Assessing the Completeness of Oncology Treatment Data from Administrative Claims: A Benchmarking Study Against Abstracted EHRs Using Patient-Level Linkages
Author(s)
Joshuah Kapilivsky, BS1, Emma Roth, BS1, Zachary Rivers, PharmD, PhD1, Adam J. Hockenberry, PhD1, Sandeep Jain, MD2, Jeremy L. Warner, MD, MS2, Emilie Scherrer, MSc1, Chithra Sangli, MA1;
1Tempus AI, Inc., Chicago, IL, USA, 2Brown University, Providence, RI, USA
1Tempus AI, Inc., Chicago, IL, USA, 2Brown University, Providence, RI, USA
Presentation Documents
OBJECTIVES: Integrating real-world data (RWD) sources can enable increasingly granular precision oncology studies. Comparisons between data sources typically report aggregate statistics from unlinked datasets, but this approach precludes analysis of patient-specific date agreements. In this study, we leverage deterministic patient linkages to benchmark claims oncology treatment data against abstracted electronic health records (EHR) in a time-aware manner.
METHODS: We extracted abstracted EHRs from a clinico-genomic database for 6487 stage 4 lung adenocarcinoma patients diagnosed between 2020 and 2023. Claims data (open and closed) were linked using de-identified patient tokens. Claims between patients’ first and last abstracted treatment dates were selected. Abstracted data were considered ground truth: claims for the same medication between abstracted start and end dates were true positives, unmatched claims false positives, and unmatched abstracted treatments false negatives. Metrics are reported as ranges across 13 infusional and 3 oral medications.
RESULTS: Closed claims enrollment periods showed greater sensitivities (50.0-95.3%) than open claims (14.3-54.8%). Sensitivities differed by route of administration, with infusions (closed: 76.5-95.3%; open: 32.4-54.8%) higher than orals (closed: 50.0-76.2%; open: 14.3-34.1%). Regardless of enrollment, positive predictive values (PPVs) were high for infusions (closed: 79.1-98.3%; open: 61.5-99.1%) and orals (closed: 84.5-94.2%; open: 91.8-96.8%). During closed claims enrollment, 45.5-82.5% of abstracted infusion start dates had exact date matches in claims, and 27.6-65.9% of abstracted oral start dates had matching claims within 7 days (accommodating prescription fill delays).
CONCLUSIONS: While EHR remains the gold standard, the PPVs for open and closed claims indicate that individual claims may be sufficient to identify patients receiving a specific treatment for an RWD study. The sensitivities and start date match rates suggest closed claims may be suitable for constructing comprehensive treatment journeys.
METHODS: We extracted abstracted EHRs from a clinico-genomic database for 6487 stage 4 lung adenocarcinoma patients diagnosed between 2020 and 2023. Claims data (open and closed) were linked using de-identified patient tokens. Claims between patients’ first and last abstracted treatment dates were selected. Abstracted data were considered ground truth: claims for the same medication between abstracted start and end dates were true positives, unmatched claims false positives, and unmatched abstracted treatments false negatives. Metrics are reported as ranges across 13 infusional and 3 oral medications.
RESULTS: Closed claims enrollment periods showed greater sensitivities (50.0-95.3%) than open claims (14.3-54.8%). Sensitivities differed by route of administration, with infusions (closed: 76.5-95.3%; open: 32.4-54.8%) higher than orals (closed: 50.0-76.2%; open: 14.3-34.1%). Regardless of enrollment, positive predictive values (PPVs) were high for infusions (closed: 79.1-98.3%; open: 61.5-99.1%) and orals (closed: 84.5-94.2%; open: 91.8-96.8%). During closed claims enrollment, 45.5-82.5% of abstracted infusion start dates had exact date matches in claims, and 27.6-65.9% of abstracted oral start dates had matching claims within 7 days (accommodating prescription fill delays).
CONCLUSIONS: While EHR remains the gold standard, the PPVs for open and closed claims indicate that individual claims may be sufficient to identify patients receiving a specific treatment for an RWD study. The sensitivities and start date match rates suggest closed claims may be suitable for constructing comprehensive treatment journeys.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
SA56
Topic
Study Approaches
Disease
SDC: Oncology, STA: Personalized & Precision Medicine