Methods/Statistics
Geographic variation in the quality of US mortality data, 1999-2022 Amy Mann* Amy Mann Mann Mann Amy.mann@chch.ox.ac.uk
Objectives: Cause-of-death data underpin public health surveillance, yet few tools quantify mortality data quality in high-income countries at a subnational level. We develop two metrics to measure mortality data quality and apply them to U.S. county mortality from 1999-2022.
Methods: We construct a Re-assignability Index (RI) that quantifies how confidently multiple-cause information can reclassify garbage-coded deaths into meaningful cause groups. We then define a standardized level-of-detail metric using entropy methods to measure coding diversity after accounting for differences in the underlying cause-of-death mix. We also estimate the proportion of deaths with garbage-coded underlying causes and combine these measures into an aggregate data quality index.
Results: Garbage coding declined nationally through 2019 but rose during the COVID-19 pandemic. Coding detail increased steadily, while RI declined modestly, consistent with growing cause-of-death diversity. All dimensions show persistent geographic inequalities, with lower-income counties exhibiting poorer data quality.
Conclusions: U.S. mortality data quality is multidimensional, geographically uneven, and systematically worse in socioeconomically disadvantaged areas. This framework enables scalable monitoring of subnational reporting quality and supports targeted investments to strengthen death certification and public health surveillance.
