On the Evaluation of Drug Benefits Policy Changes with Longitudinal Claims Data: Concepts of Income-Based Deductibles and Issues Related to Causal Inference

Cost containment in pharmaceutical benefits plans is often controversial because of its potential for unintended effects on health and overall expenditures. Thorough evaluations are needed to support benefits plans in developing evidence-based policies. Evaluations of income-based deductibles for prescription drugs using non-experimental data are challenging because the time until patients' expenses exceed their annual deductible varies with the number of drugs used, which depends on their health status and income. We sought to clarify conceptual and methodological issues in longitudinal designs to evaluate income-based deductibles for prescription drugs, and provide a framework for the evaluation of such policies using longitudinal claims data. One of the main challenges is the violation of the Exclusion Restriction Assumption, which is a standard assumption when using instrumental variables such as the start date of a new policy. This assumption requires the instrumental variable to be independent of subject characteristics, such as income and health status. The validity of an evaluation therefore depends on the ability to measure and adjust for these characteristics. Given the lack of such data, randomization of the policy may be a more valid study design.