The impact of global environmental change on vector-borne disease risk: a modelling study

Background: Vector-borne diseases, such as dengue virus, Zika virus, and malaria, are highly sensitive to environmental changes, including variations in climate and land-surface characteristics. The emergence and spread of vector-borne diseases is also exacerbated by anthropogenic activities, such as deforestation, mining, urbanisation, and human mobility, which alter the natural habitats of vectors and increase vector–host interactions. Innovative epidemiological modelling tools can help to understand how environmental conditions interact with socioeconomic risk factors to predict the risk of disease transmission. In recent years, climate-health modelling has benefited from computational advances in fitting complex mathematical models; increasing availability of environmental, socioeconomic, and disease surveillance datasets; and improved ability to understand and model the climate system. Climate forecasts at seasonal time scales tend to improve in quality during El Niño-Southern Oscillation events in certain regions of the tropics. Thus, climate forecasts provide an opportunity to anticipate potential outbreaks of vector-borne diseases from several months to a year in advance. The aim of this study was to develop a framework to incorporate seasonal climate forecasts in predictive disease models to understand the future risk of vector-borne diseases, with a focus on dengue fever in Latin America. Methods: A Bayesian spatiotemporal model framework that quantifies the extent to which environmental and socioeconomic indicators can explain variations in disease risk was designed to disentangle the effects of climate from other risk factors using multi-source data and random effects, which account for unknown and unmeasured sources of spatial, seasonal, and inter-annual variation. The model was used to provide probabilistic predictions of monthly dengue incidence and the probability of exceeding outbreak thresholds, which were established in consultation with public health stakeholders. Findings: This disease model framework, combined with seasonal climate forecasts, successfully produced real-time, probabilistic estimates of dengue outbreak risk ahead of a mass gathering in Brazil in June, 2014, for all 550 microregions. Additionally, the model predicted dengue outbreak risk after a major El Niño event in southern coastal Ecuador from January to November, 2016. Forecasts from the new model framework did better than those from benchmark models based on historical seasonal dengue averages. Interpretation: This flexible model framework can be adapted to predict any climate-sensitive disease at various spatiotemporal scales and in diverse ecological settings. Incorporating sub-seasonal and seasonal climate forecasts in disease prediction models could support public health decision makers in implementing timely disease-control and prevention strategies months in advance to mitigate the risk of imminent disease epidemics and emerging disease threats. Funding: Royal Society Dorothy Hodgkin Fellowship