The biological link between rainfall and leptospirosis transmission is established, but rigorous quantitative evidence for the insular Caribbean has been limited and largely descriptive. This study aimed to quantify the relationship between monthly rainfall and leptospirosis incidence in the Dominican Republic between 2012 and 2026, using a distributed-lag time-series framework and triangulation with a province fixed-effects panel.
Case data comprised 5,412 valid leptospirosis reports obtained from the national surveillance system (SIP-0276FA51). Rainfall data were provided by the Instituto Dominicano de Meteorologia (INDOMET) from 15 rain gauge stations distributed across 13 provinces, with gauge records spanning 2000–2026. The analytic period for disease–exposure associations covered 2012–2026.
The authors calculated the cross-correlation function between lagged monthly rainfall (lags 0–6 months) and monthly leptospirosis counts to identify candidate lag structure. They then fitted a negative binomial distributed-lag regression model including lags 0–3 months, controlling for seasonality and temporal trend. Findings were triangulated using a 13-province fixed-effects panel model. The analysis compared four different rainfall exposure metrics, conducted threshold-based sensitivity checks for extreme rainfall, and estimated the population attributable fraction (PAF) for rainfall exposure using a parametric bootstrap.
Cross-correlation analysis identified a peak association at a 1-month lag (r = 0.551; 95% CI 0.437–0.647; p < 0.001). The association remained statistically significant through a 3-month lag. In the distributed-lag negative binomial model, each of the four included monthly lags (0–3 months) showed independent associations with case counts; the strongest effect was at the 1-month lag. Specifically, an additional 50 mm of rainfall at 1-month lag was associated with an incidence rate ratio (IRR) of 1.156 (95% CI 1.090–1.226; p < 0.001). The 13-province fixed-effects panel produced a nearly identical estimate for the 1-month lag (IRR = 1.147; 95% CI 1.127–1.167), supporting robustness across modeling approaches.
The study compared multiple rainfall exposure metrics and assessed sensitivity to extreme-rainfall thresholds. Rainfall above the historical 90th percentile increased leptospirosis risk in the following month, with a reported rate ratio of 1.95 for the month after extreme rainfall. The effect magnitude increased monotonically as threshold stringency rose, indicating stronger relative risk following more extreme rainfall events.
Using parametric bootstrap methods, the authors estimated the proportion of cases attributable to rainfall above the recorded historical minimum. The population attributable fraction (PAF) was estimated at 28.2% with a 95% confidence interval of 19.0–36.6%. This quantifies the fraction of reported leptospirosis cases in the study period that could be linked to observed rainfall exposure under the model assumptions.
The analyses indicate that rainfall is a robust and consistent predictor of leptospirosis incidence in the Dominican Republic, with the strongest and most actionable signal at approximately a 1-month lag and elevated risk extending through three months after heavy rainfall. These results are broadly consistent with distributed-lag studies from other settings (Thailand, the Philippines) and represent, to the authors' knowledge, the first formal quantification of this relationship for the insular Caribbean. The findings support the potential utility of rainfall-linked early-warning systems for leptospirosis prevention and response in this endemic setting.
The study protocol was reviewed and approved by the Florida Atlantic University Institutional Review Board (IRB2606182), which determined the project exempt from federal regulation as a secondary analysis of de-identified surveillance and meteorological data. Surveillance data were provided by the Dominican Republic Ministry of Public Health (DIGEPI) and rainfall data by INDOMET; de-identified data are not publicly deposited but may be made available by the corresponding author or the data holders upon reasonable request and institutional approval. The authors declared no competing interests.