OBJECTIVES: The initiation of Medicare Part D in 2006 offers an ideal opportunity to study real-world decision-making and the role of adverse selection and other factors in insurance enrollment. Our objective was to identify predictors of Part D enrollment among individuals with a range of health conditions and insurance designs. METHODS: The sample included all individuals in both the 2005 and 2006 Medical Expenditure Panel Survey (MEPS) datasets who were enrolled in Medicare, but not Medicaid, as of December 2005. A multivariate logistic regression was used to assess the effects of sociodemographics, health status, 2005 supplemental insurance coverage, and 2005 person-level out-of-pocket (OOP) drug expenditures on the likelihood of enrolling in Part D in 2006. MEPS sample weights were used to calculate standard errors. RESULTS: Out of 1,436 persons who met inclusion criteria, 657 (45.4%) enrolled in Part D during 2006. Compared to the no-Part D group, the Part D group was slightly older, had more non-whites, rural residents, and unmarried individuals, and was slightly less educated and poorer. The Part D group had more beneficiaries with Medigap coverage only (17.2% vs. 5.7%), fewer with employer-based coverage only (18.3% vs. 37.1%), and more with no private supplemental insurance (46.1% vs. 32.1%). In multivariate analyses, significant positive predictors of Part D enrollment were having Medigap supplemental insurance only (OR: 1.99; 95% CI: 1.25-3.19) and OOP drug expenditures ≥$2500 in 2005 (OR: 1.58; 95% CI: 1.03-2.41). Most beneficiaries with employer-based coverage in 2005 maintained that coverage in 2006 (91.8%). CONCLUSIONS: Based on first-year data, fears of only the sickest beneficiaries enrolling in Part D and employers withdrawing drug benefits to retirees seem to have been unwarranted. Existing coverage and high prior drug spending drove the decision to enroll in Part D in what appears to have been a rational way.