Rapid reduction of malaria transmission following the introduction of indoor residual spraying in previously unsprayed districts: an observational analysis of Mopti Region, Mali, in 2017

A set of retrospective, observational (ecological), time-series analyses were performed using monthly incidence rates of rapid diagnostic test (RDT)-confirmed malaria cases reported in the District Health Information System 2 (DHIS2) from January 2016 until February 2018. Monthly case numbers were aggregated at the health facility level for clinics in the districts of Ségou and Mopti Regions. Baseline population estimates, based on 2012 census results, for each health facility catchment area (HFCA) were obtained from the Ministère de la Santé de la République du Mali (Ministry of Health), Direction Régionale de la Santé. IRS programme implementation data from the PMI AIRS project [13, 14] was used to stratify analyses by HFCA spray status, either IRS or non-IRS.

The districts included in this report are shown in Fig. 1 and Table 1. Ségou and Mopti contain 15 administrative districts (cercles) covering roughly 145,000 km², with an estimated 2017 population of around 5.8 million—representing about 30% of the total population of Mali [15]. The malaria burden is relatively high in the two regions (2015 Malaria Indicator Survey results estimated an u5 prevalence of 22% in Ségou and 54% in Mopti) [4], and transmission is seasonal—typically highest from August through November and coinciding with the rainy season, when An. gambiae sensu stricto (s.s.), Anopheles coluzzii, and Anopheles arabiensis are all present [9]. The most recent LLIN mass distribution campaigns had been completed in 2015 in Ségou [12] and 2017 in Mopti [6], and both regions have benefitted from continuous routine LLIN distribution through antenatal care and childhood immunization visits since at least 2013 [6]. During the study period, consistently high LLIN ownership (90–95% of households owned at least one LLIN) and use (83–86% of children u5 used an LLIN last night) were reported [16]. Local vector populations were also reported to be highly resistant to pyrethroids and dichlorodiphenyltrichloroethane but susceptible to carbamates and organophosphates throughout the study period: mortality of local vector populations exposed to pyrethroids in standard diagnostic WHO tube tests ranged from 18 to 55% against permethrin and from 45 to 77% against deltamethrin, while CDC bottle bioassays indicated that resistance remained against both chemicals at up to 10× the diagnostic dose [9, 17]. Vector populations remained 100% susceptible to PM throughout the study period [9]. Both regions have also participated in annual SMC campaigns since 2015 [18] and benefited from access to intermittent preventive treatment in pregnant women, free of cost to all residents. Across all study districts, the SMC campaigns targeted all resident children aged 3 to 59 months with a monthly course of sulfadoxine-pyrimethamine plus amodiaquine for 4 months of the rainy season, beginning each August. District-level coverages obtained by each SMC campaign were comparable during the study period and are presented in Table 1. Additionally, the districts of Ségou and Mopti have been shown to be similar to one another with respect to population density, rainfall patterns, malaria transmission seasonality, and population-adjusted Plasmodium prevalence rates [19], further strengthening the case to use the non-IRS districts and HFCAs as time- and climate-matched comparator districts for these observational studies.

IRS was implemented by the NMCP with support of the PMI AIRS project in Barouéli District of Ségou Region in 2016, and in the districts of Mopti, Bandiagara, Bankass, and Djenné in Mopti Region in 2017 [13, 14]. PM (Actellic^®300CS) was used in both campaigns. The 2016 campaign lasted from July 9 to August 12 and sprayed a total of 71,333 structures out of 73,528 eligible structures encountered (a 97% acceptance rate; Table 1) in Barouéli District, using a blanket spraying approach that aimed to spray every eligible house in the district. The 2017 campaign in the four districts of Mopti Region lasted from July 24 until August 27, and utilized a slightly different approach; only a certain percentage of HFCAs that were easily accessible, given safety and logistical considerations, were targeted for spraying in each district. The percentage of HFCAs that were targeted varied by district: 74% in Mopti, 70% in Bandiagara, 23% in Bankass, and 41% in Djenné. Within each selected catchment area, all eligible houses were targeted. Overall, the 2017 campaign sprayed 239,350 houses across the four districts, with an average acceptance rate of 96% (Table 1). In both campaigns, eligible houses were structures with one or more rooms used for sleeping and interior wall surfaces that were sprayable. Entomological monitoring from 2017 indicated that Actellic had a residual efficacy of 3–4 months, depending on wall type, as measured with standard WHO cone bioassays [9].

From January 2016 to February 2018, the DHIS2 recorded a total of 11,280 monthly malaria reports from 314 public community health facilities (Centre de santé communautaires, CSCom) across Ségou and Mopti. Only surveillance of cases from public health facilities, where national malaria case management guidelines are to test all suspected cases with an RDT, are included in the present analysis. A total of 497,498 RDT-confirmed cases of Plasmodium falciparum malaria were reported in u5 children, out of 700,767 suspected cases (patients with a fever) tested, for an overall u5 test positivity rate of 71%. Almost two-thirds (64%) of all confirmed cases were reported during the months August to January, corresponding to seasonal rainfall. During this period, 93% of all suspected cases reporting to the health system in Ségou and Mopti received a diagnosis by RDT and/or microscopy, and 94% of confirmed cases were treated with artemisinin-based combination therapy. No substantial variation in these measures were observed across districts (82–99% tested; 86–100% treated) or months (86–97% tested; 87–100% treated). District reporting rates were greater than 98% for both IRS and non-IRS districts, though district-months in which no data were reported were excluded from analysis. Additionally, since 2015, MEASURE Evaluation has actively assisted the Ministry of Health with DHIS2 data quality assurance activities at all levels of the system, including tracking commodity stockouts, which were minor during the period analysed here. Monthly malaria incidence rates were calculated by dividing the appropriate number of RDT + test results recorded in the DHIS2 database by the corresponding HFCA population denominators [15].

For this analysis of monthly trends in malaria incidence in the u5 population, a quasi-experimental time series approach was used. The regions of Ségou and Mopti were analysed separately; for each analysis, health facilities were stratified by IRS status (IRS and non-IRS) based on data compiled from PMI AIRS End of Spray Reports and national Malaria Operational Plans (http://​www.​pmi.​gov/​where-we-work/​mali). To estimate the impact of IRS, the cumulative incidence of RDT-positive malaria cases observed during the 6 months following each IRS campaign (which also corresponds to the high transmission season) was calculated for health facilities that received IRS and compared to the cumulative malaria incidence observed in health facilities from the same region that did not receive IRS. The total number of cases averted during each post-IRS window was calculated by multiplying the observed cumulative incidence reduction by the corresponding strata population estimates.

The same difference-in-differences approach was used to quantify the impact of suspending IRS operations after the 2016 spray campaign in the Barouéli District of Ségou Region. The average change in incidence from the 2016 to 2017 transmission seasons observed at the 19 health facilities in Barouéli was compared to the average change observed at the health facilities in the rest of the districts of Ségou Region (121 health facilities), which had not received IRS in either year.

Datasets were organized, cleaned, transformed, and joined using Microsoft Excel 2013 with Power Query v2.41 (Microsoft Corporation; Redmond, WA) and Tableau v10.0 (Tableau Software Incorporated; Seattle, WA). Descriptive statistics were calculated using Excel 2013 and Tableau v10.0. Confidence intervals were calculated and Student’s t-tests comparing averaged incidence rates of malaria across districts by IRS status were conducted using STATA SE 14.2 (StataCorp; College Station, TX). Geographical Information System analysis and mapping were performed using QGIS v2.16 [20]. Shapefiles were downloaded from the GADM database of Global Administrative Areas [21] in December 2017.

The average monthly incidence of RDT-confirmed cases in the u5 population, by HFCA IRS status, are presented for Ségou Region (2016 IRS campaign) and Mopti Region (2017 IRS campaign) in Fig. 2. Considering the cumulative case incidence from the six high-transmission months following the IRS campaign (September to February), the effect of IRS in Ségou in 2016 was comparatively modest but consistent: the observed case incidence at IRS health facilities was reduced by 16% in comparison to non-IRS facilities (Fig. 2a), which translates into around 5200 u5 cases averted in Barouéli that year. The observed impact of the 2017 campaign in Mopti was almost double that observed in Ségou the year prior, with case incidence rates reduced by 31% at IRS health facilities in comparison to non-IRS health facilities (Fig. 2b)—an estimated 17,500 fewer cases than expected in the u5 population.

While the reductions in incidence observed at health facilities within the IRS areas of Ségou in 2016 and Mopti in 2017 correspond both temporally and geographically with the PMI AIRS spray operations, the strength of the association can be further evaluated with a difference-in-differences approach—taking advantage of the fact that the intervention status of the IRS health facilities changed from year to year in both regions, as shown in Fig. 3.

Across all comparator health facilities where there was no IRS in either year (n = 238), peak malaria case incidence rates (September to February) fell by an average of 22% (CI₉₅ 18–30%) from 2016 to 2017 (Fig. 3). In those HFCAs of Mopti where IRS was introduced in 2017 (n = 51), peak incidence fell by an average of 42% (CI₉₅ 31–63%), a significantly greater decrease (p = 0.040) more than double what was seen in the comparator HFs (Fig. 3). The opposite effect was observed in Ségou Region, where peak incidence at those HFs where IRS was withdrawn after the 2016 campaign (n = 19) increased by an average of 106% (CI₉₅ 63–150%) from year to year, also a highly significant difference-in-differences compared to the standard no-IRS HFs (p < 0.0001) (Fig. 3).