(178) Enhancing Data Continuity: Tokenized Linkage Extends Patient Observations for Real World Evidence Generation - A Komodo Research Dataset Study

Background: Claims databases in the US historically suffer from the known limitation of inability to track individuals across multiple databases and/or insurances over time due to privacy concerns about patient health information. Tokenization technology of person identifiers provides a possible solution.

Objectives: Using the Komodo Research Dataset as an example, this study assessed the impact of encrypted linkage of tokenized person identifiers on data capabilities of sample size and longitudinality relevant to pharmacoepidemiology studies.

Methods: This cross-sectional study used the Komodo Research Dataset, a new-generation database comprising medical and prescription drug insurance claims with person identifiers native to Datavant tokens, deduplicated, and linked. A database snapshot as of May 2024 was taken. Eligible members were selected based on payer-complete data sources and required to have at least one day of medical coverage and one day of drug coverage between January 2016 through December 2023. Demographics were determined on the first day of their most recent enrollment span with simultaneous medical and drug coverages (45-day gaps allowed). Summary of enrollment spans was reported at both the enrollment span and person levels.

Results: A total of 238,606,234 unique individuals (as represented by unique patient keys) were identified, with a mean age of 33.0 years, 51.9% female, and 34.0% geographically located in Atlanta or Dallas (Health and Human Services) regions. Among the 331,290,616 enrollment spans accrued, respectively 51.6%, 37.6%, and 20.0% were 2, 3, and 5 years or longer. At the person level, 71.0%, 57.2%, and 34.3% of the included individuals contributed to 2, 3, and 5 years or longer of cumulative enrollment (gap days excluded) in the database, whereas when restricted to time since the first related medical claim, 63.4%, 48.2%, and 24.5% of 121,874,799 endocrinology patients and 58.9%, 43.2%, and 19.8% of 31,134,002 oncology patients had the corresponding years of cumulative enrollment.

Conclusions: With over 238 million observable lives and a median cumulative enrollment longer than 3 years, the Komodo Research Dataset is characterized as a robust claims data source and owns competitive advantage in elongated follow-up availability when compared to national benchmark databases such as the FDA Sentinel System. These findings demonstrated how a claims database with tokenized linkage enhances the potential of health insurance data and enables extended views of the patient journey.