Effects of climatic factors on rotavirus infection in Bangladesh: a time series analysis and future projections

BACKGROUND: Limited evidence exists concerning the association between climate factors and hospitalised rotavirus cases in Bangladesh. OBJECTIVES: To assess the association between temperature and hospitalised rotavirus cases among children aged < 5 years in Bangladesh. METHODS: We obtained weekly hospital admissions data for rotavirus from seven surveillance hospitals and climate data from the Bangladesh Meteorological Department for August 2013-June 2017. We used a maximum temperature (t-max)-based distributed-lag non-linear model (DLNM) to estimate the association between temperature and hospitalised rotavirus cases, adjusting for seasonality using natural splines. We estimated relative risk (RR), attributable number, and attributable fraction (AF), and scenario-based exploratory projections for 2040 under different Representative Concentration Pathways (RCPs). RESULTS: During the study period, 4,337 hospitalised rotavirus cases were reported among children <5 years. Countrywide, 1,781 (41%) cases were attributable to temperatures above the reference value (25.4°C), reflecting regional heterogeneity. Hospitalisation risk peaked at the 95th percentile of t-max (RR: 5.9; 95% CI: 3.6-9.7). Temperature-hospitalisation associations differed across divisions, with inverse or minimal associations observed in Dhaka, Rangpur, and Sylhet. Scenario-based exploratory projections suggested the highest risk at 0-week lag, with risk increasing again at week-4 lag above 41.7°C. CONCLUSIONS: Associations between t-max and rotavirus hospitalisation varied across Bangladesh, with higher temperatures associated with increased hospitalisation risk in some divisions but inverse or minimal associations in others, highlighting the importance of context-specific climate-sensitive interventions. Scenario-based projections beyond the observed temperature range should be interpreted cautiously, as these estimates represent exploratory extrapolations rather than forecasts of disease burden.