A study on the identification of threshold for early warning on adverse weather events based on the association of apparent temperature and years of life lost

Objective: To identify the threshold of a health warning system based on the association of apparent temperature and years of life lost (YLL). Methods: Daily mortality records and meteorological data were collected from 364 Chinese counties for 2006-2017. Distributed lag nonlinear model and multivariate Meta-analyses were applied to estimate the association between the apparent temperature and YLL rate. A regression tree model was employed to estimate the warning thresholds of the apparent temperature. Stratified analyses were further conducted by age and cause of death. Results: The daily YLL rate was 23.6/105. The mean daily apparent temperature was 15.7 ℃. U-shaped nonlinear associations were observed between apparent temperature and YLL rate. The actual temperature-caused YLL rate for the elderly was higher than the young population. The daily excess deaths rate increased with the higher effect levels. Conclusions: Regression tree model was employed to define the warning threshold for meteorological health risk. The present study provides theoretical support for the weather-related health warning system.