The Effect of Formulary Benefit Design on Utilization of Prescription Opioids by Disabled Medicare Part D Beneficiaries

Tuesday, June 14, 2016
Lobby (Annenberg Center)

Author(s): Erin A. Taylor; Andrew Mulcahy; Spencer Case; Rosalie Malsberger; Rosalie L. Pacula; Janice Blanchard

Discussant: Abby Alpert

The overuse of prescription opioids has been declared an epidemic by the Centers for Disease Control (CDC). One potential tool to address overuse of opioids is the design of prescription drug plan formularies. This design can take on three primary aspects: 1) coverage of opioids and their alternatives; 2) when covered, the amount of cost sharing charged by the plan; and 3) the use of utilization management tools, such as prior authorization, in an attempt to manage and control utilization of these drugs. Our paper examines the effect of these three aspects of Medicare Part D formulary benefit design on the utilization of prescription opioids by the disabled Medicare population.

We use Medicare prescription drug claims data from 2010 to 2012, for all Medicare beneficiaries eligible by reason of disability and living in California and Texas. We also use the complete formulary files for each plan in order to construct a measure of opioid coverage, tier placement, and utilization management tools for each plan. Finally, we merge the Part D claims data with medical claims in order to control for other variation in utilization due to differences in health status (for example, risk adjustment).

We assess two measures of utilization as the dependent variables of interest: any utilization, which measures whether or not a beneficiary filled at least one prescription for an opioid in the given year, and the number of 30-day normalized fills in a given year. Our independent variables include: opioid coverage (whether or not a plan covered the opioid, percent of opioids covered); the amount a beneficiary had to pay in cost sharing for the opioid; and the percent of opioids for the plan that had utilization management tools applied. For models looking at any utilization, we employ a logistic regression approach with fixed effects at the plan level. For the number of fills, we use a count data model with fixed effects.

We find that disabled Medicare beneficiaries enrolled in stand-alone Part D plans are less likely to use prescription opioids when cost sharing is higher.  Moreover, we also find that utilization management tools act to reduce utilization, but the overall effects are difficult to discern given that different plans define and apply utilization management tools differently and we are not able to measure this specifically in our data.  Estimates of the elasticity of demand with respect to cost sharing are provided and compared to findings obtained from another disabled population: those receiving care through Workers Compensation.