Time-series analysis of hot temperatures and mortality in Ahmedabad, 1987 to 2017

Abstract

TPS 661: Climate change: temperature effects 1, Exhibition Hall, Ground floor, August 26, 2019, 3:00 PM - 4:30 PM Introduction: Few studies have examined the impact of heat wave on mortality in developing countries most at risk due to warmer climate. Methods: We collected daily data on all-cause mortality and air temperature in Ahmedabad, India from 1987 to 2017. We used quasi-Poisson regression models to estimate associations of mortality-heat wave and mortality-temperature over lags of 0-2 days, adjusting nonlinear confounding effects of relative humidity, long-term trend and seasonality. We considered different heat wave definitions as a period of ≥ 2 or 3 consecutive days with daily maximum, minimum, or average temperature exceeding the 95th, 96th, 97th, 98th, or 99th percentile. Models performance was evaluated by one-step time series cross-validation. Effect modifications by age, gender, and residential zone were examined. Results: The model with daily maximum and minimum temperature, and without the indicator for heat wave gave the lowest root mean square error of cross-validation. The maximum and minimum temperature at lag 0 and lag 1-2 periods were statistically significantly associated with all-cause mortality. When the analysis was restricted to days with maximum temperature above 42 °C and minimum temperature above 28 °C, where dose-response curves were almost linear, an increase of 1 °C in maximum and minimum temperature at lag 0 were associated with increases in all-cause mortality of 9.6% [95% confidence interval (CI): 6.6%-12.6%] and 9.8% (95% CI: 6.3%, 13.4%), respectively. Females, older subjects, and those living in South Zone were more susceptible to high temperatures. Conclusions: Results indicate that including an indicator for heat wave in the model lead to overfitting. Both daily maximum and minimum temperature had substantial adverse effects on mortality, particularly above certain thresholds. These findings may have important policy implications on developing threshold for activating excessive heat warning system, and designing of new plans for underdeveloped regions lacking the systems.