Quantifying the lagged effects of non-optimal temperature exposure on risk of cause-specific mortality through distributed lag non-linear modeling and mixed-effects meta-regression

Thesis (Master's)--University of Washington, 2022Background: Exposure to high and low ambient temperature is associated with increased risk for mortality from a variety of causes. Previous analyses have shown that elevated risk persists for a period of time following a temperature exposure; however, no analysis has yet explored these lagged effects on the cause-specific level. We therefore assessed the relationship between non-optimal temperature exposure and mortality by creating cause-specific relative risk curves incorporating lagged effects.Methods: We matched International Classification of Diseases-coded cause of death data with administrative unit identifiers to daily and mean annual temperature estimates from the ERA5 global temperature dataset. For each administrative unit, we modelled cause-specific relative risks incorporating lagged effects for 17 communicable, non-communicable, and external causes using a distributed lag non-linear modeling framework. We meta-regressed the resulting relative risks using a Bayesian, two-dimensional spline surface specified by daily temperature and 23 mean temperature zones to predict a set of globally-applicable, cause-specific risk curves. Results: We included 52.7 million individual mortality observations from 12,641 administrative units across 8 countries for the years 1980 – 2016. For non-external causes (ischemic heart disease, stroke, cardiomyopathy and myocarditis, hypertensive heart disease, diabetes, chronic kidney disease, lower respiratory infection, and chronic obstructive pulmonary disease) we observed J-shaped relationships between daily temperature exposure and risk of mortality, while external causes (homicide, suicide, drowning, and related to disasters, mechanical, transport, and other unintentional injuries) displayed monotonically increasing relationships. For non-external causes, lagged effects existed for both heat and cold, though lagged effects for cold were more pronounced and occurred across almost all causes and temperature zones. Lagged effects for heat appeared primarily and moderate and warmer temperature zones for most non-external causes. A lag period of 21 days resulted in the highest RRs for non-external causes for cold, but for heat, lag periods of 7 or 14 days sometimes observed higher RRs than a 21-day lag period. Compared to non-external causes, external causes displayed more heterogeneous lagged effects. Conclusion: Exposure to extreme cold and heat presents substantial risks to human health, but exposure-response relationships vary by cause. Lagged effects for both temperature extremes indicate that increased risk of mortality is not limited to the single day on which an extreme temperature event occurs for most causes. Burden assessments should incorporate lagged effects to avoid underestimating non-optimal temperature’s disease burden. Medical, public health, and structural decision making must consider sustained effects of extreme temperature exposure across time when designing strategies to address climate change and extreme weather events