The US Social Security Administration Death Master File (SSDMF) appears to undercapture mortality data, with varying rates of undercapture by states, which over time could lead to artifactual temporal and geographic differences in mortality rates that should not be relied upon by clinical researchers, according to a study published in JAMA Cardiology.
Although the SSDMF’s undercapture of mortality data has been documented, and the Social Security Administration itself has noted that these records are not meant to include all deaths in the United States, researchers still often use the SSDMF to provide mortality end points in retrospective clinical studies. Furthermore, changes in the reporting of death data in 2011 likely further affected the reliability of the SSDMF.
This observational analysis was designed to assess the reliability of SSDMF mortality rates using commercial insurance and IBM MarketScan Medicare databases linked to information from the SSDMF in a cohort of patients with atherosclerotic cardiovascular disease (ASCVD). Participants from states with at least 1000 eligible participants were adults with ASCVD, had a clinical encounter between January 2012 and December 2013, and had 2 or more years of follow-up.
State-level mortality rates, stratified by database (Medicare or commercial), were estimated with Kaplan-Meier analyses. Individual Kaplan-Meier curves for temporal changes were evaluated by state, and state-by-state constant hazards of mortality were tested. For those states with a hazard of death that was constant over time, the mortality rates of participants were compared with age-group-specific, state-level, overall mortality rates in 2012, using data from the National Center for Health Statistics (NCHS).
The study analyzed data on a total of 667,516 adults with ASCVD, including 393,511 from the Medicare database — 245,366 male (62.4%) and median (interquartile range [IQR]) age of 76 (70-83) years — and 274,005 adults from the commercial insurance database (171,959 male [62.8%] and median [IQR] age of 58 [52-62] years).
Of the 41 included states, 43.9% (n=18) in the Medicare cohort and 26.8% (n=11) in the commercial cohort had a hazard of death change after 2012. Using the SSDMF, state-level mortality rates among states with constant hazard ranged widely, from 0.83 to 6.07 per 100 person-years (Medicare cohort), and from 0.06 to 1.30 per 100 person-years (commercial cohort). The between-state variability of mortality estimates using SSDMF data on adults with ASCVD greatly exceeded the overall mortality variability using data from the NCHS. No correlation between NCHS mortality estimates and those from the SSDMF was found (ρ =.29 [P =.06] for age 55-64 years; ρ =.18 [P =.27] for age 65-74 years).
Study investigators conclude, “The SSDMF appears to undercapture mortality, with variable undercapture by state and over time, which may lead to artifactual geographic and temporal differences in mortality rates. Researchers should avoid relying on mortality estimates based on the SSDMF alone and be aware of heterogeneity in SSDMF data completeness.”
Disclosure: Regeneron and Sanofi funded this study. Multiple authors disclosed affiliations with pharmaceutical companies. See the reference for complete disclosure information.
Navar AM, Peterson ED, Steen DL, et al. Evaluation of mortality data from the Social Security Administration Death Master File for Clinical Research [published online March 6, 2019]. JAMA Cardiol. doi:10.1001/jamacardio.2019.0198