Evaluating the impact of the Malakit intervention on malaria transmission in the Guiana Shield: a mathematical modelling study

Abstract

Background: Malaria elimination in mobile and hard-to-reach populations calls for novel, tailored interventions. In French Guiana, the malaria burden is high among the population working in illegal gold mining. Between April 2018 and March 2020, we implemented Malakit, a new intervention targeting gold miners, and relying on the distribution of kits for the self-diagnosis and self-treatment of malaria. In this study, we evaluate the impact of Malakit on malaria transmission.

Methods: We fitted mathematical models of the transmission of Plasmodium falciparum (Pf) and Plasmodium vivax (Pv) to surveillance data from Brazil and Suriname and to prevalence data from cross-sectional surveys to estimate the change in treatment coverages and reproduction numbers between the pre-intervention (2014-2018) and intervention (2018-2020) periods. Using counterfactual scenario analysis, we estimated the impact of the intervention on malaria incidence.

Findings: Treatment coverage increased in the intervention compared to the pre-intervention period, the effect being stronger for Pf malaria (from 20.6% [95% CrI 17.0, 25.7] to 81.7% [95% CrI 67.9, 96.4]) than Pv malaria (from 22.3% [95% CrI 19.3, 25.7] to 47.3% [95% CrI 42.6, 52.3]). Concomitantly, the mean reproduction number decreased from 0.96 (95% CrI 0.93, 1.00) to 0.34 (95% CrI 0.22, 0.46), for Pf, and from 1.13 (95% CrI 1.10, 1.16) to 0.86 (95% CrI 0.82, 0.90) for Pv. We estimated that between 2018 and 2020 Malakit helped avert a total of 1231 (95% CrI 589, 1870) and 2406 (95% CrI 1191, 3621) symptomatic infections by Pf and Pv, respectively, corresponding to a 44.2% reduction of the overall malaria incidence (95% CrI -52.5, -34.5).

Interpretation: Malakit had a significant impact on malaria transmission by improving the access to treatment for the population working in illegal gold mining in French Guiana. Building on the regional efforts of the past fifteen years, Malakit contributed to another step towards malaria elimination in the Guiana Shield.

Funding: European Union, Global Fund, MoH of Brazil, Cayenne Hospital, Health Regional Agency of French Guiana.

Keywords: French Guiana; Gold mining; Hard-to-reach population; Impact evaluation; Malaria elimination; Mathematical modelling.

Fig. 1

Map showing the location of Malakit distribution sites and pre/post intervention cross-sectional surveys enrolment sites (red dots) at the borders of French Guiana with Brazil and Suriname.

Fig. 2

Compartments used in the malaria transmission models and sources of data used for calibration, for P. falciparum (top) and P. vivax (bottom). Susceptible individuals (S) may be infected without symptoms (IA), or develop symptoms without appropriate treatment (IS), get treatment from the health system (IT) or from using a malakit (IM, during intervention). Individuals infected with P. vivax without radical cure (IA, IS, IM) become hypnozoites carriers (H). A fraction of the cases treated by the health system are notified to malaria surveillance (D). Λ: force of transmission, ε: probability of symptomatic infection, ϕ: probability of treatment (treatment coverage), m: probability of using a malakit as treatment (m = 0 before intervention), 1/μ_NT: infection duration without treatment, 1/μ_T: infection duration with treatment, ρ: probability of notification to surveillance systems (reporting fraction), μ_HI: P. vivax relapsing rate, μ_HS: P. vivax hypnozoite clearance rate. For more details about notations refer to Supplementary Table S4.

Fig. 3

Monthly incidence of reported cases of malaria (black dots), model-fitted simulations (blue solid line) and counterfactual estimates (dashed grey line) for P. falciparum (top) and P. vivax (bottom). Annual prevalence measured in PCR surveys (orange dots) and model-fitted estimates of prevalence (red dots). Maximum daily temperature (monthly average) is shown in green dashed line. The period of implementation of Malakit is shown in yellow. Vertical lines represent 95% CrI. For model-fitted and counterfactual incidence (respectively blue and grey), the dark envelopes show the uncertainty (95% CrI) associated with the joint posterior distribution, while the light envelopes account also for the uncertainty resulting from the observed Poisson process.