Abstract: Objective To explore the relationship between temperature and mortality in three cities (Beijing, Chengdu and Nanjing) in China and evaluate the temperature-related risk of non-accidental, cardiovascular and respiratory mortality. Methods Data on daily deaths of the three cities as well as meteorological factors and air pollution were collected from January 1, 2008 to December 31, 2010. Distributed lag non-linear model (DLNM) was used to assess the effects of temperature on cause-specific mortality after controlling the long term, seasonal trend and other confounders. Results Non-linear relationships between the temperature and cause-specific mortality were observed in Beijing, Chengdu and Nanjing. The associations of the cumulative hot effects of extremely hot temperature for non-accidental mortality at lag 0 in Beijing, Chengdu, Nanjing were statistically significant (p <0.05), and the relative risk (RR) were 1.09 (95% CI:1.04,1.14), 1.03 (95% CI:1.01,1.05) and 1.17 (95% CI:1.10,1.25), respectively. Meanwhile, the associations of the cumulative cold effects of extremely cold temperature for non-accidental mortality at lag 0-15 in Beijing, Chengdu and Nanjing were statistically significant (p <0.05), the RRs were 1.71 (95% CI:1.43,2.04), 3.09 (95% CI:1.57,6.10) and 1.95 (95% CI:1.21,3.16), respectively. Conclusions Extremely cold and hot temperature could increase the risk of daily cause-specific mortality in Beijing, Chengdu and Nanjing. Hot temperature could cause acute effect, the effect of cold temperature had a several days delay, but a longer persistence.

Key words: temperature, cause-specific mortality, distributed lag non-linear model