Comparability of Mortality Rates From Health Surveys Linked to National Death Index, 2015 Versus 2019: Revised...

The Linked Mortality File is vital for evaluating U.S. mortality disparities, but its 2019 iteration adopted a linkage algorithm. To determine if these methodological changes disproportionately affect mortality estimates for marginalized groups, this study compared the 2015 and 2019 Linked Mortality File versions across racial/ethnic and immigrant populations.

This cohort study analyzed 1,072,404 National Health Interview Survey participants (1998-2014, analysis in 2025). Follow-up through 2015 used deaths captured in the 2015 and 2019 Linked Mortality File versions. Cox proportional hazards regressions assessed associations. Mortality rates and hazard ratios were survey-weighted and adjusted for age and sex.

The 2015 and 2019 Linked Mortality File resulted in similar overall mortality rates (114 vs 110 per 10,000 person-years, 1%Δ), but substantial declines occurred among Latinx (124 vs 85, 31%Δ), Asian (109 vs 73, 33%Δ), and noncitizen (123 vs 75, 39%Δ) populations. Consequently, observed mortality patterns often reversed. For example, 2015 Linked Mortality File showed that Latinx participants were more likely to die than their White counterparts (hazard ratio=1.13 [95% CI=1.10, 1.15]), but 2019 version suggested the opposite (hazard ratio=0.77 [95% CI=0.75, 0.79]). The two versions diverged most significantly for populations under 50 years, where the magnitude and direction of mortality disparities by race/ethnicity and citizenship differed substantially.

Mortality estimates for Latinx, Asian, and immigrant populations different substantially between the 2015 and 2019 Linked Mortality File, questioning the validity of research using these sources. Institutions conducting complex linkages should ensure that methodologic changes yield reliable information for marginalized populations, especially those less likely to report identifiers necessary for linkage.