Fig. 1

Pooled association between relative temperature (percentiles) and all-cause mortality in 108 US cities during the summer period in 1987–2000 in Case Study 1. The x-axis is scaled so that the summer temperature distribution match the average percentiles of all the cities. The left panel shows the average heat-mortality curve estimated by the multivariate meta-analysis. The right panel illustrate the effect modification from population size, predicted from the full multivariate-meta-regression at the 10^th-90^th percentile values of the city-specific meta-variable