Unbiased temperature-related mortality estimates using weekly and monthly health data: a new method for environmental...

Exposure to environmental factors has a high burden on human health, with millions of premature annual deaths associated with the short-term health effects of ambient temperatures and air pollution. However, direct estimations of exposure-related mortality from real data are still not available in most parts of the world, especially in low-resource settings, due to the unavailability of daily health records to calibrate epidemiological models.

In this study, we have filled the crucial gap in available direct estimations by developing a method to make valid inference for the relationship between exposure and response data that uses only exposure and temporally aggregated response data. We provided the mathematical derivation of the method, and compared the results by using simulations applied to daily temperature and daily, weekly, and monthly mortality data. The method was then applied to the newly created database of the EARLY-ADAPT project.

The daily and weekly models produced similar and unbiased estimates of the temperature-related relative risks and attributable mortality, with only slightly more imprecision in the weekly model. Even the estimates of the monthly model were unbiased when using enough data, although at the expense of a substantial increase in variability. The real data analysis showed that the similarity between the regional values of two aggregation models increased with the number of years and regions of the dataset, and decreased with the difference in their degree of temporal aggregation.

Our method opens the door to conducting epidemiological studies in low-resource settings, where access to daily health data is not possible. Moreover, it allows accurate estimation of the short-term health effects of environmental exposures in near-real time, when daily health data are still not available, such as in the estimation of the mortality burden of recent record-breaking heat episodes. Overall, our method represents an important new approach to how the public health community can use data to create new evidence for research, translation and policy making.

European Research Council (ERC).