Age-specific climate sensitivity of respiratory hospitalizations in a tropical coastal city: A 20-year machine learning analysis

Respiratory diseases remain a major public health challenge in tropical coastal cities, where persistent heat-humidity coupling and climate variability create vulnerability patterns. We investigated associations between atmospheric conditions and respiratory disease hospitalizations in Maceió, Brazil, a coastal tropical city, using 20 years of data (2000–2019). Weekly hospitalization rates stratified by age (children 0–4 years, adults 5–59 years, elderly ≥ 60 years) were analyzed against meteorological variables including temperature, humidity, precipitation, atmospheric pressure, and solar radiation at 0-, 1-, and 2-week lags. Random Forest models were applied to forecast weekly respiratory hospitalization rates. Minimum temperature showed a strong inverse correlation with hospitalizations across all age groups (ρ = −0.65, p < 0.001), with effects persisting up to 2 weeks. Children exhibited immediate sensitivity to thermal and precipitation variables, while elderly populations showed delayed responses to barometric pressure and evaporation. The Random Forest model achieved excellent-to-good predictive accuracy (R² = 0.83–0.90 for children and adults; Symmetric Mean Absolute Percentage Error = 13–25% across all groups). Long-term declining trends in children and adults contrasted with stabilization and subsequent increases among elderly populations after 2010, reflecting demographic aging and heightened climate sensitivity. These findings provide a transferable framework for climate-informed respiratory risk assessment and early warning systems in tropical coastal environments, supporting age-sensitive public health planning under ongoing climate change.