Objective  To explore the relationship between ambient temperature and mortality among residents in Chengdu municipality and to provide evidences for developing relevant interventions in vulnerable populations.

  Methods  The population of the study were permanent residents in 20 districts/counties of Chengdu municipality, Sichuan province during the period from January 1, 2013 to December 31, 2018. We collected daily data on meteorology (mean temperature, relative humidity, and air pressure), atmospheric pollutants (average concentration of particulate matter < 10 microns in aerodynamic diameter PM₁₀ and ozone in 8 hours O_{3 – 8}, and number of all-cause deaths. The average daily temperatures equal or less than the 2.5th percentile of all the values were defined as low temperatures, and those equal or greater than to the 97.5th percentile of all the values were defined as high temperatures. Distributed lag non-linear model was used to analyze exposure-related lag-response relationship between average daily temperature and daily number of deaths. A minimum mortality temperature (MMT), at which mortality being the lowest, was adopted as the reference level to calculate attribute number and attribute fraction of deaths attributed to the exposure to a specific average daily temperature.

  Results  During the 6 – year period, the average daily temperature for the study area ranged from – 1.9 ℃ to 29.8 ℃, and the median was 17.5 ℃; totally 484 736 deaths were registered and the daily number of deaths ranged from 137 to 375, with a median of 214; the medians were 950.8 kPa for average daily pressure, 80% for relative humidity, 88.3 μg/m³ for PM₁₀, and 79.3 μg/m³ for O_{3 – 8}, respectively. There was a statistical correlation between the average daily temperature and daily number of deaths (P < 0.05), with an approximate L-shape exposure-response relationship and a MMT of 25 ℃. The effect of high temperature exposure on mortality was observed at lag days 0 – 5 and that of low temperature was observed at lag days 1 – 13. The cumulative relative risk of mortality at lag days 0 – 21 was 1.142 (95% confidence interval 95% CI: 1.087 – 1.199) for the exposure to high temperature and 1.405 (95% CI: 1.244 – 1.587) for the exposure to low temperature. For the whole population during the period and taking the baseline number of deaths under MMT as the reference, the estimated cumulative number of deaths attributed to unfavorable temperature at lag days 0 – 21 was 60 280 (95% CI: 30 875 – 85 953) and the estimated attributable fraction was 12.4% (95% CI: 6.5% – 17.6%); while, the estimated cumulative number of deaths attributed to low and high temperature at lag days 0 – 21 were 56 794 (95% CI: 28 080 – 82 993) and 3 493 (95% CI: 2 053 – 4 414) and the estimated attributable fraction were 11.7% (95% CI: 5.9% – 17.1%) and 0.72% (95% CI: 0.44% – 0.96%), respectively.

  Conclusion  There is an approximate L-shaped exposure-response correlation between average daily temperature and daily number of deaths among residents in Chengdu municipality and the adverse effect of low temperature is greater than that of high temperature, with higher relative risk, long lag time of effect, and larger attributable death number and fraction.