Compound and cascading effects of climatic extremes on dengue outbreak risk in the Caribbean: an impact-based modelling framework with long-lag and short-lag interactions

Background

Small islands developing states in the Caribbean are exposed to increasingly frequent and intense extreme climatic events, which can exacerbate outbreaks of climate-sensitive infectious diseases. Few forecasting tools incorporate the compound and cascading effects of multiple delayed climatic indicators on disease outbreak risk. We aimed to create an impact-based modelling framework that employs interactions between climatic predictors to forecast the probability of a climate-sensitive infectious disease outbreak 3 months ahead, and to investigate the compound and cascading effects of temperature and long-lag and short-lag standardised precipitation index (SPI) on dengue outbreak risk in Barbados.

Methods

We developed a modelling framework to predict the probability of a dengue outbreak in Barbados with a 3-month lead time. We assessed the relationships between dengue incidence and interacting long-lag and short-lag hydrometeorological predictors with confirmed cases from 1999 to 2022 and a Bayesian hierarchical framework accounting for seasonal and interannual variation. With this long–short-lag interaction model, we piloted a dengue early warning system in Barbados for the International Cricket Council Men's Twenty20 World Cup in June, 2024, as a real-world prospective example.

Findings

We found that a three-way interaction between the 3-month averaged mean temperature anomaly lagged by 3 months, 6-month SPI (SPI-6) lagged by 5 months, and SPI-6 lagged by 1 month best predicted dengue outbreak risk in Barbados. Our findings showed that long-lag dry (lagged by 5 months), mid-lag hot (lagged by 3 months), and short-lag wet (lagged by 1 month) conditions led to the greatest dengue risk. During cross-validation from 2012 to 2022, the model exhibited a true positive rate (TPR) of 81% and a false positive rate (FPR) of 29%, outperforming a baseline model representing standard practice with a TPR of 68% and an FPR of 48%. For the Twenty20 World Cup, the model predicted a 95% outbreak probability due to epidemiological and climatic conditions, which was shared with the Barbados Ministry of Health and Wellness ahead of the tournament.

Interpretation

Our impact-based modelling framework with long-lag and short-lag interactions explicitly accounted for the compound and cascading effects of drought, heat, and excessively wet conditions on dengue outbreak risk in Barbados. The model is being implemented in a national dengue early warning system with ongoing monitoring and evaluation to ensure its reliability and usefulness in operational contexts. Future work could explore the applicability of this methodology to modelling or predicting climate-sensitive infectious diseases in other endemic settings.