Comparison of real-world mortality data among patients with glioblastoma (GBM) and metastatic pancreatic cancer (mPC) treated in the community oncology setting.

Abstract

e18805 Background: Evaluation of endpoints in oncology research requires accurate assessment of patients’ vital status. For real-world investigations of cancer populations, comprehensive capture of patients’ death dates may require several data sources. The intent of this evaluation was to assess capture of documented death dates across multiple datasets in the US community oncology setting. Methods: This was a retrospective observational study of adult patients with GBM or mPC who initiated treatment between 1 January 2019 and 31 December 2019 within The US Oncology Network. Patient demographic, clinical and treatment characteristics were sourced from an electronic health record (EHR) database, iKnowMed (iKM). Patients were followed until January 2023 with death dates captured from the iKM EHR (structured and unstructured fields), the iKM claims database, the Social Security Administration’s Death Master File, as well as a commercially available mortality dataset from Datavant comprised of obituary, claims and government records. Chart review was used to confirm if patients were alive on 15 January 2023. The overall capture rate of death information from each source was evaluated descriptively. Agreement was assessed using two approaches: 1) as death dates that were an exact match between sources and 2) as death dates that occurred within 30 days of each other and/or within the same month. Results: A total of 173 patients with GBM (median age 65 years, median follow-up time 12.0 months) and 990 patients with mPC (median age 68 years, median follow-up time 7.1 months) were included. Across all sources, 91.1% of the pooled population had at least one documented death date or were confirmed to be alive (85.5% and 92.1% of patients with GBM and mPC, respectively). Among patients with full day-month-year death dates from at least 2 sources (n = 835), 95.4% of death records were matched on the exact death dates. Among patients with any death record from at least 2 sources (n = 852), 99.3% of death records were matched within 30 days of each other or within the same month. Conclusions: The findings of this study describe the capture rate of mortality and vitality status in the US community oncology setting. Future analyses are planned to augment these assessments of concordance, sensitivity, specificity, positive predictive value and negative predictive value across the sources with comparison to death dates captured by the National Death Index. These insights are important for development of composite mortality databases, comprised of death information from multiple sources using appropriate tokenization and overlap methodologies.