Objective: Little is known about how Medicare Part D utilization varies based on subjects' pre-Part D prescription coverage and comorbidities. Methods: We examined claims from a national pharmacy chain from 2005 and 2006 accounting for approximately 15% of the U.S. prescription drug market. We focused on beneficiaries ages 66-79 as of January 1, 2006. We focused on the association between pre-Part D insurance generosity, comorbid burden using a method of risk adjustment based on pharmacy claims, and Part D uptake after adjusting for patients' demographic characteristics, characteristics of pre-Part D prescription use (average 2005 annual copayments, number of prescriptions, drug utilization in pill-days, number of treatment classes as reported in the store data), pre-Part D Medicaid eligibility, and zip code linked USA Census data. Results: After multivariate adjustment, each 10% increase in 2005 insurance generosity was associated with a 16% lower likelihood of Part D utilization and each standard deviation increase in comorbid score was associated with an 8% greater likelihood of Part D utilization. Beneficiaries with the lowest yet positive insurance generosity with the greatest comorbid burden were up to 2.5 times more likely to utilize Part D than people who had generous insurance and were the healthiest before the introduction of the program (60.65% vs. 37.56%, p<0.001). Results were robust to numerous sensitivity analyses. Conclusion: We find significantly higher rates of benefit adoption among subjects with low pre-Part D prescription coverage and high pre-Part D comorbid burden. Our findings highlight the fact that implementation and uptake of changes in health policy are seldom uniform. These results also may be useful in welfare analyses of Part D, and they demonstrate the importance of considering non-random selection into Part D when considering the impact of Part D on processes or outcomes of care.