Mapping hotspots of zoonotic pathogen emergence: an integrated model-based and participatory-based approach.

Background: An increase in pandemics of zoonotic origin has led to a growing interest in using statistical prediction to identify hotspots of zoonotic emergence. However, the rare nature of pathogen emergence requires modellers to impose simplifying assumptions, which limit the model's validity. We present a novel approach to hotspot mapping that aims to improve validity by combining model-based insights with expert knowledge.

Methods: We conducted a systematic literature review to identify predictors for zoonotic emergence events in three priority virus families (Filoviridae, Coronaviridae, and Paramyxoviridae). We searched PubMed, Web of Science, Agricola, medRxiv, bioRxiv, Embase, CAB Global Health, and Google Scholar on Oct 14-28, 2021, with no restrictions on language or the date of publication. Articles suggested by subject matter experts and those identified by a review of reference lists were also included. We used regularised regression to fit a model to the data extracted from the literature and produced maps of ranked risk. In a series of workshops in five countries (Kenya, Peru, Senegal, Thailand, and Viet Nam), experts in zoonotic diseases produced qualitative hotspot maps based on their expertise, which were compared with the model-derived maps.

Findings: 425 articles were analysed, from which 19 predictors and 1068 outcome events were identified. The in-sample misclassification error was 0·365, and 89% of participant-selected zones were ranked as moderate or high risk by the model. Participant-selected zones were too large to be actionable without further refinement. Discordance was probably due to missing predictors for which no valid data exist, and homogeneity imposed by our global model.

Interpretation: Concordance between the two sets of maps supports the validity of each. Because model-based and participatory strategies have non-overlapping limitations, the results can be harmonised to minimise bias, and model-based results could be used to refine participant-selected zones. This approach shows potential for refining deployment of countermeasures to prevent future pandemics.

Funding: US Agency for International Development.