Definition of mortality-associated thermal markers for an extreme heat alert system in Belo Horizonte, Minas Gerais, Brazil.

OBJECTIVE: To estimate daily maximum temperature markers for defining heat wave alert levels in Belo Horizonte, Minas Gerais, Brazil. METHODS: the association between daily maximum temperature and mortality was estimated using quasi-Poisson regression with distributed lag non-linear models, considering lags from 0 to 21 days and adjustment for temporal trend, seasonality, and day of the week. The minimum mortality temperature was used as the reference for estimating relative risks, and high percentiles supported the definition of operational thresholds. RESULTS: The minimum mortality temperature was 29.3°C. Mortality risk increased at higher percentiles, with relative risks of 1.04 at the 90th percentile; 1.09 at the 95th percentile; 1.17 at the 98th percentile; and 1.26 at the 99th percentile. Four levels based on maximum temperature and thermal persistence were proposed: Alert 1, mild heat wave (32.0-33.1°C/3 days); Alert 2, moderate heat wave (33.2-34.6°C/3 days); Alert 3, intense heat wave (34.7-35.8°C/3 days); and Alert 4, extreme heat wave (>35.8°C/2 days). A higher frequency of heat waves was observed in years associated with El Niño. CONCLUSION: Markers based on high percentiles of maximum temperature produced levels consistent with the increased mortality risk, with potential application in public health surveillance and response.