Study site and population
Kita district, where SMC was implemented and evaluated in 2014, is located in the western region of Kayes, about 180 kms north of Bamako, the capital city of Mali. Kita was one of the 21 districts in Mali to roll out SMC in 2014 as part of the progressive national scale up of SMC. Kita has one district hospital, 40 community health centres, and 72 community health workers (CHWs). The population was estimated to be 516,649 in 2014 with approximately 77,497 children between 3 and 59 months of age and parasitaemia prevalence for the region in 2013 was 37% [13, 14]. SMC was implemented in all 336 villages/quartiers and while approximately 77,497 children were forecast to receive SMC, an estimated 104,255 children actually received the intervention. This higher number of children than expected is explained by three factors. Firstly, some children older than 5 years are likely to have received the intervention; secondly, Bafoulabe, the district next to Kita did not receive SMC and it is likely that children came from bordering villages to receive SMC; and finally, the study site covered a large agricultural area that, at the time of SMC administration, might have experienced a population swell given it was growing season so there was a need for increased labour.
SMC delivery and administration
SMC delivery refers to the process of obtaining the resources (mainly drugs and incentives) from the central stores to the end user (in this case the child). In the context of this study, incentives capture the financial compensation for supporting for SMC implementation in Kita. SMC administration (sometimes referred to in previous SMC costing papers as distribution) refers to the activities associated with giving the drugs to the child on the day(s) of SMC.
In Kita, the SMC campaign was organized in 2014 using a fixed point distribution method. Specifically, drug distribution was channelled through the district hospital and community health centres, and SMC was delivered at village focal points. Distribution was the responsibility of government employed health staff, primarily nurses and community health workers (CHWs).
Children aged 3–59 months received sulfadoxine-pyrimethamine and amodiaquine (SP + AQ) at monthly intervals over the four months of August, September, October, and November 2014. Before drug distribution started, several meetings were held from May to August 2014 at the national, regional, district and sub-districts levels in Kita in preparation for SMC implementation. Two technical groups composed of 8 staff members from NMCP and implementing partners were created: one to establish, review and update the SMC distribution modules, and develop data collection tools; and the second group to develop a communication plan for SMC implementation. SMC dispensers, who were CHWs and nurses, received a two-day training course on SMC administration using the training modules developed. Normally, the Central Pharmacy (Pharmacie Populaire du Mali, PPM) provides drugs and medical supplies to the Regional Health Directorate in Kayes. The Regional Directorate then provides supplies and drugs to health districts who dispatch them to the district and Community Health Centres. However, because of a delay in preparing for implementation, on this occasion SMC drugs were paid for by the Ministry of Health (MoH) through the NMCP and delivered directly to the district heath centre of Kita by the PPM without passing through the Regional Health Directorate. It then took five days for the district health centre to deliver the drugs and supplies to community health centres.
The SP + AQ used for SMC in Kita were not co-blistered or pre-packed conditional on age group. The drugs, therefore, had to be cut and repacked by the district health team. For this, 10 persons were recruited and worked on the drugs packaging for 25 days. After the drugs were packaged, an allocation plan was developed to distribute the SMC drugs and necessary supplies for implementation to the community health centres and villages heath workers. Information, sensitization and communication messages on SMC were developed and delivered through the two most popular local radio channels for six months (July to December 2014). Additional specific messages were developed and added during the distribution period. In addition, at the community level, mobilization and sensitization activities were carried out.
SMC drug administration was performed by Kita district health staff and comprised 588 drug dispensers (nurses and CHWs) organized into 133 teams of 2–6 health workers each at health centres (n = 37 teams) and village fixed points (n = 96 teams). The first dose of AQ and the single dose of SP was given the first day by health workers. Children were observed for 30 min, and if the child vomited within this time, another dose was administered. The second and third doses of AQ were given to parents to be administrated at home. Further details are provided elsewhere [3].
All labour used in SMC distribution was employed and paid by the MoH, except implementing partner staff who were paid by the non-government organization Save the Children [12]. Supervision and monitoring were performed during the distribution period by teams composed of staff at national level ((NMCP, Maternal and Child Survival Programme, and National Federation for Community Health Centres Association)), regional level and district level. As with all SMC programmes, incentives were given to all of those involved: NMCP personnel, Fédération Nationale des Associations Communautaires, staff from the district hospital and community health centres, CHWs and drivers. Incentives (per-diems and accommodation where required) varied according to categories of the staff and location (local, district level, regional). In Kita incentives were set at daily rates comparable to financial incentives paid for delivering other interventions and research studies conducted in the area previously. The daily incentive for a rural doctor was comparable to their daily salary, and a little less than a CHW could expect to receive for their daily duties. On average those involved in the implementation were paid approximately 36 days of incentives for helping with SMC implementation in 2014.
Costing
Costs are presented from the provider perspective using an ingredients approach whereby the relevant resources were identified and measured at the time of the SMC implementation to estimate the total cost of the intervention [5]. The choice of provider perspective was based on data availability and the intention of the study to inform decision makers how much the intervention would cost for funding proposes. Both financial and economic costs (recurrent and capital) are estimated for the 4 rounds of distribution. Financial costs reflect the actual expenditure required to deliver the intervention such as the cost of drugs and incentives. Economic costs capture the opportunity cost of all resources used to provide SMC, whether or not they incur a financial cost. For example, the time of health personnel involved in SMC delivery represents an economic cost as the staff already received a salary so there was no additional financial commitment, however they could have spent their time on other activities, so we need to capture the opportunity cost [6, 15]. Only 2014 cost data were used in this study and are presented in 2016 Communauté Financière en Afrique (CFA) and US dollars (USD). A conversion rate of CFA 494.17: USD 1 was assumed based on the average of August 2014 inflated to 2016 using US Inflation Calculator [16].
Table 2 presents the cost categories used in the analysis: planning, communication, training, drugs, personnel, equipment and transport. In addition to the total implementation cost and the total cost per round, unit costs are estimated in terms of cost per child fully adherent (i.e. a child that had the first of each of the SMC drugs under observation in all 4 rounds) and cost per child partially covered (i.e. a child that had the first SMC drugs under observation in 1, 2 or 3 rounds). Fully adherent is defined as child who received all 4 rounds and highly adherent as child who received at least 3 rounds.
Effectiveness
An aim of this district-wide SMC delivery was to see how feasible the intervention was at scale outside of trial conditions. The project focused on coverage and collected limited data on health outcomes (specifically prevalence of malaria and anaemia) [13]. Instead, modelling was used to provide an indication of the likely effectiveness of the intervention in Kita. An established dynamical transmission model was used [17]; this model has previously been used for a range of applications including to establish global investment targets for the WHO Global Technical Strategy [18]. Full details, as well as source code and instructions for compilation and running, of the model are available elsewhere [19, 20]. In brief, the model is individual-based and non-spatial, with rainfall driving seasonal patterns of mosquito emergence and subsequent patterns of the entomological inoculation rate, incidence of malaria and prevalence of infection, which then determine the level of onwards transmissibility to the mosquito. The human component of the model incorporates heterogeneity in exposure and both acquired and age-dependent immunity, which affects (i) the probability of developing severe and clinical disease, (ii) the duration of detectable blood-stage infection and (iii) onwards infectivity to the mosquito. This allows the model to capture the relationship between transmission and age-dependent patterns of prevalence and disease across a wide range of settings and to simulate the onwards impact of multiple interventions upon transmission (though the magnitude of this impact within the model is limited when SMC is only provided to under 5 s [21]. The model has also been shown to replicate seasonal patterns of malaria burden, driven by rainfall [22] and the effectiveness of SMC within trials (Cairns et al., pers. commun.), to a high degree of accuracy. Here the model was calibrated to patterns of seasonality, based upon rainfall gauge data between 2002 and 2008 in the region, smoothed by Fourier transformation, to provide typical patterns of rainfall throughout the year. Long-lasting insecticidal nets (LLIN) use within the area was also captured by the model, assuming a linear increase from 0% in 2000 to 56.9% observed in Kayes region in 2013 [3] which had risen to 65.2% in 2015 [2] following the universal coverage campaign in 2014. Usage immediately following the 2014 campaign (which was not measured) was calibrated to achieve 65% usage in the area at the end of 2015 (when the above survey was done), with no further assumed distribution in 2015–2017. The model assumes a constant per-capita rate of attrition in usage in the absence of receiving a new net of 0.2 per year [23]. Simulations were then conducted using 1000 parameter sets representing separate draws from the joint posterior distribution of the underlying transmission model. Baseline vectorial capacity in each simulation was calibrated to 2014 dry-season prevalence according to a triangle distribution with mode of 24.1% and a 95% interval between 20.7% and 27.8%, based upon the mean and binomially-distributed 95% confidence intervals (CIs) observed in this survey, and further accounting for temporal trends in seasonality, LLIN usage and insecticidal coverage.
Four rounds of SMC with SP + AQ were simulated, with a duration of prophylaxis following a Weibull distribution with scale parameter 38.1 and shape parameter 4.3 fitted to published point estimates of the protective efficacy over time from a trial in Burkina Faso [20] (see Fig. 1). This profile provides strong protection for the first 3–4 weeks after a round, declining to 50% protection around 5 weeks, and then subsequently declining rapidly. The model was calibrated to match the per-capita number of doses delivered during the intervention by setting the coverage of the intervention within the model to the overall probability of receiving a dose in any given round according to interviews with the caregivers of children who reported receiving the intervention in 3396 of 4505 (75.3%) potential per-protocol rounds within the survey [13]. Uncertainty in the timing of these doses was then captured by varying the timing of the intervention relative to the seasonal peak by summarising simulations using 1000 draws from a triangle distribution with mode centred around the seasonal peak in transmission, with extremes between a month before and after this optimal scheduling.
Goodness-of-fit was assessed by comparing the model prevalence to prevalence survey collected at the end of 2012 [3] and a month after the final round of SMC [13]. Burden outputs considered were uncomplicated clinical malaria and severe malaria in children under five years and in the wider population, with estimates of lives saved based upon a case-fatality rate of 0.3% [24] per malaria case. The modelling incorporated the universal LLIN campaign within the area in 2014; as a result, in addition to our estimates of 2014 impact, were present both the estimates of the impact of SMC in 2014 specifically, and estimates if SMC at the same coverage as in 2014 had been extended across the 3-year LLIN distribution cycle, in order to provide a measure of the likely effectiveness of SMC if applied continuously in the area.
Incremental cost effectiveness analysis
An incremental cost effectiveness ratio (ICER) was calculated by dividing the annual total economic cost of SMC by the annual modelled estimates of (1) the number of uncomplicated and severe malaria cases averted, (2) the DALYs averted, and (3) the deaths averted. The annual number of episodes averted was an average of a three-year cycle where the effects of LLINs were also included. DALYs are a measure of burden of disease and can be used as a summary measure to determine and compare the cost-effectiveness of different types of interventions across different diseases and settings [25, 26]. They allow for a more informed assessment of cost effectiveness in relation to thresholds and acceptability. There is no gold-standard for when an intervention is deemed cost-effective. Indeed, the use and mis-use of thresholds is much debated [27, 28]. In helping interpret whether SMC was cost effective we present our findings using two very conservative thresholds: (i) US $780.51, Mali’s 2016 Gross Domestic Product (GDP) per capita and (ii) an even more stringent World Bank threshold of US $250 per DALY averted [29, 30].
Ethical considerations
The study protocol was reviewed and approved by the Ethical Committee of the Faculty of Medicine, Pharmacy and Dentistry of the University of Bamako prior to the study. The clinical trial was registered at Clinicaltrials.gov as number NCT02894294.