The longer, the better: How tokenized linkage can extend observable time for cancer patients in health insurance claims databases.

e23349 Background: Claims databases have become a key to longitudinality in real-world evidence generation for US-based oncology research by extending patient journey views on diagnosis, progression, treatment patterns, supplemental endpoints, and economic burden. Yet, most claims databases are payer-segmented and lack the ability to track patients across insurances over time due to privacy concerns. Innovative technology such as tokenization is one solution. Using a new-generation claims database as the example, this study evaluated the impact of tokenized linkage on data capabilities of observable time among cancer patients. Methods: This retrospective cohort study assessed the Komodo Research Dataset, a health insurance claims database with patient-level linkage across data sources enabled by tokenized, encrypted identifiers. Eligible members contributed claims from payer-complete sources, had ≥1 payer recorded, had ≥1 day enrollment in medical and prescription drug plans plus ≥1 diagnosis of breast, prostate, or lung cancer between Jan 2016 through Oct 2024. Stratified by each cancer of interest, distributions of enrollment spans were summarized separately with and without tokenized linkage upon payer change. The continuous period during which a member had both medical and drug coverage defined an enrollment span (all gaps bridged). Results: Before cross-payer linkage, the median length of enrollment spans for patients respectively diagnosed with breast, prostate, and lung cancers was 1,096, 1,096, and 912 days in all spans and increased to 1,216, 1,237, 1,025 when restricted to spans starting from the first qualifying diagnosis. After linkage, the number of unique individuals consolidated by 26.5% to 1,847,008, 24.0% to 1,425,734, and 23.4% to 770,996, whereas the median length of enrollment spans increased by 102.9% to 2,224 days, 99.9% to 2,191 days, and 87.1% to 1,706 days. Sensitivity analyses with 45-day gap allowance suggested attenuated median length increase by 55.2%, 59.0%, and 47.5% in all spans and 62.8%, 54.9%, and 49.9% in spans since the first diagnosis. Conclusions: With recorded cancer populations of millions and up to twice elongated follow-up duration, the Komodo Research Dataset demonstrated how tokenized linkage enhances observable time for cancer patients in a claims database. Such robustness offers competitive advantages of scale and continuity among real-world data sources and can play an important role in advancing clinical knowledge for oncology.