OBJECTIVES: Insulin is a major cost driver in diabetes care. Many patients with diabetes may elect to purchase older ‘human’ insulin out of pocket (OOP) in order to avoid cost-sharing for the newer and more expensive ‘analog’ insulins. Dispenses for prescription medications paid entierly OOP do not generate insurance claims and therefore do not appear in traditional claims-based datasets, possibly causing underestimation of insulin use. We aim to quantify OOP insulin dispenses among the privately insured, Medicare and Medicaid beneficiaries, and the uninsured.

METHODS: We utilized data from years 2003 – 2015 of the Medical Expenditure Panel Survey (MEPS). Patients were included that completed all five rounds of MEPS including the pharmacy portion of the survey, reported insurance status, and had at least one dispensing of insulin. OOP dispenses were defined as those where total medication cost was equal to the amount paid OOP.

RESULTS: We identified 3,325 individuals with at least one insulin dispensing. Most (49%) were privately insured, followed by Medicare (16%), Medicaid (13%), those reporting both Medicare and Medicaid coverage (14%), and the uninsured (8%). 44.9% of patients had at least one OOP insulin dispensing, with the highest proportion of OOP insulin users among the uninsured (76%), followed by Medicare beneficiaries (51%), and privately insured (47%), compared to 34% of Medicaid beneficiaries. For analog insulin, between 3% and 12% of insured patients obtained over half of their prescriptions OOP, compared 8%-24% of insured patients for human insulin.

CONCLUSIONS: Insulin utilization, persistence, and adherence may be significantly underestimated in observational research using claims-based databases, particularly private insurers or Medicare, and particularly when human insulins are being studied. Particularly because some human insulins can be dispensed without a prescription, novel research methods may be needed to conduct rigorous pharmaco-epidemiology and pharmaco-economics on these widely used biologics.