The implementation of Medicare's drug benefit (Part D) in 2006 has important implications for mental health care. Approximately one-quarter of seniors and one-half of disabled Medicare beneficiaries used one or more psychotropic drugs in 2002 (1). Antidepressants are the most commonly used therapeutic category for general Medicare beneficiaries, followed by antipsychotics for beneficiaries with disabilities and benzodiazepines for older beneficiaries (1,2,3,4,5,6). Historically, Medicare covered merely 7% of the spending on mental health care for its beneficiaries (7,8). The implementation of Medicare Part D extensively expands Medicare's role in financing mental health care. Most psychotropic prescriptions filled by seniors are now paid by Medicare (1).
It is generally believed that Medicare Part D improves access to prescription drugs, reduces out-of-pocket payments, and will therefore increase the overall medication utilization among Medicare beneficiaries. A recent study by Lichtenberg and Sun (9) reported that Medicare Part D reduced the out-of-pocket cost for the elderly population by 18% and increased the use of prescriptions by about 13% in 2006. However, the impact of the policy change on the utilization of and expenditures on psychotropic medications is less certain, for several reasons. First, a commonly used category of therapeutic psychotropic agents, benzodiazepines, is currently excluded from Medicare formularies (10,11). Second, under capitated Medicare Part D, Medicare prescription drug plans (PDPs) have strong incentive to restrict access to expensive psychotropic medications by applying management tools such as prior authorization, step therapy, or quantity limits (1). Finally, more than two million persons with mental illness who are dually eligible for Medicare and Medicaid and whose drug benefit had been covered by state Medicaid programs may face increased cost sharing because Part D's coverage is less generous than Medicaid's (12). It remains unknown to what extent Medicare Part D has affected psychotropic utilization and the consequent expenditures for seniors. An evaluation of population-level changes can help researchers and policy makers understand the cost and access issues of the new drug benefit.
Therefore, the objective of this study was to evaluate the effect of Medicare Part D on the utilization of three most commonly used psychotropic therapeutic categories among seniors—namely, antidepressants, antipsychotics, and benzodiazepines—and related out-of-pocket expenses for those medications.
Two-year pharmacy claims data (2005—2006) were collected from one of the nation's largest retail pharmacy chains, which filled a total of 529 million prescriptions in the fiscal year of September 2005 through August 2006. Among the approximately 70 million individuals who filled prescriptions at the pharmacy chain in 2006, 11% were age 65 or older.
Retail pharmacy claims data were selected for the study for the following two reasons. First, pharmacy claims data collected directly from pharmacy chains are the best possible resource to assess out-of-pocket expenditures. Unlike pharmacy data collected by PDPs and Medicare Advantage Prescription Drug plans, retail pharmacy claims data include the utilization and payment information of prescriptions not covered by Part D (such as benzodiazepines) and the prescriptions that fall into either the deductible or the coverage gap ("doughnut hole"). Second, it is possible to directly compare drug utilization before and after Part D was implemented because seniors have been continuously using retail pharmacy services, disregarding the change of insurers.
Medicare Part D prescription drug benefit
The standard benefit for Medicare Part D for 2006 has a $32 monthly premium and $250 deductible, then coinsurance of 25% up to $2,250 in total drug costs, followed by a gap in coverage between $2,250 and $5,100, where enrollees pay 100% of the costs of their medications. After enrollees have incurred $3,600 in out-of-pocket expenses, they qualify for catastrophic coverage and pay 5% of medication costs (13).
In addition, the new drug benefit also includes substantial premium and cost-sharing subsidies (low-income subsidy) for Medicare beneficiaries with low income and modest resources (<150% of the federal poverty level), which eliminates premiums, deductibles, coinsurance, and doughnut holes (14). Beneficiaries with incomes below 100% of the federal poverty level pay a copayment of $1 for each generic prescription and $3 for each brand-name prescription. Those with incomes above 100% of the poverty level pay $2 and $5, respectively.
Psychotropic drug utilization and related out-of-pocket expenses were calculated according to therapeutic categories. The monthly utilizations of antidepressants, antipsychotics, and benzodiazepines were measured as total number of prescriptions filled by seniors (age 65 or older) in each month. The out-of-pocket expense related to psychotropic medications was measured with the proportion of prescription expenditure paid out of pocket, which was calculated as prescription drug cost charged to individuals (out-of-pocket spending) divided by total pharmacy reimbursement.
Total pharmacy reimbursement was defined as the sum of third-party reimbursement and individual out-of-pocket spending. The percentage was used to measure the financial burden of seniors because the net out-of-pocket expense might have increased because of the increase in utilization.
Research design and analysis
Segmented regression of interrupted time-series analysis was used to estimate the effects of the new Medicare drug benefit on psychotropic drug utilization (15,16). We analyzed psychotropic drug prescriptions filled by seniors in the retail and mail-order pharmacies over a period of 24 months, including 12 months before Part D became effective (January 2005 through December 2005) and 12 months after Part D became effective (January 2006 through December 2006).
Interrupted time-series analysis is characterized by a series of measurements over time that are interrupted by an event, such as an intervention. This method allows us to assess how much intervention—in this case, the introduction of Medicare Part D—changed specific outcomes (utilization of antidepressants, antipsychotics, and benzodiazepines) immediately and over time. For this study, the model is preintervention segments O1, O2, O3, … O11, and O12, interrupted by intervention X, followed by postintervention segments O13, O14, O15, … O23, and O24, where Oj=monthly utilization of antidepressants, antipsychotic drugs, or benzodiazepines; j=month number, where j is numbered, for example, 1 for January 2005, 2 for February 2005, and 24 for December 2006; and X=introduction of Part D in January 2006.
The time-series segmented linear regression model was fit with monthly utilization estimates to evaluate changes after the activation of Medicare Part D (January 1, 2006), as follows:
Y=β0 + β1X1 + β2(X1 - 11) X2 + β3X2 + ϵi
where Y=monthly utilization by therapeutic categories, β0=intercept, β1=secular trend before Part D implementation (months 1 through 12), X1=month number (months 1 through 24), β2=secular trend after Part D implementation (months 13 through 24), X2=design variable (before Part D=0; after Part D=1), β3=one-time effect of the month of joining Part D, and ϵi=error term.
The time series consisted of two segments, before and after the policy activation date, that could be evaluated for both changes in the linear slope (the rate of utilization) and an instantaneous change in magnitude immediately after Part D became effective (January 2006). The coefficient β0 indicates the y intercept of the regression model and represents the initial utilization level at time 0 (January 2005), β1 indicates the initial monthly rate of utilization occurring from January 1, 2005, through the Medicare Part D implementation date (January 1, 2006), and β2 indicates the change in utilization rate (that is, the slope change) after the implementation of Part D. The changes occurring immediately after the implementation of Part D are evaluated by β3.
In ordinary least-squares regression analysis, it is assumed that error terms are uncorrelated. Error terms of consecutive observations are often correlated because time is a predictor in segmented regression analysis. Underestimation of standard errors and overestimation of significance of intervention effects will result if autocorrelation is not corrected (16). To overcome this problem, the model was corrected for autocorrelated errors using the SAS PROC AUTOREG procedure (17). The Durbin-Watson statistic by this procedure was used to gauge serial autocorrelation of error terms. Values near or greater than 2.00 are evidence of no correlation (16,17).
The number of seniors who filled at least one antidepressant, antipsychotic, or benzodiazepine prescription at the retail pharmacy grew 7%, from 1.19 million in 2005 to 1.28 million in 2006. The most significant growth was observed in antidepressant usage (10.6%), followed by antipsychotic usage (8.5%) and benzodiazepine usage (4.4%). Despite the increase in senior psychotropic users since Medicare Part D became effective, the demographic characteristics remained stable. The mean±SD age of the study population was 75.2±7.6 in 2005 and 74.9±7.4 in 2006. Most were female (68.5%, or 818,008 of 1,193,831 in 2005, and 68.7%, or 877,987 of 1,277,793 in 2006) and most were from the Midwest (28.8% in 2005 and 28.4% in 2006), followed by the South Atlantic (25.8% in 2005 and 25.9% in 2006), West (19.4% in 2005 and 19.7% in 2006), other Southern states (18.8% in 2005 and 18.8% in 2006), and the Northeast (7.1% in 2005 and 7.2% in 2006).
Psychotropic prescription costs
Table 1 presents the roles of different payers of psychotropic medications for seniors from 2005 to 2006. Before Medicare Part D became effective, Medicaid was one of the most important third-party payers of psychotropic medications. In 2005 Medicaid paid 27% of total pharmacy reimbursement for antipsychotics and 12% for antidepressants. With the implementation of Part D, Medicare replaced Medicaid in paying these two therapeutic categories for seniors. In 2006 Part D covered 40% of the total pharmacy reimbursement for antipsychotics and 29% for antidepressants. At the same time, the proportion of psychotropic medication reimbursement paid by Medicaid declined to 1% for antidepressants and 2% for antipsychotics.
With respect to out-of-pocket burden, the proportion of psychotropic prescription pharmacy reimbursement paid out of pocket by seniors decreased 18% for antidepressants (from 40% in 2005 to 33% in 2006) and 21% for antipsychotics (from 28% in 2005 to 22% in 2006), which represents $4.52 savings per antidepressant prescription and $5.71 savings per antipsychotic prescription. Unlike antidepressant and antipsychotic medications, benzodiazepines were excluded from the Part D formulary. Therefore, the out-of-pocket share that seniors paid for benzodiazepine prescriptions increased 19% (from 63% in 2005 to 75% in 2006) during the same period, which represents a net increase of $2.79 per prescription.
Psychotropic drug utilization
The segmented linear regression model equation shown previously was used in all models for assessing the utilization of psychotropic drugs. The Durbin-Watson test results produced by the SAS procedure for each of the models were significant, indicating that serial autocorrelation was detected. Therefore, all models were corrected for autocorrelation.
Table 2 and Figures 1, 2, and 3 present the change in psychotropic utilization during 2005 through 2006. Before Part D was implemented, the monthly levels of psychotropic drug utilization were 271,763 prescriptions for antidepressants, 44,559 for antipsychotics, and 228,152 for benzodiazepines, with significant upward trends of 1,537 prescriptions per month for antidepressants, 901 prescriptions for benzodiazepines, and a significant downward trend of -140 prescriptions per month for antipsychotics in the segmented linear regression analysis (Table 2).
The implementation of Part D led to an immediate 5% drop in the level of benzodiazepine utilization (-12,339 prescriptions) but no significant change in trend (416 prescriptions). In contrast, there were no significant changes in levels of antidepressant and antipsychotic utilization after the implementation of Part D (antidepressant prescriptions, 851; antipsychotic prescriptions, 198), but there were significant increases in trends: 1,679 prescriptions per month for antidepressants and 567 prescriptions per month for antipsychotics. In December 2006, 12 months after Part D went into effect, senior patients filled 329,884 antidepressant prescriptions and 48,276 antipsychotic prescriptions, which represent 7% and 18% growth in the utilization of antidepressant and antipsychotic medications, respectively, compared with the expected level estimated on the basis of preintervention trends.
Findings from this study demonstrate that the Part D drug benefit has made Medicare the most important third-party payer of antidepressant and antipsychotic medications for seniors. In 2006 Medicare spent a total of $279 million and $58 million on antipsychotic and antidepressant prescriptions dispensed, respectively, to seniors at the study's retail pharmacy chain, accounting for 40% of all payments for antipsychotic medications and 29% of all payments for antidepressant medications. Seniors who have been using these medications tend to benefit more from the new drug benefit than the general elderly population. The average out-of-pocket expenditure per prescription in 2006 dropped 21% (from $22.06 to $17.54) for antidepressants and 17% (from 32.72 to $27.01) for antipsychotics when compared with 2005 costs. These savings are higher than the average 14% savings for all seniors that Lichtenberg and Sun (9) estimated using the same data.
Unlike antidepressant and antipsychotic medications, another important psychotropic therapeutic category for seniors—benzodiazepines—was excluded from the Medicare Part D formulary. Our results showed that Part D was associated with a substantial drop in the share of psychotropic medication costs paid by the third parties other than Medicare and Medicaid, even including the psychotropic medications excluded from the Part D formularies. The share paid on benzodiazepines by other third parties dropped from 29% to 21% with the implementation of the new Medicare drug benefit, which implies that many Medicare enrollees lost benzodiazepine coverage when their prescription benefit changed to Part D from employer-sponsored insurance for retirees, Medigap, or Medicare health maintenance organizations. Therefore, instead of gaining any reduction in out-of-pocket spending, seniors paid 19% more out of pocket on benzodiazepine prescriptions (from 63% in 2005 to 75% in 2006) compared with the period before Part D was implemented. It is important to note that despite the continuous Medicaid coverage of benzodiazepines for persons dually eligible for Medicare and Medicaid benefits after Part D implementation, the share of benzodiazepine costs paid by Medicaid dropped significantly, by 50% (from 8% in 2005 to 4% in 2006). Obviously, the financial burden associated with benzodiazepine prescriptions had shifted from employers and Medicaid to seniors during the post-Part D period.
The reason for blanket exclusion of benzodiazepines under Part D was that benzodiazepines, more than many other medications, are subject to abuse and misuse by a small proportion of those who take them. The adverse effects associated with benzodiazepine use by seniors include oversedation, falls and hip fractures, and dependency (18). Most benzodiazepines have been identified in the Beers List, a guide that identifies medications or classes of drugs that should be avoided by seniors (19,20). However, benzodiazepines are inexpensive and are often prescribed to seniors for the management of acute anxiety, panic attacks, seizure disorders, and muscle spasms. Considering that nearly one-quarter of disabled persons with dual eligibility rely on benzodiazepines to manage their mental and neurological disorders, coverage under Part D is needed for these agents. Many nonprofit organizations, such as the Medicare Rights Center, the National Alliance on Mental Illness of Minnesota, and the American Psychiatric Association, have exerted extensive effort in urging the coverage of benzodiazepines by Part D both before and after Part D was implemented. Finally, in July 2008, Congress decided to remove the current ban in Part D to allow coverage for benzodiazepines.
The implementation of Medicare Part D had various effects on psychotropic prescription utilization among seniors. The most significant change was found in the use of antipsychotics (18% increase), followed by antidepressants (7% increase) and benzodiazepines (5% decrease). Although the extent of variation in utilization is positively correlated with the net savings per prescription (antipsychotics, $5.71 savings; antidepressants, $4.52 savings; benzodiazepines, $2.79 increased expense), the implementation of Medicare Part D had relatively small effects on the utilization of psychotropic prescriptions compared with the change in out-of-pocket spending. The less significant change in utilization is probably because many seniors (approximately two million) who have been using psychotropic medications extensively were dually eligible for full Medicare and Medicaid benefits rather than being uninsured (21). The prescription drug benefit of these seniors was covered by Medicaid before 2006 (14). With Medicare Part D, dually eligible beneficiaries' drug benefit shifted from Medicaid to Medicare. A great share of Medicare spending on antidepressant and antipsychotic medications was probably used to fill the coverage gap left by Medicaid for persons with dual eligibility. Our results show that Medicaid was one of the nation's major payers of psychotropic medications for seniors, accounting for more than a quarter of antipsychotic spending and 12% of antidepressant spending at the pharmacy chain in 2005. With the implementation of Medicare Part D, the share of antidepressant and antipsychotic costs paid by Medicaid dropped to near zero (2% for antipsychotics and 1% for antidepressants).
In addition, we noticed that there was a month-to-month fluctuation in the utilization of benzodiazepines according to the plot of the original data (Figure 3). Unlike antidepressant and antipsychotic prescriptions, which are usually used for an extended period, benzodiazepines are often prescribed for short-term use. The renewal of benzodiazepine prescriptions usually requires a reevaluation of the patient's condition by his or her physician, which might create a gap between two consecutive benzodiazepine prescriptions because of the scheduling of appointments.
The changes in psychotropic utilization suggest that Medicare Part D has increased access to commonly used psychotropic medications (antidepressants and antipsychotics) in the elderly population by reducing out-of-pocket expenses while discouraging the utilization of psychotropic medications that may sometimes be unsafe (benzodiazepines). However, it is also possible that the current new drug benefit has pushed patients to switch from established benzodiazepine regimens to other, more expensive and less effective medications covered under Part D. It remains unknown how these changes in psychotropic medication use have affected the outcomes of treating psychiatric and neurological disorders among seniors. With Congress passing the bill to extend the coverage of benzodiazepines to all Part D beneficiaries, it will be interesting to track the experiences of various subgroups of Medicare beneficiaries with mental disorders to see how the policy change (exclusion and inclusion of benzodiazepine) will affect psychiatric practice and patient outcomes.
Some limitations must be considered in interpreting the study results. First, a concurrent comparison group was not available in this study. The results might have been biased because of changes in the composition of the study population and simultaneously occurring intervention (15). However, our analysis of population demographic characteristics has shown that the characteristics of the study population remained stable during the study period. With respect to simultaneously occurring interventions, if there were any, they should not have had an impact great enough to reverse the Part D effect because Medicare Part D is the most significant health care policy that has been implemented since Medicaid and is also the only prescription drug policy that affects the entire senior population. Furthermore, segmented regression analysis addresses important threats to internal validity such as history and maturation. For instance, the design takes into account the increase in sales over time as a result of the company's business expansion by making multiple assessments of the outcome variable before and after the intervention.
Second, the business focus of the retail pharmacy chain is the community rather than institutionalized populations. Therefore, the estimated utilization trend cannot be generalized to elderly persons staying in psychiatric hospitals and long-term-care facilities. However, we believe that the study results are applicable with the general community-dwelling elderly population even though the data were collected from only one pharmacy chain. First, there was not obvious selection bias in terms of patients covered by different prescription insurances considering that the pharmacy chain accepts almost all third-party contracts nationwide. Second, the study population is geographically representative of the dwelling patterns of the senior population. The pharmacy chain's retail outlets can be found in 47 of 50 states. Seniors who filled psychotropic medications at the retail pharmacies were evenly distributed across most geographic regions (West, Midwest, South, and South Atlantic). Third, the sample of our study was large. The retail pharmacy chain is one of the largest nationwide in terms of prescription sales. Our study sample included more than one million seniors who were psychotropic users both before and after the implementation of Medicare Part D. Even in the most underrepresented geographic region (Northeast), more than 85,000 seniors filled at least one psychotropic prescription annually.
Finally, psychotropic medications filled outside of the pharmacy chain were not included in the analysis. One of the major concerns about the value of retail pharmacy data for research purposes is that complete patient-level prescription records are usually not available. However, this issue was not a concern for our study because it was a prescription-level rather than a patient-level analysis.
Psychotropic medications are the mainstay of seniors with mental disorders. Our findings confirm the success of Part D in helping this population and improving seniors' access to psychotropic medications covered in Part D formularies. Seniors paid less out of pocket for antidepressant and antipsychotic prescriptions and used more of these medications in 2006 than they had in 2005. Our findings also highlight the fact that senior benzodiazepine users had to pay a greater proportion of benzodiazepine cost out of pocket compared with the period before Part D because of the blanket Part D exclusion of this therapeutic category. Future studies are needed to evaluate the clinical impact of Part D on patients with mental and neurological disorders.