Private insurers are documenting changes in the prescription utilization of health benefi ciaries. Such detailed prescription monitoring includes impact analyses of drug benefi t program costs, as well as general health outcome research.1-10 Although studies that evaluate capped benefi ts for Medicare recipients have been conducted, limited member adjudication research has been published on the new Medicare Part D prescription drug program. The Medicare Part D prescription drug program has largely been a success, enrolling just more than 20 million total prescription drug plan members by midyear 2006 with approximately 10 million enrolled in the standard Prescription Drug Plan (PDP) in the fi rst year of the program.11 The initial Part D PDP in 2006 included a $250 deductible with the beneficiary responsible for a coinsurance payment of 25% until total drug spending equalled $225. After reaching $2250 in drug expenditures, the member was responsible for 100% cost sharing until $5100 in total drug expenditures was incurred, with a maximum of $3600 of out-of-pocket expenses. This gap in coverage is commonly referred to as the “doughnut hole” because members are responsible for 100% of prescription drug costs; thus, a “hole” exists in the coverage design. The deductible and spending benefit thresholds increase each year. The standard benefit in 2010 had a $350 deductible and 25% cost sharing to the initial coverage limit of $2830.11

Several studies do suggest that increasing the costs to patients results in alterations of these patients’ prescription fulfi llments.2,10,12-19 Furthermore, out-of-pocket costs have been shown to be highest among the elderly, with the average senior taking at least 6 medications.20 From a policy perspective, there is extreme interest in determining the relationship between decisions Part D seniors make regarding filling, switching, delaying, or stopping their prescription medications and the Part D benefit design structure. Hence, the present study investigates whether there is a relationship between prescription medication utilization and the doughnut hole phase of the Part D PDP among a sample of Part D PDP among a sample of Part D Medicare enrollees.

METHODS
Data for this retrospective case-control study were obtained from a national HMO Part D benefit plan’s pharmacy administrators via their pharmacy prescription–tracking database. The sampling frame consisted of HMO Part D participants who had exceeded $2250 in drug spending. A total of 750 active HMO Part D participants (deidentified for HIPAA compliance) 62 years and older were randomly selected for inclusion in the study. This group was subsequently stratified by benefit design (standard PDP) or Enhanced Drug Coverage (enhanced plan), and a random sample was drawn from each of these strata. Thus, there were 2 study groups: cases (ie, standard PDP members, n = 500) and controls (ie, enhanced plan members, n = 250). We compared 500 patients who had no doughnut hole coverage with 250 patients who had prescription drug coverage through the doughnut hole.

Members with less than $2250 in annual drug expenditures were excluded from the analysis, as they failed to reach the doughnut hole: the independent variable of interest. Dual eligibles, members who possessed additional drug coverage benefit to the Part D plan, were also excluded from the analysis. Some members did not reach catastrophic expenditure levels.

The independent variable, doughnut hole status, was measured by summing drug expenditures for each member until the threshold of $2250 was met. Prescription fulfillment decisions were operationalized into the dependent- variable nominal categories of filled, switched, delayed, and stopped. Each prescription of each Part D participant within the HMO was followed monthly throughout 1 full year via the health plan’s pharmacy adjudication database. Enrollee data were analyzed to ascertain the effect of doughnut hole status on prescription fulfillment decisions over the course of 1 full year. Each member’s prescriptions by month were reviewed for any changes in prescription fulfillment decisions, measured by whether the member switched, delayed, or stopped the medication. All study data were collected in accordance with guidelines set forth by the Trident University International Institutional Review Board.

RESULTS
There were 323 male and 427 female participants. Participants in both the doughnut hole and non–doughnut hole groups were taking an average of 5.9 different classes of medications (SD = 1.60) from 8 total medication classes categorized in this study (Table 1). Each member obtained a prescription sometime within the year for an average of 17.5 different medications (SD = 6.93). Almost two-thirds of members (65.3%, n = 490) were shown to need prescriptions for 5 to 7 classes of medications (23.3% [n = 175] for 7 classes, 21.2% [n = 159] for 6 classes, and 20.8% [n = 156] for 5 classes). Those members with the standard PDP plan spent 144.7 days (SD = 50.72) on average in the doughnut hole (Figure). Members with the enhanced plan experienced 0 days in the doughnut hole.

The median number of days in the doughnut hole was 151, and the most frequently reported number of days spent in the hole was 167.Of 500 standard PDP members, 135 (27%) spent 5 to 6 months in the doughnut hole, while 22% (n = 112) spent 4 to 5 months in the hole. Almost three-fourths (73%) of standard Part D members who experienced the doughnut hole (n = 363) spent more than 4 months there. Table 2 depicts the odds ratios [ORs] for occurrence of each of the coded prescription fulfillment behaviors in the study. Members who experienced the doughnut hole were 1.4 times more likely to switch their medication (OR 1.39, 95% confidence interval [CI] 0.86-2.27), 2.3 times more likely to delay their prescriptions (OR 2.33, 95% CI 1.48-3.66), 3.3 more times likely to switch and delay their prescriptions (OR 3.29, 95% CI 1.53- 7.05), and 4.9 more times likely to switch, stop, and delay their prescriptions (OR 4.88, 95% CI 1.47-16.23) compared with members who had coverage through the doughnut hole period. This means those members that experienced the doughnut hole were 3 to 4 times more likely to engage in “multilayering” within prescription fulfillment decision making; that is, they decided to change their medications via more than 2 prescription fulfillment behaviors.

Additionally, members who experienced the doughnut hole were 1.3 times more likely to switch and stop medications (OR 1.32, 95% CI 0.63-2.78) and 1.2 times more likely to delay and stop their prescriptions (OR 1.21, 95% CI 0.71-2.07). Of all 750 study participants, 18.3% (n = 137) decided to delay their medication, 12.3% (n = 92) switched to a different medication (typically a generic version), and 9.5% (n = 71) both delayed and stopped some of their medications.

In Table 3, the standard PDP group (n = 500) is compared with the enhanced plan group (n = 250). The standard PDP members struggled a great deal more than enhanced plan members to fill their prescriptions regularly. Compared with enhanced group members, standard PDP members were about half as likely to fill their prescriptions. The fulfillment decisions made by standard PDP members included delaying their prescriptions (n = 110, 22%), making product switches (n = 67, 13.4%), and stopping a medication entirely (n = 30, 6%). Some standard PDP members delayed and stopped medications (n = 50, 10%); switched and delayed their medication (n = 49, 9.8%); switched, delayed, and stopped some of their medications (n = 28, 5.6%); or switched and stopped some medications (n = 26, 5.2%). Only 28% of participants (n = 140) who experienced the doughnut hole filled their medication versus 54% of enhanced members (n = 135). More than 28% (n = 140) of standard PDP members made prescription alteration decisions that affected 1 class of medication. Another 26.2% (n = 131) made prescription alteration decisions that affected 2 classes of medications, 13.2% (n = 66) made prescription alteration decisions that affected 3 classes of medications, and 3.8% (n = 19) made prescription alteration decisions that affected 4 classes of medications. In the enhanced plan group, 22.8% (n = 57) made prescription alteration decisions that affected 1 class of medication, 14.4% (n = 36) made prescription alteration decisions that affected 2 classes of medications, 4% (n = 10) made prescription alteration decisions that affected 3 classes of medications, and 4.8% (n = 12) made prescription alteration decisions that affected 4 classes of medications. However, more than half of the enhanced plan members filled their original prescriptions without alterations throughout the entire year.

The number of classes affected by members’ medication fulfillment decisions varied significantly between members experiencing the doughnut hole (mean = 1.36, SD = 1.13) and members not experiencing the doughnut hole (mean = 0.83, SD = 1.12) (F[1, 748] = 37.03, P <.001). Of 750 members, 475 (63.3%) altered their prescription fulfillment decisions, which did increase the number of medication classes affected. An increase in the number of affected medication classes produced a steady rise in the mean differences in prescription fulfillment decisions.

In Table 4, the final regression model demonstrated that age (OR 1.02, 95% CI 1.00-1.05), sex (OR 0.97, 95% CI 0.69-1.35), and medication classes (OR 0.92, 95% CI 0.80-1.07) did not significantly influence whether members altered their prescription fulfillment behavior. In the model the number of medications a person was taking significantly contributed to a decision not to fill prescriptions (OR 1.10, 95% CI 1.06-1.14). Equally, the number of days spent in the doughnut hole contributed to whether or not members filled all of their prescriptions (OR 1.02, 95% CI 1.0-1.02). The tests for goodness-of-fit for the model demonstrated high applicability.

DISCUSSION
The purpose of this retrospective, case-control pharmacy adjudication prescription fulfillment study in Medicare members with standard PDP and enhanced Part D plans was to determine whether the presence of the doughnut hole (a gap in Part D coverage requiring full cash payments for prescriptions) influenced whether these members filled their prescriptions. Research on other types of drug benefits suggests the potential for prescriptions to be delayed, switched to another product, or stopped completely when patients are faced with a change in the dynamic of their drug coverage.2,10,12-17,19,21,22 Current estimates have the doughnut hole surpassing $6000 by the year 2016.23 Therefore, it is important to understand how changes in benefit design affect prescription fulfillment behaviors.

The 750 members took an average of 17.5 different medications (SD = 6.93) in a year, encompassing an average of 5.9 classes of medications (SD = 1.60). According to the American Society of Consultant Pharmacists, more than 75% of seniors aged 65 to 79 years have 1 or more chronic diseases. The society notes that “on average, individuals 65 to 69 years old take nearly 14 prescriptions per year, individuals aged 80 to 84 take an average of 19 prescriptions per year.”24 The outliers for the distribution included 2 members with only 1 medication and 6 members with more than 38 medications, including 1 member with 44 different medications used within the year. One group of 500 standard PDP members experienced the doughnut hole, and a second group of 250 enhanced plan members did not experience the doughnut hole. However, a doughnut hole time period was calculated based on drug expenditures for the enhanced plan group. The Part D members who experienced the doughnut hole remained in the hole for a considerable amount of time. Standard PDP members spent a mean of 144.7 (SD = 50.72) days in the doughnut hole; enhanced plan members spent none. Twenty-seven percent of standard PDP members remained in the doughnut hole for 5 to 6 months. More than 1 in 5 members spent 6 months or more in the doughnut hole.

The covariates in the study include age, sex, classes of medication, and number of medications. Age and sex had no significance as confounders. Also, the classes of medications were not statistically significantly different between study groups, thus showing evidence of equality in study populations with respect to the types of medications members are taking. No single class of medication was more prevalent in 1 study group compared with the other. Statistical significance was demonstrated by the covariate of the number of medications a person was taking. However, that was due to its relative exponential relationship to alterations in prescription fulfillment decisions. Thus, it was not surprising that the number of medications was an influential covariate regarding members’ decisions to fill prescriptions. However, logistic regression models showed that age, sex, and classes of medications played no significant role as confounders. There was no significant relationship between the classes of medications and prescription fulfillment decisions (OR 0.92, 95% CI 0.80-1.07).

Despite the enhanced plan group being half the size of the standard PDP group, they filled almost as many prescriptions. Members with the enhanced plan were more likely to always fill their prescription medications. Restricting the analysis to situations where the impact of the fulfillment decision was greater than 9%, 22% of the standard PDP members delayed their prescriptions, 13.4% switched medications, 10% delayed and stopped some medications, and more than 9% switched and delayed their medications. The enhanced plan group, which had no such doughnut hole, only delayed filling prescriptions in 10.8% of cases and switched their prescriptions in 10% of cases. Some studies in seniors report compliance rates comparable to those found in this study.13,22,25,26 One such study of people 65 years and older from 8 different states showed that 22% of participants did not fill prescriptions 1 or more times in a full year because of costs; 23% of respondents reported that they skipped doses of medications.27 These fulfillment rates are very similar to our findings in the standard PDP group. While similar comparative research is sparse, if not nonexistent, results are mixed. Some studies do show in nearly similar populations that an increase in drug expenditure relates to an increase in drug use.12,21,28 However, other studies show that as copayments rise, drug utilization will decrease.14-16 In a study comparable to this one, the Kaiser Foundation showed that of 8 drug classes analyzed, 20% of enrollees in the doughnut hole in 2007 either stopped taking a medication in that drug class, reduced their medication use (eg, skipped doses), or switched to a different medication in that class when they reached the gap. Specifically, 15% stopped taking their medication, 5% switched to an alternative drug, and 1% reduced their medication use.22 Similarly, Hsu et al found that 15% of study participants switched to a cheaper medication and 8% ceased taking their medication.26 These findings are similar to ours.

The prescription fulfillment decisions made by members in the doughnut hole tended to affect more than 1 class of medication. The decision not to fill a prescription affected 1 class of medication for approximately 28% of doughnut hole members. Alarmingly, 43.2% of standard PDP members’ prescription fulfillment decisions affected 2 or more classes of therapy. Most importantly, the data showed that people who experienced the doughnut hole were 1.4 times more likely to switch 1 or more of their prescription medications (OR 1.39, 95% CI 0.86-2.27), 2.3 times more likely to delay 1 or more of their medications (OR 2.33, 95% CI 1.48-3.66), 3.3 times more likely to switch and delay 1 or more medications (OR 3.29, 95% CI 1.53-7.05), and 4.9 times more likely to switch, stop, and delay their medications (OR 4.88, 95% CI 1.47-16.23).

Health Implications

Standard PDP members remained in the doughnut hole for a mean of 144.7 (SD = 50.72) days. A 2007 Kaiser study found that members in the doughnut hole made changes to their drug use regimen, including stopping their medications altogether, and although ramifications from cessation of prescription medications may not be initially apparent, in a disease like osteoporosis, such a decision could increase the risk for hip fractures.22 The doughnut hole may add to the complexity of noncompliant morbidity and mortality rates. Seniors in this study were 2 times more likely to delay and more than 4 times more likely to switch, stop, and delay their medications due to the doughnut hole. Stopping and/or delaying prescription medications may lead to poorer health outcomes.9,28-33 Tamblyn et al found an 88% increase in “hospitalizations, institutionalization in nursing homes, and even death, and a 78% increase in the rate of emergency room visits.”32 The intent of the Part D program should be centered on providing seniors with prescription medications to improve or stabilize their health condition and prevent unnecessary healthcare costs such as hospitalizations, emergency department visits, and physician appointments. The findings in this study indicate the need for follow-on Part D prescription fulfillment studies, Part D health outcomes research that looks at all services impacted by the doughnut hole type design (eg, physician, hospital), and specific class-by-class Part D doughnut hole analyses.

Policy Ramifications
Our findings also have policy implications for the Medicare Part D program. The Centers for Medicare & Medicaid Services may want to consider other benefit design options that support more steady, but higher, cost-sharing designs to allow greater consistency for beneficiaries throughout a year. Notwithstanding private or public payer interests and control, our research at least notes concern over a benefit design that inflicts a large and sudden patient cost responsibility midstream through prescribed therapies. Part D premiums rise each year, placing an increasing financial burden on seniors enrolled in a design that already asks them to participate in volatile cash outlay responsibilities. Some plans have dramatically increased premiums (some by more than 100%) from 2006 to the present.34 Apparently, concerns over the cost ramifications of the Part D doughnut hole resonated with private and public stakeholders. The government and the pharmaceutical industry together heard the cries of doughnut hole alumni, as evidenced by the Patient Protection and Affordable Care Act. The law has provisions for the purpose of reducing and ultimately closing the doughnut hole by 2020.35 Members will first receive a $250 rebate and 50% of the doughnut hole will be removed by 2011 through pharmaceutical manufacturers’ contributions. Regardless of any particular viewpoint about “ownership” of a senior’s drug benefit, the doughnut hole design of the current Part D PDP alters prescription fulfillment decision making.

Limitations
There are several limitations to this analysis. Income data were not available from the HMO pharmacy claims database. Seniors’ level of income has been determined to be a potential confounding factor in similar research.4,5,7-9,36 Using a cross-sectional study design could allow for recording of participant income levels; however, the benefits of obtaining actual information on prescription activity of members via a retrospective, longitudinal sample seemed to far outweigh any benefit from the factoring of income levels. Patient interactions or communications with physicians specifically regarding cost-sharing responsibilities for medications also have been noted to be a contributing factor in prescription compliance. This information was not available with pharmacy claims data and thus was excluded from our analysis.

One other important limitation is that patients might have appeared to stop medications if they stored pills, split pills, or otherwise altered their intended prescription regimen as a means to offset costs. However, members’ behavior before and after the doughnut hole phase helped determine whether cash alternative or self-altering medication regimens were being utilized. Patients would have had to have purchased a medication from an external source such as a Canadian online pharmacy to avoid recording of their prescription transaction. Another delimiting factor of the study is whether the first year of Part D utilization was indicative of utilization in future years. Extrapolation of findings from one year’s benefits to the subsequent years may not be appropriate. Still, our research should encourage similar analyses for subsequent years.

CONCLUSIONS
This pharmacy adjudication, prescription fulfillment study demonstrates how the sudden drop of prescription drug coverage does have an effect on what patients decide to do with their prescriptions. This can have longterm implications for patients’ health. Seniors must be cautious about disrupting the appropriate timing of filling their prescriptions without talking about this issue with their healthcare providers. The current study has shown the effect of the doughnut hole on prescription fulfillment decisions, but not the clinical ramifications of such behaviors. Further studies are necessary.