OBJECTIVES: Most healthcare research, including oncology studies, requires accurate vitality/mortality information. This retrospective, observational cohort study examined mortality/vitality completeness across multiple real-world sources among patients with high mortality burden conditions, glioblastoma multiforme (GBM) and metastatic pancreatic cancer (mPC).

METHODS: Adult patients diagnosed with GBM or mPC between 1/1/19–12/31/19 were followed from diagnosis through 8/31/23. Mortality data were sourced from electronic health records, US Oncology’s Financial Data Warehouse (FDW), the Limited Access Death Master File, commercially-available mortality dataset linked by Datavant tokens, and the National Death Index (NDI). Vitality status was confirmed on 1/15/23 through chart abstraction. Overall survival (OS) with and without NDI data was estimated with Kaplan-Meier methods; censoring patients without a death record on their last contact date.

RESULTS: 173 patients with GBM (median age 65 years, median 12.0 months follow-up) and 990 with mPC (median age 68 years, median 7.1 months follow-up) were included. In total, 157 (90.8%) and 929 (93.8%) of patients with GBM and mPC, respectively, had a death record in at least one source. Of the 77 patients without death records in any source, 46 patients (59.7%) were confirmed alive through abstraction (97.3% overall mortality-vitality capture). Among the 852 patients with death records from multiple sources, 844 (99.1%) had the same month-year across sources. With only structured EHR, FDW, and commercial sources, 993 (91.4%) of all death records (n=1,086) were captured. Among GBM patients, median OS was 13.7 months (95% confidence interval [CI] 11.4-14.7) and 13.5 months (95% CI 10.8-14.4) with and without NDI data, respectively.

CONCLUSIONS: Mortality data completeness across all data sources was over 90% in this study, with strong agreement across sources. As 91% of death records were captured without chart abstraction or the NDI, survival estimation without these sources can be a fit-for-purpose and scalable option for oncology research.