Social vulnerability and mortality attributable to non-optimal temperature in the United States: A county-level...

To quantify county-level mortality attributable to non-optimal temperature in the United States and examine whether social vulnerability modifies this relationship.

Ecological time-series analysis from 2000-2020.

We analyzed 1,514 counties representing 91.2% of the 2010 U.S. adult population (ages 25-84), including 33,395,241 deaths (after imputation: 33,421,054) which were linked to monthly mean temperature. A two-stage modeling framework was used. First, quasi-Poisson models with natural cubic splines estimated county-specific non-linear temperature-mortality associations, with population offsets and spline-based control for seasonality. Second, coefficients and variances were pooled using random-effects multivariate meta-analysis to identify the minimum mortality temperature (MMT) and percentile (MMP). Effect modification was assessed by incorporating Social Vulnerability Index (SVI) quartiles into a multivariate meta-regression. Attributable deaths were estimated using 5,000 Monte Carlo draws and classified as heat-related (above MMT) or cold-related (below MMT); crude rates were expressed per 100,000 person-years.

The pooled MMT was 22.7 °C (95% CI: 22.2-23.2), corresponding to the 78.4th percentile (95% CI: 76.1-80.8). Nationally, an estimated 72,361 (95% CI: 68,837-76,153) cold-attributable and 6,129 (95% CI: 5,309-7,227) heat-attributable deaths occurred annually, equivalent to 40.1 and 3.4 per 100,000 person-years. More socially vulnerable counties had higher MMTs [SVI Q4: 23.9 °C vs Q1: 21.0 °C] and higher heat- (2.0 vs 0.7) and cold-related mortality rates (40.9 vs 28.0).

Cold accounted for most temperature-related deaths in the U.S., and social vulnerability intensified both cold- and heat-related mortality. These findings support geographically targeted, vulnerability-responsive public health strategies.