Linking climate and dengue in the Philippines using a two-stage Bayesian spatio-temporal model

TL;DR

This study models the relationship between climate factors and dengue in the Philippines using a Bayesian spatio-temporal framework, revealing complex associations with temperature and rainfall across different regions.

Contribution

It introduces a two-stage Bayesian modeling approach that incorporates climate uncertainty and spatial variation to better understand dengue-climate links.

Findings

Temperature positively correlates with dengue, but extremely hot conditions reduce risk.

Rainfall's effect on dengue varies: negative in consistently rainy areas, positive in seasonal regions.

Significant spatial and temporal variation remains unexplained after accounting for climate factors.

Abstract

Dengue is an infectious disease which poses significant socioeconomic and disease burden in many tropical and subtropical regions of the world. This work aims to provide additional insight into the association between dengue and climate in the Philippines. We employ a two-stage modelling framework: the first stage fits climate models, while the second stage fits a health model that uses the climate predictions from the first stage as inputs. We postulate a Bayesian spatio-temporal model and use the integrated nested Laplace approximation (INLA) approach for inference. To account for the uncertainty in the climate models, we perform posterior sampling and then perform Bayesian model averaging to compute the final posterior estimates of second-stage model parameters. The results indicate that temperature is positively associated with dengue, although extremely hot conditions tend to have a negative effect. Moreover, the relationship between rainfall and dengue varies in space. In areas with uniform amounts of rainfall all year round, rainfall is negatively associated with dengue. In contrast, in regions with pronounced dry and wet season, rainfall shows a positive association with dengue. Finally, there remains unexplained structured variation in space and time after accounting for the impact of climate variables and other covariates.