Plan selection in Medicare Part D: evidence from administrative data

Abstract

We study the Medicare Part D prescription drug insurance program as a bellwether for designs of private, non-mandatory health insurance markets, focusing on the ability of consumers to evaluate and optimize their choices of plans. Our analysis of administrative data on medical claims in Medicare Part D suggests that fewer than 25% of individuals enroll in plans that are ex ante as good as the least cost plan specified by the Plan Finder tool made available to seniors by the Medicare administration, and that consumers on average have expected excess spending of about $300 per year, or about 15% of expected total out-of-pocket cost for drugs and Part D insurance. These numbers are hard to reconcile with decision costs alone; it appears that unless a sizeable fraction of consumers place large values on plan features other than cost, they are not optimizing effectively.

Keywords: Administrative data; C25; D12; H51; Health insurance demand; I11; I18; Insurance claims data; Medicare; Prescription drugs.

Figure 1

Cumulative distribution functions of annual total drug bills

Figure 2

Distribution of differences between actual and simulated OOP costs

Figure 3

Empirical means of 2007 simulated out-of-pocket (OOP) costs for alternative plan types and the Part D standard plan designed benefit schedule, conditioned on annual total drug bill

Figure 4

Cumulative distribution functions of 2008 total drug bill conditioned on 2007 total drug bill percentile

Figure 5

Mean and 95^th percentile of 2008 total drug bill, conditioned on 2007 total drug bill

Figure 6

Probability that standard plan enrollment in 2008 at a $30 per month premium gives lower consumer cost than non-enrollment, conditioned on total drug bill in 2007 (log scale)

Figure 7

2008 expected consumer inclusive cost (OOP plus premium) of “No Plan”, “Silver” (Standard) plan, and “Gold” (gap generic coverage) plan choices, given 2007 total drug bill

Figure 8

Figure 8a. Reduction in CIC associated with optimal choices implied by benchmark decision rules, relative to the chosen plan’s CIC (2008, no substitution) Figure 8b. Reduction in CIC associated with optimal choices implied by benchmark decision rules, relative to the chosen plan’s CIC (2008, with therapeutic substitution)