Mapping the short-term exposure–response relationships between environmental factors and health outcomes and identifying the causes of heterogeneity: A multivariate-conditional-meta-autoregression-based two-stage strategy

Studying the spatial distribution of short-term exposure–response relationships (ERRs) between environmental factors and health-related outcomes and identifying the causes of spatial heterogeneity are of great importance on making region-specific environment-related public health policies. However, the widely used multivariate meta-regression (MMR)-based two-stage strategy does not consider the spatial dependence between regions, which may give unsatisfactory results, even a false policy implication. More importantly, possibly due to the limitation, the spatial distribution of short-term ERRs is less frequently focused on. In this work, we combined the conditional autoregression with MMR to construct an extended model called MCMAR. Then a MCMAR-based two-stage strategy is developed to map the ERRs and identify the causes of heterogeneity. A published motivating example and a simulation study were used to validate the efficiency of our strategy. Results show that the MCMAR-based strategy achieved considerably better fit performance in terms of the Akaike information criterion, obtained a more reasonable spatial distribution of ERRs, and identified more accurate causes of heterogeneity than the classic strategy. As numerous spatial ERR datasets have been and are being produced, we believed that MCMAR-based two-stagy strategy will have an important and wide application value.