Assessing dengue fever risk in Costa Rica by using climate variables and machine learning techniques

Abstract

Dengue fever is a vector-borne disease affecting millions yearly, mostly in tropical and subtropical countries. Driven mainly by social and environmental factors, dengue incidence and geographical expansion have increased in recent decades. Therefore, understanding how climate variables drive dengue outbreaks is challenging and a problem of interest for decision-makers that could aid in improving surveillance and resource allocation. Here, we explore the effect of climate variables on relative dengue risk in 32 cantons of interest for public health authorities in Costa Rica. Relative dengue risk is forecast using a Generalized Additive Model for location, scale, and shape and a Random Forest approach. Models use a training period from 2000 to 2020 and predicted climatic variables obtained with a vector auto-regressive model. Results show reliable projections, and climate variables predictions allow for a prospective instead of a retrospective study