Mortality risk due to non-optimum temperatures by cause of death, age, and sex: a multi-country time-series study

Abstract

Background: Non-optimum temperatures have been linked to increased mortality, but the cause-, age-, and sex-specific impacts remain largely unclear. This study investigates the temperature–mortality relationships for nine causes of death across multiple countries/territories and explores subgroup differences by sex and age. Methods: We analysed the non-linear and lagged associations between temperature and mortality in 1117 locations from ten countries or territories (Australia, Brazil, Canada, Chile, Mexico, New Zealand, Philippines, South Korea, Taiwan, and Thailand) covering country-specific periods within 2000–2019, using a two-stage time-series design. In the first stage, a quasi-Poisson generalised linear regression with a distributed lag non-linear model was fitted to estimate the location-specific mortality risk associated with temperature. We then applied a meta-regression model in the second stage to synthesise the associations for cause-specific mortality. Stratified analyses were conducted by sex and age group, and the attributable fraction (AF) of mortality was subsequently estimated. Findings: We identified three distinct exposure-response patterns: an inverse J-shaped curve with higher risks from extreme cold for most causes, a U-shaped curve for all-cause and respiratory mortality, and a J-shaped curve for injury and external causes with greater risks at extreme heat. Significant differences in temperature-related mortality risks were observed across age and sex groups, with the direction and magnitude of these differences varying by cause of death. We estimated that 2.03 million deaths were attributable to non-optimum temperatures during the study period, corresponding to 4.38% (95% CI: 2.02, 6.59) of all-cause mortality. The highest AFs were observed for mental disorders (6.53%), nervous (6.40%), and cardiovascular causes (5.71%). For most causes of death, cold temperatures accounted for the largest proportion of mortality. However, for deaths related to infectious as well as injury and external causes, heat exposure contributed to the majority of the mortality. Interpretation: Our findings demonstrated substantial variations in temperature-related mortality by cause of death, sex, and age. This analysis highlights the need for tailored public health strategies that address the unique vulnerabilities of specific demographic groups across different causes of death, with targeted interventions to mitigate temperature-related health risks. Future work should focus on improving estimation in data-sparse subgroups and developing cause-, age-, and sex-specific projections of temperature-related health risks under future climate scenarios. Funding: The Australian Research Council, Australian National Health and Medical Research Council, VicHealth, and National Research Council of Thailand.