An ingredients-based approach to costing integrated malaria elimination strategies: an example from Myanmar

Background Mass drug administration (MDA) has received growing interest to accelerate the elimination of multi-drug resistant malaria in the Greater Mekong Subregion. The potential effectiveness of delivering targeted MDA was demonstrated in a recent intervention in Kayin State, Myanmar. Policymakers and funders need to know what resources are required if MDA is to be included in elimination packages beyond existing malaria interventions. Methods We used nancial data from a malaria elimination initiative, conducted in Kayin State, to estimate the programmatic costs of the MDA component using a micro-costing approach. Three activities (community engagement, identication of villages for MDA, and conducting mass treatment in target villages) were evaluated. We then estimate the programmatic costs of implementing targeted MDA to support P. falciparum malaria elimination in Kayin State. A costing tool was developed to aid future analyses. Results The cost of delivering MDA within an integrated malaria elimination initiative in Eastern Kayin State was approximately US$ 910 000. The cost per person reached for MDA was 2·5. Conclusion This cost analysis can assist policy makers in determining the resources required to clear malaria parasite reservoirs. The analysis demonstrated the value of using nancial data from research activities to predict programmatic implementation costs of MDA in different numbers of target villages.

This cost analysis will inform policy makers' contemplating the selection of interventions to be included in their national malaria control programs, aimed at meeting their elimination timelines, on the expected cost of malaria interventions packages that include MDA.
Countries across the malaria endemic world are aiming to eliminate malaria and committed to identifying approaches aimed at interrupting its transmission [1]. Although substantial progress has been made through the scaling up of existing malaria interventions, the gains achieved are fragile and unevenly distributed among regions and countries. The development of artemisinin resistance in South-East Asia, followed by partner drug resistance [2], led to a call by all country governments in the Greater Mekong Subregion (GMS) to hasten Plasmodium falciparum malaria elimination [3].
In 2010, the World Health Organization (WHO) proposed mass drug administration (MDA) as a strategy to accelerate the elimination of multi-drug resistant P. falciparum malaria [4]. This strategy aims to treat every individual in a community with three rounds of a full dose of antimalarial drug, regardless of whether they have malaria symptoms. Many eld trials and programmatic implementations of MDA have been carried out over the past century with varying degrees of success [5]. However, a trial MDA conducted in Zanzibar showed no impact on malaria incidence [6].
A series of targeted MDA projects have recently been piloted in countries in the GMS [7][8][9][10] and Africa [11,12]. These trials have demonstrated that MDA is feasible and well-accepted by communities, with high levels of community participation. MDA using a therapeutic dose of an effective antimalarial medication can clear the malaria parasite reservoir, including asymptomatic infections that would otherwise not be treated [13]. This approach can rapidly reduce malaria parasite prevalence if highly e cacious antimalarial drugs are used [14].
In addition to the effectiveness of interventions, policymakers must be able to evaluate the costs of malaria elimination, to determine which intervention or package of interventions should be invested in, given the constraints of budget limitations. Even when the epidemiological impact of intervention packages is well characterised, e cient resource-allocation decisions can only be made when all resources consumed are explicitly valued and visible. Cost analyses have been conducted for most malaria control activities [15], but to date, few costing exercises of MDA have been performed [16][17][18].
The Malaria Elimination Task Force (METF), established by the Shoklo Malaria Research Unit, which is based on the Thai-Myanmar border, delivered an integrated malaria elimination strategy that layered targeted MDA over a series of malaria control and research activities in the region. Here, we used nancial data from METF targeted MDAs in 61 selected villages in Kayin State, Eastern Myanmar as a basis for a costing exercise to analyse the costs of delivering MDA in targeted Myanmar villages and calculate the scalable cost of delivering targeted MDA to support P. falciparum malaria elimination.

Study area
Kayin State lies on Myanmar's international border, with Thailand to the East. It is a mountainous region, with the rocky Dawna mountain range running the length of eastern Kayin State. The climate is hot and humid, with average maximum temperatures of between 29 and 37°C and average annual rainfall of approximately 5000 mm [19]. The METF covers a population of 365 000 people. The residents of Kayin State traditionally rely on agriculture for their livelihoods; major crops include rice, rubber, sugarcane, coffee, and seasonal fruit and vegetables.
Kayin State has experienced decades of armed con ict between various Kayin militant groups and the national government [20]. Therefore, it is a politically sensitive area, and government accessibility to the region is limited. Consequently, basic infrastructure, such as roads, electricity, schools, and health care facilities, are under-developed.

Data collection
A series of consultations were held with METF, and the importance of the costing exercise was explained. Permission to access nancial reports relating to the MDA conducted by METF (2015)  First, we estimated the cost of implementing MDA during 2015 based on 2015 nancial data; these costs were then extrapolated to a total of 61 villages that received targeted MDA.
All resource ingredients required to perform the three MDA implementation activities were identi ed, measured, and valued by reviewing nancial reports and conducting interviews with key staff. Then, each cost was assigned to a primary resource cost centre. Costs in resource centres were re-classi ed into relevant activity cost centres.
Cost model ingredients 1. Staff costs. These comprised basic salaries for both national and international staff, plus their allowances, which included bene ts and overtime. These staff costs were shared resources, so the allocation of these shared resources was based on the proportion of the time these staff contributed to various services.
2. Travel costs. These included all transportation costs, including bus, taxi, and boat fares; toll fees; petroleum consumed during the project; motorbikes and their maintenance; and other travel-related expenses, such as accommodation and per diems. The cost of transportation of medical and nonmedical products and travel expenses for monitoring and training were also included in the travel cost centre.
3. Consumables costs. These included the cost of media, pamphlets, and other health education materials for community engagement activities; uPCR sample preparation and analysis; antimalarial drugs; and medication to treat any adverse effects. The costs of consumables used for uPCR sample collection and analysis were estimated based on the mean number of uPCR samples and the unit cost of sample collection.

Malaria in Kayin State
Malaria transmission is low, seasonal, and spatially heterogeneous in Kayin State [21]. Five decades ago, malaria was a signi cant problem, but following rigorous treatment and implementation of malaria posts by the METF, annual incidence of malaria has gradually decreased in the region. Malaria posts are operated by trained members of the community. Malaria infections, con rmed at malaria posts by a rapid diagnostic test (RDT), comprise approximately 12% of all febrile illness [13]. In Myanmar, malaria infection is frequently asymptomatic [22]. A recent survey using uPCR showed a malaria prevalence of 21%, with P. falciparum and P. vivax comprising 3% and 15% of infections, respectively [13]. MDA was conducted in 61 selected villages in ve phases over two years, with more than 80% participation of village members. As a result, the incidence of P. falciparum malaria was reduced by 92%. Following targeted MDA in the region, 71% of METF villages reported no P. falciparum malaria [23].

Results
The METF implemented an integrated malaria elimination initiative in 1226 villages in Kayin State from May 2014 to December 2019. This included targeted MDA in 61 villages in areas with high levels of submicroscopic P. falciparum malaria, to contain further spread of multi-drug resistant malaria. The total cost of the targeted MDA was estimated using a micro-costing approach. Figure 2 shows the detailed breakdown of total cost of targeted MDA in Kayin State, which involved three activities: i) community engagement (CE), ii) identi cation of target villages by uPCR, and iii) targeted mass treatment in selected villages. Villages were selected for MDA if the measured uPCR prevalence for P. falciparum was more than 20%. CE is a crucial activity required both before and during MDA. The total cost of CE activity for 61 villages (5% of 1226 villages) was US$ 76 330. The average cost per person for providing CE was US$ 0·20.
Villages for MDA were identi ed based on their malaria prevalence relative to other villages in Kayin State. A malaria prevalence survey was conducted, using a uPCR assay, to measure the true prevalence of malaria in 272 randomly selected villages. Consumables comprised the largest proportion of the total cost of identifying MDA villages, accounting for 80% of all costs. This was because of the cost of uPCR analysis, which was approximately US$ 25 per test. The high cost of the tests was due to the need for expensive equipment, reagents and consumables for high-volume PCR to obtain the desired sensitivity [24]. The total cost of the prevalence survey for 272 villages was US$ 541 042, with the average cost per village of US$ 1 989.
The cost of providing three rounds of three-day antimalarial mass treatment in 61 villages was US$ 291 759. The staff costs were the largest contributor to the total cost of mass treatment (46%), followed by consumables (21%). This was because the METF provided supervised treatment for every dose of antimalarial treatment. Therefore, staff needed to stay for at least four days in a village in each treatment round to cover any latecomers and to watch for any side effects of the treatment.. The average cost per village for providing a full course (three-day treatment) of antimalarial drugs for three consecutive months was US$ 4 455. The cost per person reach for providing three rounds of antimalarial mass treatment was US$ 0·8.
The total cost of an integrated MDA initiative, including detection of hotspots and mass treatment in 61 villages (5% of villages), would be approximately US$ 910 000 over 2 years. The cost per capita for three rounds of targeted MDA was estimated to be US$ 2·5. The intervention was an integrated strategy for the prevention, early detection, and treatment of clinical malaria in all villages, combined with MDA in targeted villages; therefore, the cost per capita quoted is derived from the total cost of the integrated MDA package of interventions but excluding early detection and case management divided by the total population of the area where it was delivered (including MDA villages and non-MDA villages  * Cost per village is estimated by dividing the total cost of targeted MDA by the total number of villages in the four townships (1226 villages). These targeted MDA costs will be shared among all villages in the region because targeted MDA is provided in addition to other malaria interventions, so the total cost is distributed among all villages in the region.
** Cost per person reached is calculated by the total cost divided by the total population in that area (365 000).
MDA, mass drug administration; uPCR, ultrasensitive polymerase chain reaction Comparing the programmatic cost of targeted MDA using different infection diagnostic approaches We estimated the programmatic cost of targeted MDA using different molecular diagnosis methods, such as uPCR, RNA testing and ELISA testing. The unit cost of molecular tests varies, so the programmatic cost of targeted MDA in Kayin State will differ if we use techniques other than the more expensive uPCR method. The percentage of the prevalence survey was kept the same as the MEFT project (22% of all villages); only the percentage of villages where CE and mass treatment were provided was varied. Table 2 shows the comparison of the programmatic cost of targeted MDA using three different molecular diagnosis methods. The unit costs set for the RNA and ELISA tests per sample were US$ 20 and US$ 5, respectively (F. Nosten, personal communication). The total programmatic cost would be decreased by approximately 37% if METF used the ELISA method to identify villages to target for MDA. Similarly, the cost of targeted MDA would be reduced by 9% if uPCR was switched to RNA testing.

Discussion
The malaria map is shrinking in both Kayin State [23] and the whole of Myanmar [25]. Myanmar has set a goal to interrupt transmission and eliminate P. falciparum malaria from the entire country by 2025 [26]. Malaria elimination requires a substantial level of investment, especially for detecting and responding to small numbers of remaining malaria cases [27]. Which malaria intervention packages to use and the resources needed to eliminate this disease nationally and sub-nationally is a challenging question for a developing country that largely relies on external funding to achieve this goal.
Malaria case management and intensive vector control are core interventions in malaria control, but achieving elimination goals is likely to require other population wide measures, particularly in the context of increasing multi-drug resistant malaria. A population-wide, medicine-based strategy, such as MDA, can accelerate the reduction in transmission [28]. Several targeted MDA projects have been conducted in the GMS, including in Kayin State [13], and have been shown to reduce the incidence and prevalence of malaria [14,29]. The implementation of targeted MDA, however, requires a signi cant investment in terms of resources and time to mobilise the targeted villages. The higher cost relative to standard approaches to malaria control and elimination was considered an acceptable given the risk of multi-drug resistance and the measures deemed necessary to address this.
The cost of identifying target villages was the largest contributor in this cost analysis. When the prevalence of malaria is declining, its management is focused on subclinical infection. In low transmission settings, asymptomatic infection dynamics should be adequately identi ed using highly sensitive diagnostic methods. Molecular techniques are more sensitive than other diagnostic methods.
The detection limit of PCR is approximately 22 parasites per mL. METF used high-volume uPCR to identify villages to target for MDA. An alternative assay to uPCR, such as RNA or ELISA, would reduce the cost of the prevalence survey while maintaining su cient sensitivity [30].
Studies on the cost of deploying malaria MDA are limited, with one study examining the delivery costs of MDA in two island settings and an emergency setting [16].  [32] revealed that the spread of multi-drug resistant P. falciparum malaria in GMS countries was accelerating, highlighting the urgent need to adopt an effective strategy to eliminate malaria. Recently published studies [9,13,29,33] suggest that targeted/focal MDA with a high degree of community participation can rapidly reduce malaria infections to zero when used in conjunction with intensive vector control and standard case management. There was no signi cant increase in any of the genetic markers for resistance after MDA [29]. The components necessary for a successful integrated malaria elimination strategy are predicted to be highly dependent on the setting [34], and Myanmar is expected to require MDA or other more intensive interventions.
Policymakers must therefore consider a trade-off between investing in rapid elimination strategies that might stave off the threat of resurging drug resistant malaria, or slower (and cheaper) elimination strategies. GMS countries need to buy time to halt the spread of multi-drug resistant malaria while new antimalarial are developed. This analysis provides the added cost of targeted MDA to rapidly eliminate malaria on top of existing malaria surveillance and control costs. We estimated the programmatic cost of targeted MDA in Kayin State using nancial data from the METF implementation, and developed a malaria mass intervention costing tool to support policy decisions towards P. falciparum malaria elimination in other settings. The key features of this costing tool are its ease of use, the exibility to explore different targeting strategies, and the cost predictions for any single malaria intervention or package of interventions.
The costing tool was designed based on the targeted MDA initiative in Myanmar; nonetheless, the tool can also be used to predict programmatic costs in other GMS countries, by adjusting the unit costs of resources and the proportion of villages undergoing interventions. For example, the tool could be deployed in other regions in Myanmar, such as Chin state, where P. falciparum malaria incidence and mortality is high in comparison with other regions [25]. The added bene t, beyond addressing the multidrug resistance issue, would be the additional lives saved by accelerating elimination to a date earlier than 2030. Table 1 allows the exploration of the cost implications of delivering this programme in a higher prevalence setting, which would probably mean a higher number of villages ful lling the conditions for being designated a hotspot. Table 2 allows the exploration of the trade-off between costs and the use of cheaper alternative screening options.
There are several limitations to this cost analysis. Different teams providing MDA include staff members of differing levels of seniority, so there may be some variations in estimating staff costs. However, this variation between MDA teams is negligible. As the percentage of villages provided with MDA increases, programme managers have the option of training more staff or using their existing team for an extended period. In our cost estimation, we used the same team to provide MDA. Cost variations may result if a programme manager makes the trade-off of recruiting more staff to complete MDA in less time. The targeted MDA initiative in this cost analysis was operated by a Thailand-based organisation. Therefore, staff compensation and travel costs to access the villages were based on staff travelling from the Thai side, so some cost variations will be seen if villages were accessed from the Myanmar side. However, this variation would be minimal, since most of the costs were incurred within the country.
Our costing model can predict the costs of a particular malaria elimination package design, but cannot make any assurances on the likelihood of success of such a package in achieving elimination elsewhere. It is designed to be used in concert with detailed knowledge of the target area and/or with mathematical models that can simulate the impact of various strategy designs on the prevalence and incidence of malaria [34].

Conclusions
This cost analysis quanti es the costs of accelerating P. falciparum malaria elimination. Such cost analysis makes a useful contribution to determine the level of resources required to clear the residual malaria parasite reservoir. It also provides a framework for projecting the cost of similar programmes in settings with different epidemiology and/or the exploration of the cost of alternative designs. The study demonstrated the use of nancial data from MDA research to project the programmatic implementation cost of MDA with a different number of targeted villages. None of the authors declare any competing interests.

Funding
The Wellcome Trust of Great Britain and the Bill and Melinda Gates Foundation supported this cost analysis. However, the funding organisations had no role in the design of the study, analysis, interpretation of data and in the writing of the manuscript.
Authors' contributions SSK, TD, WP, YL, LJW, and FN were involved in designing the study. SSK and TD performed the nancial data collection. SSK and WP analysed the costs of targeted MDA. SSK and OC developed the costing tool. SSK, WP, RJM, and LJW wrote the original draft. All authors have read and approved the nal manuscript.