (and more recently LLINs) are highly cost-effective malaria control interventions [15, 16]. More recently focus has shifted towards examining the relative cost-effectiveness of these two interventions in comparison to each other [17]. However, there is still a dearth of evidence on the cost-effectiveness of malaria control interventions in epidemic prone areas [6] and economic evaluations of interventions in epidemic or seasonal transmission areas commonly fail to account for key differences in assessing malaria control interventions in epidemic as opposed to endemic settings. A previous paper has shown how the effectiveness of IRS (in terms of cases prevented) varies according to the magnitude of transmission each year, as well as the timing of IRS in relation to the transmission season [18].

This paper examines how the cost-effectiveness of IRS varies depending on the severity of transmission and level of coverage. It also investigates how efficiency could be improved by incorporating climate information into the decision making process of malaria control programmes as part of an integrated Malaria Early Warning and Response System (MEWS) [19]. Put another way this study aims to show how more health impact can be obtained through use of the same or similar resources if climate information was used to inform programme decisions.

Firstly, available evidence on the cost and cost-effectiveness of IRS in sub-Saharan Africa is reviewed to establish what is already known about the key drivers of cost and cost-effectiveness of IRS and highlight limitations of existing studies. Secondly, novel cost data on an operational IRS programme from Zimbabwe is presented. Thirdly, investigation is made into how cost-effectiveness and marginal cost-effectiveness of IRS varies depending on severity of transmission and simulated level of IRS programme coverage. Finally, a discussion focuses on how the cost-effectiveness of IRS and other malaria control interventions can be improved through the incorporation of climate information into malaria control programming decisions.

The economic costs of the IRS programme in Hwange District, of Zimbabwe were captured for a twelve month period aligned with the 1998/99 malaria transmission season using standard economic costing methodology [29]. Full details of the study area, IRS programme, costing methods and results are given in Additional file 2.

Using the same model presented in a previous paper [30], the unit cost data was combined with estimates of the effectiveness of IRS over a six year period to examine how the average and marginal cost-effectiveness of a programme varies depending on (i) the level of spray coverage achieved and (ii) the severity of transmission. The model was then used to simulate the number of cases that would occur given five alternative levels of spray coverage, effective from January each year. The coverage levels examined were 0%, 24%, 50%, 75% and 100%.

24% coverage is the actual level of coverage achieved overall in Hwange, District this is because the IRS programme targeted administrative units smaller than districts, achieving high levels of coverage (80–95%) in these sub-units but an average coverage level of 24% throughout the District. Hence what our coverage levels refer to is the percentage of the district targeted for IRS with the implicit assumption that the level of coverage achieved is sufficiently high (> 85%) for the IRS to achieve impact.

For each year and level of spray coverage the total cost of the spray programme was calculated, the number of cases prevented and the average cost per case prevented compared to the no coverage (0%) scenario. Next the marginal cost per case prevented of additional coverage levels compared to the baseline (24%) scenario was calculated. Full details and results of simulations on cases and cases prevented can be found in [30].

In order to examine the impact of severity of transmission, each year was categorized in our dataset as low, medium, high or epidemic depending on the number of simulated cases with no intervention) (0–50,000; 50,001–100,000; 100,001–150,000 and 150,001–200,000 respectively) as shown in columns 1 – 3 of Table 2.

Next the potential efficiency implications of a MEWS which would allow programme managers to choose the level of IRS programme coverage in advance based on information about the likely severity of the forthcoming malaria transmission season was examined. It was assumed that the forecast would influence decision making regarding IRS programme coverage, as shown in columns 2 and 4 of Table 2, as part of a MEWS. This was labelled the "implied level of coverage", i.e. the level of coverage that the MEWS indicated would be appropriate given predicted transmission potential each year. Note that there are no years where 0% coverage would be recommended and alternatives are being compared to the baseline level of coverage (24%) hence the key variable of interest here is marginal (as opposed to average) cost-effectiveness.

For this type of system to be of practical use (and reap the hypothetical efficiency and health benefit gains presented in this paper) there are two interrelated key factors which need to be considered. The first is the accuracy (ability to predict malaria season severity) of the information and the second is the timeliness of the information, especially in relation to health services capacity to respond. The present analysis assumes that the MEWS is accurate (does not predict false epidemics or miss real epidemics) and that the information is available sufficiently early for programme managers to use it as a basis to make IRS programme coverage decisions. WHO has, since 2002, promoted a framework for malaria early warning in which four components are operating [31]. The components are: 1) vulnerability monitoring; 2) seasonal climate forecasting; 3) environmental monitoring and 4) sentinel case surveillance; each informing elements of control planning, preparedness and response.

Vulnerability monitoring of factors such as HIV AIDS incidence, drug resistance and food insecurity helps keep track of factors which increase the severity of disease outcome should an epidemic occur. Seasonal climate forecasts provide advance warning of the potential for malaria epidemics to occur. Environmental monitoring, e.g. of rainfall can be used to monitor the accuracy seasonal climate forecasts, yet it still offers some lead time to inform programme decisions. Sentinel case surveillance data is of paramount importance in early detection of epidemics but may come too late to trigger a full preventive response. The WHO framework to malaria early warning has been adopted in a number of epidemic prone countries in Southern Africa [32]. In Botswana an analysis of national confirmed malaria incidence over 20 years revealed that year to year variation in rainfall could account for the year to year variations in incidence once long term vulnerability changes had been taken into account [33].

Furthermore seasonal climate forecasts issued in November, four to five months prior to the peak malaria season, were found to accurately predict five low malaria transmission years and four out of five high transmission years in Botswana [34].

Of the southern African countries (Botswana, Madagascar, Mozambique, Namibia, South Africa, Swaziland and Zimbabwe) involved in these efforts, Botswana has perhaps the most advanced and well resourced system. There is evidence that in a recent wet year following several dry years (2005–06 malaria season), cases were ten times lower in Botswana than in a previous year with similar conditions (1996–97). In Zimbabwe, where political and economic constraints have hampered efforts to implement MEWS and malaria control more generally, cases were down by half. MEWS may have contributed to these successes but changes in other factors (e.g. availability of drugs and increased ITN use) will also have had an effect [13].

For practical reasons, programme managers are not likely to be interested in such detailed analysis of average and marginal cost per case prevented. However, on a broad operational level the results clearly show that compared to spraying the same level of coverage each year, programme managers could ensure a more effective and efficient use of resources by using information on likely transmission severity to inform IRS programme coverage. Accurate advance knowledge of the severity of the transmission season could help managers make coverage decisions which optimize resource use and exploit efficiency gains afforded by the greater effectiveness of IRS in high transmission seasons. They can also avoid using resources unnecessarily in a low transmission year.

The efficiency gains examined in this paper could only be realized if a fully integrated MEWS was in place within a health system with sufficient flexibility to modify control plans in response to forecasts. This is the case in Botswana where it is likely that some efficiency gains and greater health benefits are being realized, if not actually quantified. Increasing the use of MEWS and flexible response systems further, especially in relation to IRS programme planning and decision making, has the potential to ensure malaria control programme decisions are more efficient and critically have greater health impact.

Increased resources are available for both malaria control and IRS in particular. However, resources are still scarce in relation to the burden of malaria. There is thus a need to ensure resources are used as efficiently as possible to generate maximum health benefit. We have the scientific understanding to predict malaria transmission using climate and other indicators [34]. There is also increasing evidence and examples of the use of climate based early warning systems being used as a decision making tool in malaria control programmes [13, 32]. More countries and programmes should be supported to use the best available evidence and scientific know how to integrate climate informed MEWS into decision making within malaria control programmes. The current focus on climate change adaptation and improved management of climate sensitive risk in national development sectors offers opportunities for Ministries of Health to request greater engagement with their national counterparts in climate and weather services. There are currently only a few countries pioneering this approach, however national and international development donors are seeking to invest strongly in improving climate risk management in coming years [13, 35]