bayesian-models-map-argentina-dengue-risk-but-cannot-forecast-season-magnitude
Bayesian Models Map Argentina Dengue Risk but Cannot Forecast Season Magnitude

Bayesian Models Map Argentina Dengue Risk but Cannot Forecast Season Magnitude

A national spatio-temporal analysis of dengue incidence across northern and central Argentina found that large-scale climate patterns, seasonal weather conditions, and local waste sites are jointly associated with reported infections, but spatial models cannot forecast epidemic magnitude.

Before this modeling study, few investigations had analyzed the determinants of reported dengue incidence across Argentina’s heterogeneous transmission zones at a national scale to separate large-scale inter-annual climate variability from local environmental and infrastructure conditions.

To examine these drivers, the research team analyzed 817,699 dengue cases reported across 473 departments in northern and central Argentina from 2009 to 2025. Across this study period, reported disease incidence displayed pronounced inter-annual fluctuations, culminating in the 2024 peak.

The researchers implemented Bayesian hierarchical spatio-temporal models using the integrated nested Laplace approximation framework. Their modeling strategy employed a zero-inflated negative binomial distribution to address excess zeros and overdispersion in departmental case reports. The model incorporated spatial, temporal, and space–time random effects alongside climatic, environmental, and sociodemographic covariates. Predictive performance was evaluated through forward-chaining temporal cross-validation to test how effectively the model generalized to future transmission periods.

Large-scale climatic variations were characterized using the continuous Oceanic Niño Index averaged over the September-to-April transmission window. Warmer El Niño–Southern Oscillation conditions were associated with increased reported dengue incidence independently of seasonal climate. At seasonal scales, higher current-summer temperatures and increased relative humidity were associated with elevated transmission risk, whereas previous-spring temperatures exhibited a non-linear inverse association with reported dengue incidence in the following transmission season.

Local infrastructural conditions also showed clear associations with departmental transmission patterns. Reported dengue incidence was higher in departments where a larger share of the population lives within 500 meters of waste-disposal sites. This finding indicates that neighborhood waste conditions are linked to reported departmental case burdens, highlighting how localized environmental factors operate alongside broader climatic drivers in shaping infection risks.

Out-of-sample evaluations highlighted distinct practical boundaries for model application. While the spatio-temporal framework consistently ranked departments by relative transmission risk, it did not predict the intensity of an upcoming dengue season. The authors concluded that although Bayesian risk mapping provides a sound foundation for spatial targeting in prevention planning, the empirical evidence does not support using these models to forecast epidemic magnitude.

Subject of Research: Spatio-temporal modeling of reported dengue incidence in Argentina

Article Title: Spatio-temporal modeling of dengue in Argentina (2009–2025): factors associated with reported incidence, spatial patterns, and implications for prevention planning

Article References: Piette, J., Linard, C., Morlighem, C., & Andreo, V. (2026). Spatio-temporal modeling of dengue in Argentina (2009–2025): factors associated with reported incidence, spatial patterns, and implications for prevention planning. International Journal of Health Geographics. https://doi.org/10.1186/s12942-026-00513-2

Image Credits: AI Generated

DOI: 10.1186/s12942-026-00513-2

Keywords: dengue, Argentina, Aedes aegypti, spatio-temporal modeling, Bayesian hierarchical models, integrated nested Laplace approximation, Oceanic Niño Index, El Niño-Southern Oscillation, waste-disposal sites, zero-inflated negative binomial, relative humidity, summer temperature, forward-chaining cross-validation, spatial risk mapping, epidemic forecasting, disease surveillance, environmental covariates