Sub-national tailoring of seasonal malaria chemoprevention in Mali based on malaria surveillance and rainfall data

Abstract

Background: In malaria endemic countries, seasonal malaria chemoprevention (SMC) interventions are performed during the high malaria transmission in accordance with epidemiological surveillance data. In this study we propose a predictive approach for tailoring the timing and number of cycles of SMC in all health districts of Mali based on sub-national epidemiological surveillance and rainfall data. Our primary objective was to select the best of two approaches for predicting the onset of the high transmission season at the operational scale. Our secondary objective was to evaluate the number of malaria cases, hospitalisations and deaths in children under 5 years of age that would be prevented annually and the additional cost that would be incurred using the best approach.

Methods: For each of the 75 health districts of Mali over the study period (2014-2019), we determined (1) the onset of the rainy season period based on weekly rainfall data; (ii) the onset and duration of the high transmission season using change point analysis of weekly incidence data; and (iii) the lag between the onset of the rainy season and the onset of the high transmission. Two approaches for predicting the onset of the high transmission season in 2019 were evaluated.

Results: In the study period (2014-2019), the onset of the rainy season ranged from week (W) 17 (W17; April) to W34 (August). The onset of the high transmission season ranged from W25 (June) to W40 (September). The lag between these two events ranged from 5 to 12 weeks. The duration of the high transmission season ranged from 3 to 6 months. The best of the two approaches predicted the onset of the high transmission season in 2019 to be in June in two districts, in July in 46 districts, in August in 21 districts and in September in six districts. Using our proposed approach would prevent 43,819 cases, 1943 hospitalisations and 70 deaths in children under 5 years of age annually for a minimal additional cost. Our analysis shows that the number of cycles of SMC should be changed in 36 health districts.

Conclusion: Adapting the timing of SMC interventions using our proposed approach could improve the prevention of malaria cases and decrease hospitalisations and deaths. Future studies should be conducted to validate this approach.

Keywords: High transmission season; Malaria; Rainfall; Sub-national; Tailoring.

Fig. 1

a, b Maps showing the predicted onset of the high transmission season in 2019, with the colour scale representing the week (W) of onset: a predicted onset using the App-A predictive approach (based on median rainfall data and median lag data for the 2014–2018 period, b predicted onset using the App-B predictive approach (based on rainfall data for 2019 and median lag data for the 2014–2018 period). c, d Maps showing the prediction error in weeks, with the colour scale representing the error value (note: a difference between - 2 and + 2 weeks was considered acceptable): cprediction error using App-A, d prediction error using App-B

Fig. 2

Density of standardised scores with confidence intervals. Standardised scores for App-A and App-B are shown in red and blue, respectively

Fig. 3 The Blue line on a silver background represents the smoothing curve and the confidence intervals

Duration and seasonality of malaria transmission in four representative health districts. a Bimodal seasonality, b usual seasonality (from July to December), c irregular seasonality, d low seasonal transmission

Fig. 4

Number of cycles of seasonal chemoprevention needed in each health district based on the duration and seasonality of malaria transmission. Colour scale represents the required number of cycles. SMC, Seasonal malaria chemoprevention

Fig. 5

Maps showing the additional cost that would be incurred and the number of cases that would be prevented annually using App-A. a Additional cost incurred per health district, with the colour scale representing the difference between the cost of using the current approach and the cost of using App-A. b Number of malaria cases prevented per health district, with the colour scale representing the estimated number of cases that would be prevented using App-A