The effect of non-optimal ambient temperature on daily mortality in Colombia 2010-2019.

Adverse effects of non-optimal temperatures on mortality have been reported in different climates. However, only a low number of studies have been conducted in tropical locations where these effects might differ. Here, we estimate the association between ambient air temperature and all-cause mortality and assess the impact of non-optimal temperatures on attributable mortality at national and subnational levels in Colombia during 2010-2019. We obtained daily temperature and mortality data for 32 departments and conducted a two-stage analysis. In stage-1 we fitted a time-series Poisson model for each department and estimated the mortality-temperature association using distributed-lag-nonlinear models with 7-28 days of lag. In stage-2, we pooled these estimates using a multivariate meta-analytic model including mean temperature, relative humidity, and a multidimensional poverty index. We calculated attributable deaths and fractions due to non-optimal temperatures and due to overall heat and cold. We analyzed 2,561,561 deaths and found marked differences in exposure-response curves of mortality-temperature, where most departments showed acute heat effects but no cold effects for the 7-days lag. This lag-response curve for heat showed that the risk of death is higher during the same day (lag 0) of extreme temperatures and decreases after the third day. The country attributable fraction due to non-optimal temperature was higher for heat (1.77, 95% CI 1.16 - 2.31) during the 0-7 days lag, but higher for cold temperatures (4.68, 95% CI 2.34 - 6.72) during the 0-28 days lag. There was high heterogeneity in the estimated risks between departments. These results should inform planning adaptation strategies for climate change differentiated at subnational level.