Effect Modification of Temperature Variability on the Association Between Ambient Heat and Hospitalization: A Nationwide Study in Brazil Between 2000 and 2015.

Little evidence is available about how temperature variability (TV) may modify the association between ambient heat and health conditions, especially in less developed countries, for example, Brazil. This study explored the change in heat-hospitalization association from days with low to high temperature variabilities (TVs) in Brazil. We collected data on daily hospitalization and weather from 1,814 Brazilian cities during the 2000-2015 hot seasons. For each city, the heat-hospitalization association was estimated using the quasi-Poisson regression with constrained lag model on days with low and high TVs, respectively. City-specific effect estimates were then pooled using random-effect meta-analysis. Stratified analyses were performed by region, sex, age-group and cause category. Our results showed that at the national level the risk of hospitalization increased by 4.4% [95% confidence interval (CI): 4.0%-4.7%] and 2.7% (95%CI: 2.4%-3.0%) per 5°C increase in daily mean temperature on days with low and high TVs, respectively. The effect modification of TV on heat-hospitalization association was significant in the northeast, southeast and south, and for most population groups except for the children aged 0-4 years and the elderly aged ≥80 years. Significant effect modification of TV was observed in injury and poisoning, respiratory and cardiovascular diseases. Our findings suggest that in Brazil, the heat-health association is substantially modified by exposure to TV. The high heat susceptibility on the day with low TV warrants particular concern in order reducing the risk of disease burden.