Impact of Medicare Part D on Utilization and Expenditures

, , , ,
The American Journal of Pharmacy Benefits, June 2010, Volume 2, Issue 3

Examination of pharmacy behaviors before and after implementation indicated that restrictions in pharmacy benefits may not deter utilization for Part D.

The Medicare Part D benefi t implemented on January 1, 2006, is unique because after initial coverage is offered to members and a spending limit is reached, a gap or loss of coverage ensues while members accrue additional out-of-pocket (OOP) costs. Thus, the Part D benefit has a restriction on prescription coverage built into its benefits.1

A primary source of Part D coverage is through Medicare Advantage (MA) plans. Prior to 2006, MA plans were referred to as Medicare+Choice (M+C) plans, which also provided prescription drug coverage but in a very limited capacity.2 Prior to 2006, most M+C plans had caps on pharmacy benefits. In 2005, more than 50% of the M+C plans had annual caps for brand-name and generic medications of $1000 or less, of which 30% had annual caps of $500 or less.3

The Part D benefi t comprises 3 phases: pregap, gap, and the postgap. The pregap phase consists of (1) a $250 deductible; (2) a monthly premium that averaged $32 a month in 2006; (3) 25% coinsurance on formulary drugs up to a maximum of $500; and (4) $2250 initial coverage limit on total drug expenses including the deductible, member copays, and health plan costs for prescription medications. In the gap phase, individuals have $2850 in OOP costs followed by a $3600 OOP limit (which is the sum of the $250 deductible, $500 maximum coinsurance, and $2850 gap OOP costs) reached after total drug costs equal $5100. The postgap phase provides catastrophic coverage, with no more than 5% coinsurance for formulary drugs.

Figure 1

shows a graphical representation of the Part D benefit in 2006.

Prior to 2006, the impact of restrictions on pharmacy benefi ts in M+C plans was examined. In 2003, Hsu and colleagues compared the impact on various outcomes of a $1000 annual cap on pharmacy benefi ts with no cap on pharmacy benefi ts.4 The authors found that members with capped benefi ts were more likely to be nonadherent with medications for chronic conditions and had higher rates of emergency department visits, nonelective hospitalizations, and mortality. Tseng and colleagues surveyed M+C members who exceeded annual caps of $750 or $1200 in 2001 and compared them with a control group who did not exceed a $2000 cap.5 A higher proportion of members who exceeded their caps reported using fewer prescribed medications, switched medications, and had difficulty paying for prescriptions.

Rector and Venus surveyed 1500 Medicare members with chronic diseases enrolled in M+C plans with a $200 to $300 annual drug benefi t and found that 32% of members did not fi ll a prescription or reduced a prescribed dosage because of OOP costs.6 Researchers at RAND compared pharmacy use among retirees in plans with annual caps of $1000 and $2500 with pharmacy use among retirees in noncapped plans between 2003 and 2005.7 By the end of each calendar year, the use of antidepressants, antihypertensives, antihyperlipidemics, and diabetic medications was 15% to 28% lower among members in capped plans versus those in noncapped plans.

Like the caps under the M+C benefi ts, the gap under the Part D benefit is a restriction on the pharmacy benefit. The impact of the gap under Part D also has been investigated. Members who reach the gap appear to stop their medications, switch to alternatives, or reduce their existing regimens of medication.8 Current evidence suggests that both the cap restrictions under the M+C plans and lack of coverage in the gap under the Part D benefi t change medication-purchasing behaviors for Medicare members. Studies solely examining the impact of Part D reveal that the benefit results in a 5.9% to 12.8% increase in the use of prescription medications and a 13.1% to 18.4% reduction in member OOP expenditures.9-12 However, what remains unknown is whether the expected generosity of the Part D benefi t (despite the presence of the gap) improves prescription use and adherence among Medicare members, specifically those with chronic diseases who previously faced a cap on their pharmacy benefi ts.13

Our goal was to compare prescription use, medication persistence, and OOP expenditures for Medicare members who were enrolled in an M+C plan with a capped pharmacy benefi t in 2005 and who switched to an MA plan in 2006 that provided no gap coverage. Our analysis assessed the policy question of whether the Part D benefi t truly does improve medication use for Medicare beneficiaries who faced restrictions in their coverage before and after Part D.

METHODS

Data Sources

Data sources were enrollment, pharmacy, and medical claims for Medicare benefi ciaries who were enrolled in an M+C plan in 2005 and switched to an MA plan in 2006 that was offered by a large insurer. This insurer provided coverage to more than 3.1 million individuals, including 58,561 Medicare beneficiaries in MA plans in 2006.

Description of the M+C Plan and the MA Plan

The M+C plan in 2005 and the MA plan in 2006 were point-of-service plans with both medical and pharmacy benefi ts (

Table 1

). The prescription benefi t for the M+C plan (Table 2) was a $100 deductible followed by a 3-tier prescription benefi t plan and a $600 annual cap for formulary and nonformulary brand-name medications. The annual cap of $600 for brand-name drugs was the sum of total drug costs, which included the member’s share and the health plan’s share. The member’s share included $10 copays for generic (tier 1), $20 for formulary brand-name (tier 2), and $35 for nonformulary brand-name (tier 3) medications for retail prescription purchases. Mail-order purchases had twice these copays for each tier.

The MA plan in 2006 had a $250 annual deductible, followed by a 4-tier plan with retail copays of $10 for generic, $19 for formulary brand-name, and $39 for nonformulary brand-name medications, and 25% coinsurance for specialty and self-injectable drugs until members reached the $2250 cap, which included both the member’s and the health plan share of drug costs. Therefore, the deductible amount, prescription copays/coinsurance, and health plan costs for medications counted toward the $2250 cap. After this cap was reached, members paid 100% of their prescription costs until their OOP costs reached $3600 or total pharmacy expenditures (member OOP costs and health plan costs) reached $5100. After this limit, members paid no more than 5% coinsurance for formulary drugs.

Sample Inclusion Criteria

The study included individuals who were enrolled in a M+C plan for at least 6 consecutive months during calendar year 2005 and switched to an MA plan and were enrolled for at least 6 consecutive months in 2006. A subset of individuals with diabetes, hypertension, dyslipidemia, or congestive heart failure (CHF) also were identified. These members were considered to have 1 of these diseases if they had either 2 or more claims for outpatient visits or 1 or more claims for a hospitalization or an emergency department visit with either a primary or secondary diagnosis for the disease of interest during the calendar year 2005.14

Measures

Characteristics of the Study Cohort. Demographic and clinical traits of members in the study cohort that were available in the database included age, sex, and risk score. The numbers of members with hypertension, dyslipidemia, CHF, or diabetes in 2005 also were identified. Time to reach the cap or gap was determined, as well as time spent in the gap.

Pharmacy Utilization (Overall and Disease Specific). The numbers of members purchasing prescriptions in the study cohort in 2005 (before Part D was implemented; the preperiod) and 2006 (after Part D was implemented; the postperiod) were determined on a per member per month (PMPM) basis. Medication persistence (disease specifi c) or “therapy persistence” is a measure of how many days in the study period a member was in possession of at least 1 drug within a drug class. Therefore, persistence was measured using the proportion of days covered for any drug used to treat 1 of the 4 disease states, and an overall rate for each disease was computed. The proportion of days covered was defined as the ratio of the total days supply from the first to the last prescription for any drug within a drug class (plus the last day supply) to the length of the time period.15 Mean numbers of adherent members in the preperiod and postperiod were determined for each disease state.

Member OOP Expenditures. The mean member OOP pharmacy costs, defined as the sum of the deductibles, copays, coinsurance, and any other OOP costs for prescription medications members paid in the preperiod and postperiod for all prescriptions, were determined separately for brand-name and generic medications.

Independent Variables

The independent variables were age (stratifi ed into 4 categories of ≤64, 65-74, 75-84, and 85+ years, with 65-74 years as the reference category); sex, with male as the reference category; a risk adjuster for each individual determined as the prescription drug hierarchical condition category (RxHCC) score16; indicators for the 4 chronic conditions; and an indicator to designate the preperiod and the postperiod, with the preperiod as the reference category.

Statistical Methods

We conducted descriptive analyses on member characteristics of the study cohort. We used generalized estimating equation regression models17 to estimate the impact of time period (preperiod vs postperiod) on the outcome measures. These models have the general form:

g(μ) = Xβ = η

where

g = link function

X = design matrix (a matrix of the explanatory variables)

β = vector of parameters to be estimated

and

E [Y] = g-1(η)

where

Y = outcome measure

E [Y] = expected value of the outcome measure

The estimated parameters are interpretable as marginal or population averaged and were exponentiated for interpretation. In the case of the Poisson regression, the result was an incident rate ratio (IRR); in the case of logistic regression, the result was the odds ratio (OR). The software allowed us to specify the distribution and link function, providing flexibility for the range of distributions we encountered in the data. Specifically, differences in the mean number of prescriptions were estimated using Poisson regression, and differences in the number of individuals taking any prescription and differences in the number of adherent individuals in the preperiod and postperiod were estimated using logistic regression.

Expenditure data were not normally distributed; for descriptive statistics on the differences between cohort expenditures, we used the bootstrap, resampling with replacement method.18 The 95% confi dence intervals (CIs) were estimated by the percentile method. When the CI excluded 0, the difference was statistically different from 0. All analyses were done using SAS version 9.2 (SAS Institute, Cary, NC).

RESULTS

Sample Characteristics

Characteristics of the study cohort based on calendar year 2005 are presented in

Table 3

. The mean age was 71 years, with members between the ages of 65 and 74 years representing 54.8% of the study sample; 56% of the cohort were women. Hypertension was the most prevalent disease among the 4 examined (38.7%), followed by dyslipidemia (19.1%), diabetes (17.6%), and CHF (5.1%).

In 2005, 36.73% of members met the $600 annual brand-name cap by the end of the year (Table 4). Members with dyslipidemia were most likely to meet the cap (46.7%), followed by those with hypertension (42.86%), CHF (40.52%), and diabetes (40.56%). In general, members took 297 days (9.8 months) to reach the brand-name cap in calendar year 2005.

In 2006, 29.5% of all members reached the gap phase and remained in the gap for the rest of the year; 5.2% left the gap at the end of 2006, and 58.5% never reached the gap by the end of 2006 (

Table 5

). More members with CHF entered the gap (48%), followed by those with diabetes (44.9%), dyslipidemia (36.3%), and hypertension (36.1%). On average, members took 234 days (7.8 months) from the start of the calendar year to reach this phase.

Mean number of days spent in the gap phase was 115 days (3.8 months) for all members, including those who reached the postgap phase. Diabetic members spent the most amount of time in the gap phase (128 days), whereas hypertensive members spent the least amount of time in the gap (116 days).

Pharmacy Utilization

Observed differences between the preperiod and postperiod, the primary explanatory variable, for overall pharmacy utilization are shown in

Table 6A

. On average, a 17.5% increase was found in all prescriptions PMPM in the postperiod, which translated into a 20.8% increase in the use of generic prescriptions and a 13.63% increase in the use of brand-name medications. All 3 differences were statistically signifi cant as shown by the adjusted IRRs, the 95% CIs for these ratios, and the P values in Table 6A.

The adjusted IRR is the incidence rate of the dependent variable in a given group or condition relative to the referent for categorical covariates, or the relative change attributable to a unit increase in a continuous covariate.19 For the mean number of all prescriptions PMPM in Table 6A, the IRR is 1.08 (95% CI = 1.07, 1.10), which reflects a statistically significant increase in the mean number of all prescriptions in the postperiod (P <.0001). Similarly, the IRR for the mean number of generic prescriptions is 1.12 (95% CI = 1.10, 1.14) and the IRR for the mean number of brand-name prescriptions is 1.05 (95% CI = 1.02, 1.07), showing a statistically signifi cant increase in the mean number of both generic and brand prescriptions in the postperiod relative to the preperiod.

The IRRs for all the covariates in each model used to estimate the adjusted pre/post differences in Table 6A are shown in

eAppendix A

(available at www.ajpblive.com). These estimates indicate the direction of the difference in each of the independent variables relative to the reference category. When the IRR is greater than 1, there is a greater incidence rate of the outcome variable in a category relative to the reference category. For example, in eAppendix A, women had more prescriptions per member than men for 2 models: for mean number of all prescriptions (IRR = 1.04; 95% CI = 1.01, 1.08) and for mean number of brand-name medications (IRR = 1.08; 95% CI = 1.03, 1.13). Similarly for mean number of all prescriptions, members who were ≤64 years of age (IRR = 1.09; 95% CI = 1.03, 1.15), between 75 and 84 years of age (IRR = 1.06; 95% CI = 1.01, 1.11 ), and more than 85 years of age (IRR = 1.04; 95% CI = 1.01, 1.08) had more prescriptions than the reference group of members between 65 and 74 years of age. Finally as expected, the mean number of all prescriptions was an increasing function of RxHCC.

Observed differences between the preperiod and postperiod for the number of individuals purchasing prescriptions are shown in

Table 6B

. The ORs for all the covariates for each model used to estimate the adjusted pre/post differences in Table 6B are shown in

eAppendix B

(available at www.ajpblive.com) with similar interpretations for the direction of differences as described for the IRR. That is, when the OR is greater than 1, the odds for the outcome are greater in a given category than the odds for the outcome in the reference category. There was a 7% increase in the number of members who purchased any prescription in 2006 compared with 2005 (adjusted OR = 1.88; 95% CI = 1.71, 2.07). This increase was distributed more toward generics (9.5% increase) than brand names (4.9% increase). The adjusted ORs were 1.71 (95% CI = 1.59, 1.85) and 1.18 (95% CI = 1.09, 1.27), respectively. The odds for individuals to be taking any prescription were a signifi cantly increasing function of RxHCC and lower for each age group relative to the age 65 to 74 reference category; none of the other covariates was signifi cant. These observed differences were similar for individuals purchasing all, brand-name, or generic prescriptions.

Medication Persistence

Table 7

presents the observed differences in the number of persistent members in the preperiod and postperiod for the individual disease states. For all disease states, the number of persistent members increased in the postperiod, with the highest increase for those with dyslipidemia (9.1%) and smaller increases for those with CHF (5.6%), diabetes (5.5%), and hypertension (5.4%). Adjusted ORs show that the odds for persistence were greater in the postperiod for members in all disease states, with the exception of CHF, where this difference was not statistically signifi cant (P = .10). Odds ratios for the covariates in each model used to estimate the adjusted ORs for the postperiod and preperiod in Table 7 are shown in eAppendix C (available at

www.ajpblive.com

). Only being female (OR = 1.49; 95% CI = 1.00, 2.00) and having hypertension (OR = 0.51; 95% CI = 0.33, 0.79) were significant covariates in the subset of members with CHF; age of <64 years (OR = 0.74; 95% CI = 0.58, 0.93) and being female (OR = 1.17; 95% CI = 1.01, 1.36) were significant in the subset of members with hypertension. None of the other covariates except the postperiod indicator were significant in the subsets of patients with diabetes or dyslipidemia.

Expenditures

Mean member OOP expenditures increased from $45.66 to $81.83 PMPM for all prescriptions in the postperiod, representing a 79.21% increase (

Figure 2

). Per member per month OOP costs for generic medications increased from $11.70 to $22.93 (96.06% increase), whereas similar costs for brand-name medications increased from $34.00 to $58.90 (73.38% increase) in the postperiod. The substantial increases in OOP costs members experienced in the postperiod were infl uenced by the OOP expenditures that 29.5% of the study cohort who reached the gap in 2006 faced. We therefore examined preperiod and postperiod member OOP expenditures, excluding members who reached the gap (Figure 2). Members still experienced increases in OOP expenditures, but the increases were not as large: a $9 (12.76%) increase in the postperiod for all prescriptions, a $3.39 (6.27%) increase for brand-name medications, and a $5.77 (32.74%) increase for generic medications.

DISCUSSION

Our primary results show that the Part D benefit appears to encourage the use of prescription medications and improves persistence levels for members with chronic conditions despite the presence of a gap compared with a capped pharmacy benefit. The implementation of Part D appears to have encouraged generic use, which had a 21% increase in 2006 and a 10% increase in the number of members purchasing generics. Use of brand-name medications also increased by approximately 14%, but the number of members purchasing brand-name prescriptions was only 5% higher in 2006. We would anticipate that the brand-name cap in 2005 discouraged the use of brand-name medications. Thus, removing this restriction would increase brand-name use in 2006, as we observed. Therefore, some of the increase in brand-name use may be attributed to the removal of this cap. However, it appears that despite the concerns about its limitations, the Part D benefit encourages the use of generics more than brand-name medications.

The number of members who achieved persistence levels of ≥80% also was greater in 2006, ranging from 5.4% to 9.1% for the specific diseases we examined. The most surprising fi nding was the increase in prescription use and persistence we observed in 2006, despite members experiencing signifi cant increases in OOP expenditures. One reason for the large increase in expenditures was the costs incurred by 29.5% of members who reached the gap in 2006. Under the capped benefit, members were responsible for 100% of the costs of all brand-name medications only; under the gap, members paid 100% of the costs for all medications. In addition, although more members reached the brand-name cap in 2005 (36.73% vs 29.5%), they took longer to reach the cap in 2005 (9.8 months) compared with the time it took for all members to reach the gap in 2006 (7.8 months). In addition, members who reached the cap in 2005 only spent, on average, 1.8 months without coverage for brand-name medications, whereas members who reached the gap in 2006 spent, on average, 3.8 months without coverage for any prescription medications. Clearly, members who reached the gap faced signifi cantly higher costs, and including these members in the pre/post analysis of member OOP expenditures contributed to the substantial increases in costs that were observed in 2006.

With the exclusion of these members, OOP expenditures increased by only $9 PMPM, which consisted of a $5.77 PMPM increase in OOP expenditures for generics and a much smaller $3.39 PMPM increase for brand-name medications. Although prescription behavior within the gap was not examined, our results appear to show that the presence of a gap under the Part D benefit did not reduce use or persistence for members who switched from a capped brand-name benefi t to the Part D benefit.

Other ancillary findings were that in general women were more persistent with their medications and also consumed more medications than men in all the results we observed. In addition, as the risk score for an individual increased, so did their use of prescription medications. Both findings have been corroborated in previous research on the Medicare population, which has shown that being female and having increasingly complex healthcare conditions are positively correlated with more utilization.19,20

Limitations

Our study had several limitations. Most important was the lack of a comparison cohort to examine the differences in pharmacy use patterns and expenditures between those who experienced a change in pharmacy benefits and those who did not. The pre/post approach we used meant that members served as their own historical controls. Because this was a national benefi t that began in January 2006, all MA plans and most employer groups nationwide changed or modifi ed their prescription benefits in some way. Therefore, fi nding an appropriate Medicare comparison cohort would have been diffi cult. Comparisons with Medicare-ineligible members younger than age 65 years who did not experience a change in prescription benefits have already been done and may not represent the most appropriate comparisons.

Our second key limitation was the small sample size of the primary cohort. We examined a total of 6876 individuals who were enrolled at some point in the preperiod and the postperiod. By eliminating a continuous 24-month enrollment criterion (which would have reduced our sample size considerably), our sample included individuals who might not have had an opportunity to reach the brandname cap in 2005 or the gap in 2006, either because members had until May 15, 2006, to join a Part D plan or because of the enrollment criteria. Therefore, our results may be somewhat underestimated. Because a majority of Part D members in the MA plans we observed in 2006 were members who “rolled over” from the M+C benefit with the same health benefit in 2005, we did not expect a majority of members to have had fewer than 12 months of enrollment in each time period.

Our third limitation was that we had to estimate the brand-name OOP expenditures for members who reached the cap in 2005 after they reached the cap. Because members were responsible for all OOP expenditures after the cap was reached in 2005, the actual costs were not captured in the pharmacy claims. We therefore estimated member OOP expenditures for brand-name drugs after the cap was reached using the sum of health plan and member OOP expenditures for these drugs that members incurred before they reached the cap. Because most health plans are able to negotiate discounted prices for their members, we expected that a majority of members continued to purchase their brand-name medications at the discounted rates offered by the health plan, thereby validating the estimation methods we used. However, a possibility still exists that we either overestimated or underestimated member OOP costs for brand-name medications after the cap was reached.

Finally, the capped benefit we examined in 2005 was more generous than what the M+C marketplace offered in the same year. Although the $600 annual cap on brand-name medications was restrictive, members had access to unlimited generic medications, available at a $10 copay for retail purchases or a $20 copay for mail-order purchases. A majority of the drug classes we examined in this study had generic alternatives, either within a particular drug class or through therapeutic generic alternatives within other drug classes. Thus, members were already exposed to a more generous M+C benefi t compared with the marketplace norm in 2005, which was a cap on all prescription purchases, not just brand-name prescriptions. Individuals could, therefore, continue to use generic medications for the treatment of the more common chronic conditions. Despite the comparative generosity of the M+C benefit, we still saw increases in use and persistence in 2006 under the Part D benefit, suggesting that members were responding to what they perceived as a more generous benefit in 2006.