Vital status ascertainment in cancerlinq discovery (CLQD): Improvement in mortality capture with a supplemental data source.

Abstract

7064 Background: Overall survival (OS) is the gold standard outcome in clinical cancer research but many clinical trials cannot assess long-term OS. Real-world data sources can be used to calculate long-term OS, but only if vital status is accurately captured. Methods: The primary goal was to assess concordance of death dates from CLQD and an external source, and the effect of incorporation of external death data on estimates of OS. CLQD obtains electronic medical record (EMR) data from participating US oncology sites. Under-reporting of vital status is a common problem with EMR data; we investigated the value of including death data from a commercially available database, obituarydata.com (OBD) into CLQD. OBD pulls death data from published obituaries across the US. A matching algorithm is used to match patients in CLQD and OBD. OBD death data from breast, lung, ovarian, and pancreatic cancer patients diagnosed between 2010–2018 supplemented the CLQD in this study. OS was calculated using Kaplan-Meier estimation; Pearson correlation was used for comparing time to death. Results: The addition of OBD modestly changes OS estimates (see Table). Among a subset of patients with death dates in both CQLD and OBD, dates were highly correlated for breast (r = 0.98), lung (r = 0.93), ovarian (r = 0.99), and pancreatic (r = 0.88) cancers. When date differences existed, they were ≤10 days for > 95% of the patients. These results suggest death dates are reliable in CLQD EMRs. OS curves were as expected, with OS decreasing by stage and age at diagnosis. Conclusions: Incorporating OBD modestly improves OS estimates and shows that when death data is present in CLQD, it is reliable. Future enhancements will focus on improving sensitivity of mortality ascertainment with external data linkages, without compromising specificity. [Table: see text]