A Bayesian Approach to Estimating Changes in Small Area Life Expectancy Disparities in Philadelphia

There is substantial geographic variation in life expectancy (LE). The COVID-19 pandemic caused a mortality crisis, but it is unclear how small-area variations in LE changed during the pandemic. We estimated gender-specific LE by census tract in Philadelphia, Pennsylvania (2017-2019 and 2020-2022), and compared disparities in LE at birth before and during the pandemic. To account for small counts, we employed a Bayesian Poisson model borrowing strength from neighboring census tracts by using conditional autoregressive priors, and from known age-specific mortality patterns by using an adaptation of the TOPALS (tool for projecting age patterns using linear splines) method. We then summarized LE and computed total and income-based disparities in LE by period. The median [10th-90th percentile] change in LE during the pandemic was -2.10 [-5.39 to 0.40] and -1.37 [-3.23 to 0.39] years for men and women, respectively. Declines were concentrated in the same areas that had lower LE at baseline. We found widening disparities due to a steeper decline in LE in lower-income areas during the pandemic. Our study leveraged Bayesian models to produce reliable local-level estimates in small areas, highlighting the disparate impacts that the pandemic had on population health.