The Medicare Part D prescription drug program was implemented to improve access to medications among all Medicare enrollees. Achievement of these goals was mixed in the NH setting. In this analysis of over 800,000 long-staying NH residents, we found a modest decrease of 3 percentage points (11% in 2005 vs. 8% in 2006, p<.05) in the proportion of prescription drugs paid for out-of-pocket. This finding compares to the 9%–13% relative out-of-pocket savings due to Part D observed in the community-setting.(16
) This difference may have been due to the high proportion of NH residents with Medicaid, since that program generally requires little or no cost-sharing for medications. NH residents without drug coverage prior to Part D experienced the largest relief in drug costs, although they were a relatively small group (<3% of the study population.
The Part D program did not appear to have expanded drug coverage in the NH setting, unlike in the community-setting where rates of having no drug coverage dropped from 30% to 10% between 2005 and 2006.(17
) Instead, the percentage of NH residents without any drug coverage remained at a stable 3% in the 1 year before and after Part D. Part D substituted for former sources of drug coverage, primarily Medicaid. It is possible, though, that temporary drug payment assistance, such as Medicare Part A, may have masked expansions in drug coverage.
The NH residents who did not enroll in Part D exhibited characteristics of some of the most vulnerable segment of the NH population, namely the oldest old (aged 85+), those without prior drug coverage, and those with the highest comorbidity burden. These findings are unique from those in the community-setting where Part D non-enrollers tended to be in better health status than the enrollers.(17
) By the end of 2006, approximately 11,000 Medicare enrollees in NHs still had no drug coverage. Our analysis cannot explain this finding but it is troubling and suggests the need for facilitated Part D enrollment for all Medicare-eligible NH residents who have no evidence of other drug coverage.
Overall patterns of drug use in NHs were disrupted in the early months of Part D; initiation of the new program coincided with a statistically significant decrease in average monthly prescription use per resident of half a prescription relative to 2005 levels (10.1 vs. 9.6, p<.003). The reduction appears to have been temporary as average drug use gradually returned to 2005 levels by December of 2006. This finding is also different from the experience of Medicare enrollees in the community where, overall drug use increased slightly (1.1%) after Part D.(16
) Our findings of a modest but temporary reduction in drug use in NHs are somewhat substantiated by a small survey of NH stakeholders (n=31) who reported that Part D increased the drug management processes in the NH setting and shifted drug use within therapeutic classes.(11
) Part D plans often employed prior authorization, step therapy protocols, tiered formularies, and quantity limits. Furthermore, one long-term care pharmacy provider indicated a need to be more aggressive in seeking reimbursement for dispensed drugs as the magnitude of drug claims rejected by Part D plans grew.(11
) However, this same group also reported no overt changes in gross drug use in NHs following the implementation of Part D. It is yet to be determined if this temporary disruption in medication use caused any untoward consequences to the health of the NH population.
Lastly, the Part D program decreased but did not eliminate Medicaid’s burden in paying for the medications of dual-eligibles in NHs. The fact that Medicaid paid for 12% of the medications dispensed in 2006 to our study population indicates the scope of gaps in Part D coverage. Medicaid wrap-around coverage varies by state but generally covers medications rejected from coverage by the Part D provider or excluded by Medicare (e.g., benzodiazepines or prescription vitamins.)
It should be noted that our analyses were limited to NH residents whose prescriptions were dispensed by 1 long-term care pharmacy provider and these results may not be nationally representative of all Medicare enrollees in NHs. However, a comparison of the geographic residence of our study sample to that of the nursing home residents in the December 2006 CMS OSCAR Data survey shows a similar distribution (Northeast: 24% vs. 23%; Midwest 36% vs. 29%; South 28% vs. 34%, and West 11% vs. 14%).(20
) Our study represents one of the first analyses of the effects of Part D on NH residents, and it reflects the observed experiences of nearly half of the entire Medicare population living in NHs.
Our approach has several strengths. First, we conducted these times-series analyses of changes in costs and drug utilization using an intention-to-treat study design, without regard for enrollment into Part D. This meant that we avoided introducing the selection bias inherent in comparing changes in drug use between individuals who selected into Part D and those who did not, a confounding potentially affecting the estimates of the non-Medicaid NH population. In addition, time-series analyses are robust to many of the threats to the validity of weaker observational designs, particularly in unmeasured changes in the composition of the study population or in survivor bias of cohort studies requiring a long observation in a NH population.(14
Limitations of the study include the following. First, our dataset was quite limited in a number of potentially important characteristics of the nursing home residents; we did not have information on race, socio-economic status or education. Thus, we could not ascertain if certain NH populations experienced more difficulty in enrolling in Part D or more disruptions in drug use than other groups. The comorbidity scores were based on only medication use but these are validated and robust measures and the recommended approach when the pharmacy claims data are the only source of information on health status.(21
) Secondly our times series were short, and far less than the 100 data points recommended to rule out seasonal effects, although this possibility is unlikely given the visual pattern of data.(14
) Short time series are valuable though, especially in helping to specify the degree of the immediate impact and the persistence of any delayed impact, which is difficult or impossible to do with other study designs.(22