Objectives
The purpose of the proposed study is to assess how variation in Part D benefit design (cost sharing and formulary provisions) influences use of prescription drugs, other health care services, and clinical and economic outcomes for enrollees with chronic diseases enrolled in a Medicare Health Support program.
Research Questions
We have two primary hypotheses, each with several subhypotheses.
- Medicare Health Support program enrollees in worse health status will systematically select more generous Part D plans, but the sorting will not be “perfect.”
- Medicare Health Support program enrollees who have selected Part D plans with high levels of cost sharing will be less adherent to their medications and have worse health outcomes.
Study Design
The study will be based on the experiences of the 24,325 enrollees in a Medicare Health Support program. We will analyze the subset of about 8,000 who have chosen Part D plans. The Health Support program has gathered all utilization and outcome variables necessary for the analysis, and we will be able to merge with that data we gather from PDP plans. These data will allow us to relate plan design to important economic and clinical outcomes.
We will use several dependent variables in our analysis. They include measures of utilization such as prescription drug utilization (measured my medical possession ratios) and utilization of other health care services (including hospitalizations, nursing home admission, and emergency room use). The latter utilization measures capture important clinical events and have important economic consequences. We will also measure total medical spending, which will be disaggregated by type of spending (pharmaceutical, inpatient, outpatient, physician, nursing home, and home care spending). Medical spending will be analyzed in aggregate and separately by beneficiary out-of-pocket spending and Medicare program spending (by parts A, B, and D). In addition, we will analyze clinical outcomes including intermediate health outcomes (e.g., blood pressure, glucose control, and cholesterol control), process measures (e.g., annual glucose and cholesterol testing, flu and pneumococcal vaccination), serious cardiac events such as acute myocardial infarction, quality of life (e.g., self-report of pain and health status), and mortality. We will define appropriate clinical measures separately for CHF and diabetes patients and analyze them separately, recognizing some beneficiaries will be in both samples.
The primary independent variables will be measures related to the structure of the Part D plan the beneficiary chose. We will construct measures of formulary restrictiveness and financial generosity.
