Presentation (CTI)

OBJECTIVES: Federated data networks have arisen as a solution to sample size and privacy challenges in observational studies. Yet, their distributed structure suffers from inability to track individuals across multiple databases, or insurance plans in the case of claims, over time. Tokenization is an advanced method to address these issues. Using a next-generation database as example, this study assessed the impact of tokenized linkage on data capabilities of sample size and longitudinality.
METHODS: This retrospective cohort study examined the Komodo Research Dataset, a claims database with person-level linkage across sources enabled by encrypted, tokenized identifiers. Eligible members must have had payer information and ≥1 enrollment day in medical and drug plans between January 2016 through October 2024. The analyses summarized distributions of the continuous enrollment spans separately with and without tokenized linkage upon payer change and of the extensions at both person and span levels.
RESULTS: Of 197,556,385 individuals identified, the mean age was 35.0 years, 52.2% were female, and 50.1% were geographically located in Atlanta, Chicago or Dallas regions. After cross-payer linkage, the number of accrued enrollment spans decreased by 10.6% from 316,099,620 to 282,445,948 and the median length of these spans increased by 55.8% from 489 to 762 days. Personally, the length of enrollment spans on average increased 554 days from the shortest single-payer span to the longest mixed-payer span. Sensitivity analysis suggested a 23.8% decrease in the number of spans and 136.9% increase in their median length, when enrollment gap allowance widened from 45 days to infinite.
CONCLUSIONS: This is one of the first evaluations to quantify the effect of tokenized linkage on a claims database. Comprising observable lives of hundred millions and their elongated time in database up to 2.4 times, the Komodo Research Dataset demonstrates robustness and evidentially offers competitive advantages of scale and continuity among real-world data sources.