What temperature index is the best predictor for the impact of temperature on mortality / 中华预防医学杂志

ABSTRACT

<p><b>OBJECTIVE</b>To explore the suitable temperature index to establish temperature-mortality model.</p><p><b>METHODS</b>The mortality and meteorological information of Guangzhou between year 2006 and 2010 were collected to explore the association between sendible temperature, heat index and deaths by adopting distributed lag non-linear model to fit the daily maximum, mean and minimum temperature with and without humidity. Q-Q plots based on the standardized residuals of each model were used to qualitatively access the goodness of fitting. The minimum Akaike information criterion (AIC) and residual sum of squares (RSS) value were used to explore the most suitable temperature index for model establishment, and to further analyze the fittest temperature index for different diseases, ages and cold and hot effect.</p><p><b>RESULTS</b>Guangzhou features a subtropical monsoon climate, with an annual average temperature at 22.9°C and daily average relative humidity of 71%. The standardized residuals of all models followed normal distribution. For all death, death from circulation system diseases, the 65-84 years old aging groups and cold effect models, the daily average temperature fit better, whose AIC (RSS) values were the smallest as 11 537 (1897), 9527 (1928), 10 595 (2018) and 11 523 (1899), respectively. However, for death from respiratory system disease, groups aging under 65 years old or over 85 years old and hot effect models, the daily average sendible temperature fit better, whose AIC (RSS) values were the smallest as 8265(1854), 675 (1739), 8550 (1871) and 11 687 (1938), respectively. In comparison with the model controlling both temperature and relative humidity, different diseases, aging groups and cold and hot effect models fitted by sendible temperature index showed smaller AIC (RSS) values. The relative risk (RR) value of the cold effect lagging 0 - 3 days fitting by daily maximal temperature was < 1, and the RR value of it fitting by daily minimum temperature was > 1.04. The RR value of the hot effect lagging 0 - 1 days fitting by daily maximal temperature was < 1.16, and the RR values of it fitting by daily minimum temperature and daily average temperature were > 1.16.</p><p><b>CONCLUSION</b>There were no best temperature indicators for different diseases, ages and cold and hot effect. The model using sendible temperature index better fit the model including relative humidity as a covariable.</p>