Designing spatial adaptive surveillance for the emerging malaria vector Anopheles stephensi in Eastern and Horn of Africa

Sedda, Luigi and Ochomo, Eric and Tadesse, Fitsum G. and Khaireh, Bouh Adbi and Demissew, Assalif and Demisse, Mulugeta and Getachew, Dejene and Guelleh, Samatar K. and Ibrahim, Mohamed M. and Abongo, Bernard and Moshi, Vincent and Muchoki, Margaret and Polo, Brian and Maige, Janice and Kipingu, Andrea M. and Mlacha, Yeromin P. and Sangoro, Onyango and Adeleke, Monsuru and Adeogun, Adedapo O. and Ayodele, Babalola and Okumu, Fredros O. and Pang, Xiaoxi and Ferguson, Heather M. and Kiware, Samson (2026) Designing spatial adaptive surveillance for the emerging malaria vector Anopheles stephensi in Eastern and Horn of Africa. Other. medRxiv.

Abstract

The spread of Anopheles stephensi into the Horn of Africa represents one of the main challenges for malaria control, given the species’ ecological plasticity and resistance to multiple insecticides. In response to the World Health Organization’s 2022 vector alert, an adaptive, model-based spatial surveillance framework was developed and evaluated to improve detection, mapping accuracy, and operational responsiveness during invasion. Adaptive surveillance utilises initial observations to guide subsequent surveillance, linking the surveillance design to the underlying geographical characteristics of Anopheles stephensi distribution through observed data. This dynamic approach targets areas of high uncertainty and/or abundance, making the design responsive rather than predetermined. Focusing on Djibouti and selected regions of Ethiopia and Kenya, the adaptive surveillance was designed on previous in-country Anopheles stephensi surveillance data integrated with assembled open-source environmental, epidemiological, and demographic covariates. Key driver factors of the average monthly Anopheles stephensi catches varied geographically, although seasonality was universally important. Adaptive site allocation was optimised using a multicriteria target function which combines the trapping probability and uncertainty from previous surveys, with a simulation based on peaks-over-threshold (generalized Pareto) modelling of exceedances and Bayes factor–guided prioritisation. The selected adaptive surveillance design is the one that minimise the uncertainty in Anopheles stephensi trapping probability in hotspot areas. Optimal adaptive designs required between 50 to 59 sites per country, with uncertainty reductions in the probability of trapping projected up to 36% in Djibouti and more than 60% in Ethiopia and Kenya, with more than 60% site implementation halving uncertainty in Djibouti and Kenya and reducing it by up to 75% in Ethiopia. The proposed adaptive surveillance framework operationalises WHO guidance, accelerates hotspot identification, and inform targeted ecological studies and control interventions. It is extensible to other urban vectors (e.g., Aedes aegypti), enabling integrated, cross-border surveillance essential to contain Anopheles stephensi during ongoing invasion.