Does Formulary Benefit Design Affect Opioid Use and Misuse for Disabled Medicare Beneficiaries?

According to the CDC, health care providers wrote nearly a quarter of a billion prescriptions for opioid analgesics in 2013, enough to provide a full bottle of pills to every adult in the U.S. (CDC 2017). The problem of overprescribing and overuse of opioids has been a major focus of state and federal policy makers, concerned about the opioid epidemic, but tools are available to private insurers to tackle this problem as well. In this paper, we consider a particular tool that has been used to address overutilization of other pharmaceuticals, namely prescription drug benefit design. Specifically, we consider the extent to which the amount of cost sharing (determined by placement of opioids in particular tiers) and use of utilization management tools within the Medicare Part D plan formularies are associated with prescribing and/or misuse.

We use Medicare prescription drug claims data from 2010-2012 for all disabled Medicare beneficiaries living in California and Texas, and link them to the formulary files for each plan to construct plan-level measures of opioid cost sharing and use of utilization management tools. We also control for other patient- and plan-level factors associated with use, such as age, gender, and plan premium. We focus our analyses on beneficiaries who are never dually eligible for Medicaid, as we would expect this group to respond less to changes in plan cost sharing since they face low, fixed cost sharing that does not vary by plan.

Using Ordinary Least Squares (OLS) regressions incorporating plan level fixed effects, we examine the impact of cost sharing and a variety of prescription drug utilization management tools on two primary outcomes: any utilization of opioids, and the TROUP score, which measures the likelihood a patient misused an opioid during the given year. We consider models grouping all opioids in a given plan as well as those that examine differences across different types of opioids.

We find that beneficiaries in our two state sample are not responsive to higher opioid cost sharing; for all opioids, an increase of $10 in the cost share is associated with a 1.6% decrease in the likelihood of utilization. This is statistically significant but small in magnitude. We find similar effects across most opioid types. Responsiveness to utilization management tools was generally insignificant across all of the models as well. Our findings are similar for the TROUP score, where we see no consistently significant effect of opioid cost sharing on likelihood of misuse. Our findings suggest that the disabled Medicare population are generally unresponsive to standard formulary design tools, especially utilization management tools, at least within the ranges tried by plans in our data. Medicare Part D plans may need to focus on provider-oriented incentives to reduce excessive use among the Medicare disabled rather than traditional patient-oriented tools.