(F) Time Trends in Performance of Different Mortality Data Sources Compared to The National Death Index in The Healthcore Integrated Research Database

Publication Number: 664

Background: Ascertainment of mortality is critical to many non-interventional studies but is challenging due to the lack of systematic recording of mortality data in health claims or electronic medical records. The National Death Index (NDI) is considered a gold standard for mortality ascertainment in the US. However, linking real-world data sources to NDI data often involves considerable time and resources, and privacy restrictions prevent linkage to NDI data for some populations. While alternative sources may be used, their performance may vary over time.

Objectives: To investigate time trends in the performance of death data from the Social Security Death Master File (DMF), hospital discharge codes, health plan disenrollment, obituaries, and other sources compared to the NDI in the HealthCore Integrated Research DatabaseSM (HIRD).

Methods: This study is a secondary analysis of de-identified data originating from a standing cohort of individuals in the US with advanced cancer with an established NDI linkage in HIRD from January 2010 to December 2018. NDI early release data was obtained for 2018 and may not be complete. Mortality data sources, including inpatient discharge status, health plan disenrollment, DMF, and online obituary data, all updated monthly in the HIRD, were compared to the NDI. Death status from each source and the NDI was used to calculate sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and 95% confidence intervals (95% CI) for each year of the study.

Results: We identified death dates for 27,396 of 40,692 advanced cancer patients from 2010 to 2018 (NDI: 25,761; alternate sources: 24,612; overlap: 22,977). The sensitivity for the DMF declined from 0.77 (95% CI: 0.76, 0.79) in 2010 to 0.13 (95% CI: 0.11, 0.14) in 2018. In contrast, the sensitivity of obituary data rose from 0.12 (95% CI: 0.11, 0.13) in 2010 to 0.64 (95% CI: 0.62, 0.66) in 2018. Sensitivity was above 0.80 in each year throughout the study period for all death sources combined. Time trends in specificity, NPV, and PPV were consistent across death sources. NPV was above 0.97 each year for the composite source. Specificity was above 0.89 each year, while PPV remained above 0.97 from 2010 to 2016 before dropping to 0.75 (95%CI 0.72, 0.77) in 2018.

Conclusions: Our results demonstrate that the sensitivity of the mortality alternatives to the NDI can vary substantially over time. Combining death data from multiple sources may be necessary to achieve adequate performance for studies with longer follow-up periods.