Thirty-six observational studies of heat exposure and chronic obstructive pulmonary disease were pooled using a three-level random-effects model to separate morbidity from mortality and to test how the definition of heat affected the estimate. The pooled risk ratio was 1.15 overall, 1.26 for mortality and 1.05 for morbidity. Heatwave and threshold-based definitions both yielded 1.21 against 1.05 for continuous temperature, and adjustment for air pollution did not materially change the associations.
Why it is interesting: The gap between mortality and morbidity estimates, and the dependence of the effect size on how heat is defined, bears directly on how heat-attributable deaths are counted.