Temperature extremes and infant mortality in Bangladesh: Hotter months, lower mortality.

BACKGROUND:Our study aims to obtain estimates of the size effects of temperature extremes on infant mortality in Bangladesh using monthly time series data. METHODS:Data on temperature, child and infant mortality were obtained for Matlab district of rural Bangladesh for January 1982 to December 2008 encompassing 49,426 infant deaths. To investigate the relationship between mortality and temperature, we adopted a regression with Autoregressive Integrated Moving Average (ARIMA) errors model of seasonally adjusted temperature and mortality data. The relationship between monthly mean and maximum temperature on infant mortality was tested at 0 and 1 month lags respectively. Furthermore, our analysis was stratified to determine if the results differed by gender (boys versus girls) and by age (neonates (≤ 30 days) versus post neonates (>30days and <153days)). Dickey Fuller tests were performed to test for stationarity, and since the time series were non-stationary, we conducted the regression analysis based on the first differences of mortality and temperature. RESULTS:Hotter months were associated with lower infant mortality in Bangladesh. Each degree Celsius increase in mean monthly temperature reduced monthly mortality by 3.672 (SE 1.544, p<0.05) points. A one degree increase in mean monthly temperature one month prior reduced mortality by 0.767 (SE 0.439, p<0.1) for boys and by -0.0764 (SE 0.366, NS) for girls. Beneficial effects of maximum monthly temperature were on the order of 0.623 to -0.712 and statistically significant for girls and boys respectively. Effect sizes of mean monthly temperature were larger for neonates at 1.126 (SE 0.499, p<0.05) than for post-neonates at 0.880 (SE 0.310, p