This study presents a modeling framework for West Nile virus (WNV) risk in Europe that integrates spatiotemporal dynamics, reporting heterogeneity, and a causal perspective on explanatory variables. The authors identify three common limitations of zoonotic risk maps — lack of a spatiotemporal autocorrelation component, failure to account for heterogeneity in case reporting, and absence of a causal framework for variables — and address these in a unified model system built for WNV, a vector-borne pathogen transmitted in a bird reservoir and affecting humans and horses.
The authors built a spatiotemporal occupancy model and fitted it to notified WNV case data for humans and horses. Surveillance data used in the analysis were obtained from the European Centre for Disease Prevention and Control (ECDC) after a formal request. The model explicitly incorporated spatiotemporal autocorrelation to better represent the spread and persistence of infection across space and time. Specific model parameter values and technical details are provided in the original preprint and in the code repository cited by the authors.
To estimate causal effects, the investigators based their analysis on a directed acyclic graph (DAG). They separated explanatory variables into two conceptual groups: (1) conjectural weather variables that change in the short term, and (2) structural variables that represent long-term characteristics of the bird community. The DAG guided identification of direct and indirect pathways linking weather and structural factors to WNV circulation in the avian reservoir and to detected cases in humans and horses.
A key component of the modeling approach was explicit modeling of heterogeneity in case reporting — that is, under-reporting of WNV cases across space and time. By fitting the occupancy model to notified human and horse case data and including a reporting component, the authors sought to separate true circulation in the bird reservoir from variation introduced by surveillance practices and reporting completeness.
Using the fitted model, the authors computed population attributable fractions to quantify the relative contributions of conjectural weather variables and bird community structure to WNV outbreaks in Europe. They report that the contribution of short-term weather variables to outbreaks was, at the continental scale, globally higher than the contribution of the structure of the bird community. The preprint emphasizes this finding as evidence that short-term climatological or meteorological variability can be a dominant driver of outbreak occurrence, when accounting for spatiotemporal autocorrelation and reporting heterogeneity.
The study highlights methodological elements that can improve risk mapping for vector-borne zoonoses: incorporating spatiotemporal autocorrelation, explicitly modeling reporting heterogeneity, and using a causal framework (DAG) to clarify the roles of short- versus long-term drivers. For WNV in Europe, the findings suggest that surveillance and response planning should consider the outsized role of short-term weather variability in outbreak emergence, while continuing to account for structural ecological factors that shape baseline risk.
The authors have made the code reproducing the study available at: https://github.com/JonathanBas/WNV_Europe. Funders declared in the preprint include Conseil Régional d'Île-de-France (DIM1Health) and the Agence Nationale de la Recherche (ANR 24 PEPZ 0004). The authors declared no competing interests.
The preprint states that relevant ethical guidelines were followed. The dataset of West Nile virus surveillance data was accessed from the ECDC after formal request, and the authors confirm that appropriate patient/participant consent and institutional approvals or exemptions were obtained and documented. The manuscript was posted to medRxiv on August 23, 2026.
Note: This summary and rewrite reflect the facts reported in the cited medRxiv preprint. Technical specifics, model equations, parameter estimates and additional results are available in the original preprint and accompanying code repository; those details were not fully enumerated in the source text provided here.