Studies have documented significant geographic divergence in U.S. mortality in recent decades. However, few studies have examined the extent to which county-level trends in mortality can be explained by national, state, and metropolitan-level trends, and which county-specific factors contribute to remaining variation. Combining vital statistics data on deaths and Census data with time-varying county-level contextual characteristics, we use a spatially explicit Bayesian hierarchical model to analyze the associations between working-age mortality, state, metropolitan status and county-level socioeconomic conditions, family characteristics, labor market conditions, health behaviors, and population characteristics between 2000 and 2017. Additionally, we employ a Shapley decomposition to illustrate the additive contributions of each changing county-level characteristic to the observed mortality change in U.S. counties between 1999–2001 and 2015–2017 over and above national, state, and metropolitan–nonmetropolitan mortality trends. Mortality trends varied by state and metropolitan status as did the contribution of county-level characteristics. Metropolitan status predicted more of the county-level variance in mortality than state of residence. Of the county-level characteristics, changes in percent college-graduates, smoking prevalence and the percent of foreign-born population contributed to a decline in all-cause mortality over this period, whereas increasing levels of poverty, unemployment, and single-parent families and declines manufacturing employment slowed down these improvements, and in many nonmetropolitan areas were large enough to overpower the positive contributions of the protective factors.

研究记录了近几十年来美国死亡率的显着地域差异。然而，很少有研究检验了县级死亡率趋势在多大程度上可以用国家、州和大都市级别的趋势来解释，以及哪些县级特定因素导致了剩余的变化。将死亡人口生命统计数据和人口普查数据与随时间变化的县级背景特征相结合，我们使用空间显式贝叶斯层次模型来分析工作年龄死亡率、州、大都市地位与县级社会经济条件、家庭特征之间的关联、劳动力市场状况、健康行为和 2000 年至 2017 年的人口特征。此外，我们采用 Shapley 分解来说明每个变化的县级特征对 1999-2001 年和 2015-2017 年间美国县观察到的死亡率变化的附加贡献，超过了国家、州和大都市-非大都市死亡率趋势。死亡率趋势因州和大都市地位而异，县级特征的贡献也是如此。大都市地位比居住州更能预测县级死亡率的差异。在县级特征中，大学毕业生百分比、吸烟率和外国出生人口百分比的变化导致这一时期全因死亡率下降，而贫困、失业和单亲家庭的水平上升制造业就业的下降减缓了这些改善，