Zero-inflated and distributed lag nonlinear models with random effects for assessing environmental impacts on...

Climate change and air pollution are key determinants of public health, particularly in the onset and exacerbation of respiratory diseases. The main objective is to quantify the lagged and nonlinear effects of climate and air pollution on respiratory hospitalizations in peripheral regions of Costa Rica.

This study presents a methodological framework that combines Distributed Lag Nonlinear Models (DLNM) with Generalized Linear Mixed Models (GLMM), incorporating fixed and random effects, to assess the lagged and nonlinear effects of climatic variables and atmospheric pollutants on hospitalizations due to respiratory causes. The response specification was carried out using zero-inflated distributions, aiming to adequately capture the overdispersion and excess zeros present in the data. The analysis focused on peripheral climatological regions and subregions of Costa Rica-territories outside the Central Valley, including Caribbean and Pacific coasts and border areas, characterized by low population density. Weekly data (2000-2019) on temperature, precipitation, relative humidity, and aerosol optical depth (AOD) were combined with seasonal effects and a population offset to account for subregional differences.

Northern and Central Pacific regions show similar climate-pollution impacts on respiratory health, while the South Pacific exhibits stronger and more persistent risks from moderate to high pollution, and Atlantic regions show consistently higher risks associated with intense rainfall and high humidity. Overall, precipitation extremes, high humidity, and AOD contribute more to respiratory hospitalizations than temperature.

This approach improves explanatory and predictive performance, yields robust relative risk estimates, and captures regional sensitivity to environmental conditions, supporting spatiotemporal health analysis and early warning systems in rural tropical settings.