(C02) Anonymized tokenization technology to link real world data and assess the potential for exposure misclassification for pharmacoepidemiology research
Director, RWE Scientist Pfizer New York, United States
Background: While administrative claims data are widely used to conduct comparative effectiveness analysis, in the United States eligible patients may also receive treatment via unpaid mechanisms such as patient assistance programs (PAP). This study leveraged tokenization technology to link Centers for Medicare & Medicaid Services (CMS) Medicare fee-for-service (FFS) claims with PAP data among users of tafamidis, a medicine indicated for treating transthyretin amyloid cardiomyopathy a rare cardiovascular disorder. This linkage enabled access to paid and unpaid utilization which provided more complete prescription journey for each included patient.
Objectives: To assess the potential for exposure misclassification when identifying tafamidis users based on Medicare FFS claims.
Methods: Patient data in the CMS Medicare FFS and PAP data was tokenized using tokenization software from a third party to protect patient information which allowed linkage between Medicare FFS beneficiaries and PAP program participations. The tokenization algorithm leveraged different combinations of personally identifiable information elements inclusive of first and last name, date of birth, gender, and address to create unique, irreversible tokens. Elderly adults (≥65 years old) who received ≥2 tafamidis prescription claims between 03 May 2019 and 31 December 2021 either via Medicare FFS or PAP were included in the study. The date of the first prescription was defined as the index date, with follow-up time defined as the time between the first and last prescription plus the days’ supply from the last refill.
Results: There were 8,131 PAP patients matched to Medicare FFS beneficiaries based on 1 token, 7,968 based on 2 tokens, and the number of patients matched stabilized after matching on 8 tokens (n=6,769). Among 7,554 patients who matched on at least 8 tokens and met the inclusion criteria, 51% received tafamidis through Medicare FFS only, 35% through PAP only, and 14% through both Medicare FFS and PAP. Overall, mean age was 80 years old (SD: 6 years) and over 80% were male. Mean time to discontinuation with a 60-day gap was 217 days (SD: 185 days) among those who received therapy through Medicare FFS only, 197 days (SD: 151 days) for PAP only, and 234 days (SD: 170 days) for both Medicare FFS and PAP.
Conclusions: While all patients were enrolled in Medicare FFS, more than 1 in 3 tafamidis users solely received prescriptions via the PAP and more than 1 in 8 patients received prescriptions through both Medicare FFS and PAP. Given this, when conducting a comparative effectiveness study of tafamidis users, only relying on insurance claims may be subject to substantial potential selection bias into nontreatment group and exposure misclassification.
