The objectives of this study were to estimate the effects of temperature on cardiovascular mortality in 26 regions in the south and west of China from 2008 to 2011, and to identify socioeconomic and demographic factors contributing to such inter-region variation in the temperature effect. A separate Poisson generalized additive model (GAM) was fitted to estimate percent changes in cardiovascular mortality at low and high temperatures on a daily basis for each region. The model used the smooth functions to model the nonlinear effects of temperature and humidity and to control for the seasonal factor using the calendar time variable. Given variation in the magnitude of the temperature effect on cardiovascular mortality, we employed a Bayesian network (BN) to identify potential region-specific socioeconomic and demographic factors that may explain the variation. In most regions, an increasing trend in high or low temperature was associated with an increase in cardiovascular mortality, with variation in the magnitude of the temperature effects across regions. Three factors, including per capita years of education (as an indicator of economic status), percentage of the population over 65 years of age and percentage of women had direct impact on cold-related cardiovascular mortality. Number of hospital beds (as an indicator of the availability of medical resources), percentage of population engaged in industrial occupations, and percentage of women showed direct impact on heat-related cardiovascular mortality. Due to the socioeconomic and demographic inequalities between regions, the development of customized prevention and adaptation programs to address the low/high temperatures in vulnerable regions should be prioritized.
Keywords: Bayesian network; Poisson generalized additive model; cardiovascular mortality; socioeconomic and demographic factors; temperature.