OBJECTIVES: Enrollment in Medicare Part D is voluntary; however, mechanisms exist to encourage early enrollment and improve risk pooling, including a permanent premium penalty associated with delayed enrollment and restricted enrollment periods. This study examined whether a health shock would create adequate incentives to overcome the penalties associated with late enrollment. METHODS: Using enrollment and claims from a random 5% sample of Medicare beneficiaries from 2006 to 2008, we observed Part D enrollment decisions among beneficiaries who had failed to enroll in Part D at first eligibility (N=207,674). A health shock was defined as a hospital admission due to a drug-intensive chronic condition. Multivariable logistic regression examined the impact of a health shock on the probability of late Part D enrollment, controlling for beneficiary demographics, pre-existing chronic conditions, preventive service use, and admission to a facility. We also examined whether timing of the hospitalization relative to the next available enrollment period influenced the likelihood of Part D enrollment.  RESULTS: Eighteen percent of beneficiaries in the cohort enrolled late into Part D. Initial and subsequent hospitalizations for drug-intensive conditions were associated with 5 and 7 percentage point increases in the probability of Part D enrollment, respectively (p<0.01). A gap from the time of hospitalization to the next coverage period was associated with a lower likelihood of enrollment among non-Low Income Subsidy (LIS) recipients, but had no relationship for LIS enrollment, which is not restricted to enrollment periods. CONCLUSIONS: Health shocks were associated with an increased likelihood of late Part D enrollment, but many beneficiaries remained without Part D coverage despite deterioration in their health status and expected increased need for drugs.  Non-enrollees were forced to either absorb the full cost of medications or forgo them, which can have negative effects on health and, the potential to increase Parts A and B spending.