Evaluating Consumer Preferences for Medicare Part D Using Conjoint Analysis

Abstract

The establishment of an outpatient prescription drug benefit in Medicare, Medicare Part D, was one of the more significant events in United States healthcare reform history. Many seniors have chosen to enroll in the plan as the program has an enrollment of over 27 million. One central premise of Medicare Part D was that the plan would be administered entirely through private insurance plans. Because many plans would be competing against one another for potential beneficiaries, it was expected that companies would offer seniors plans with high quality benefits at lower costs that would presumably be attractive to seniors. However, in order for this strategy to succeed, seniors must be sensitive to differences in price and the quality of benefits amongst different plans. This study sought to evaluate the relative importance seniors place on different attributes of Medicare Part D drug plans using a marketing technique called conjoint analysis. An Adaptive Choice Based Conjoint Analysis Survey was constructed and administered to 497 seniors in the Greater Memphis/Shelby County Area. Seniors were recruited from local senior centers or senior apartment homes. The survey consisted of 7 attributes: premiums, brand copayments, generic copayments, deductible amounts, doughnut hole coverage, formulary provisions, and eligibility requirements for medication therapy management services. The most important attribute in the aggregate sample was premiums with an importance score of 50.3%, followed by the formulary provisions (importance score of 12.1%), deductible (11.6%) and pharmacy access (10.1%). Seniors with higher incomes, those taking more than seven medications, and those with monthly medication costs above $80 were less sensitive to premium increases than other groups. As a group, the seniors were willing to pay approximately $20 more in monthly premiums to go from plans with the maximum allowed deductible to plans with no deductible and only $3 more to have full coverage of generics in the doughnut hole; however these values differed amongst different segments. For example, seniors with reported medication costs of at least $80 per month are willing to spend on average $20 per month for doughnut hole coverage. Adaptive Choice Based Conjoint Analysis represents a feasible way to elicit preferences for features in Medicare Part D drug plans. Seniors are sensitive to differences in cost and breadth of coverage across plans, although only a subset of seniors is sensitive to the doughnut hole. Seniors are particularly concerned with low premiums in their particular plans, although seniors with higher incomes, those taking more medications, and those with higher out of pocket medication costs are less sensitive to changes in premiums. The delivery system characteristics are also an important consideration for seniors in purchasing Medicare Part D plans, as most seniors exhibited strong preferences for being able to retain their relationship with their current pharmacy. vi TABLE OF CONTENTS CHAPTER 1.

Bayesian evaluation of energy dependent neutron induced fission yields*

“…This distribution answers predictive queries about unseen data using expectations. Each possible configuration of the weights, weighted according to the posterior distribution, makes a prediction about the unknown label given the test data item x [63].…”

Developing mobile services

“…This becomes relevant since, due to the lack of research on mobile application characteristics in a tourism setting, there is no "a priori" knowledge regarding preferences. A significant difference between attributes only exists if the upper confidence interval (CI) for one attribute is greater than the lower CI for another attribute (Manalo, 1990;Wingate, 2011). This means that any attribute comparison with overlapping values in their CI will be considered not statistically different.…”

“…H2 is designed to infer whether the levels of a particular attribute are statistically significant. The zero-centered part-worth estimates are examined at the 2.5 percentile and the 97.5 percentile to obtain a 95 per cent CI and determine whether they were statistically significantly different from zero (Wingate, 2011). If the CI values for a specific level cross zero, then they are not assumed to be significant.…”