Cost-Lowering Strategies Used by Medicare Beneficiaries Who Exceed Drug Benefit Caps and Have a Gap in Drug Coverage FREE

The new national Medicare prescription drug benefit will provide important coverage for Medicare beneficiaries. However, it is unknown how one aspect of the currently proposed benefit design, a gap in coverage once a beneficiary's total drug costs reach a $2250 annual cap, will affect medication use.¹ This coverage gap occurs because the benefit pays 75% of prescription costs for beneficiaries until their total drug costs exceed $2250 and then provides no further coverage for the rest of the year unless total drug costs exceed $5100.¹

Patients who exceed the cap and whose drug costs fall within this gap in coverage may be at risk for decreasing their use of medication. These patients must pay the entire cost of their prescriptions, and it is well documented that Medicare beneficiaries without drug benefits use less medication.^{2 - 8} However, the temporary nature of the increases in out-of-pocket cost during this coverage gap may also mean that some beneficiaries may be willing and able to pay higher drug expenses until their benefits restart the following year. Thus the exact impact of caps and this coverage gap on medication use is uncertain.

Beneficiaries enrolled in Medicare managed care (Medicare + Choice) face a similar coverage gap. Two thirds of the 4.5 million Medicare + Choice enrollees receive drug benefits through their health plan.⁹ The vast majority of these drug benefits have dollar limits or caps.⁹ Beneficiaries who exceed caps also pay the entire cost of their prescriptions for the remainder of the year. Unlike the national Medicare drug benefit, these caps are usually applied to a health plan's share of drug costs rather than to total drug costs (ie, health plan costs plus patient's share of costs). However, Medicare beneficiaries with prescription expenditures high enough to exceed Medicare + Choice caps are also likely to exceed caps in the new national Medicare drug benefit.¹⁰ The impact of caps and coverage gaps on medication use is also important since at common Medicare + Choice cap levels such as $750 per year, 1 in 5 Medicare patients exceeds his/her capped benefits before the end of the year.¹⁰ In addition, the escalating costs of providing drug benefits have led many Medicare + Choice plans to lower the generosity of their benefit caps in recent years. Approximately 80% of Medicare + Choice caps were set at $750 or less in 2002 compared with 50% of caps in 2001.⁹

With the risk for coverage gaps in the newly adopted national Medicare drug benefit and the decreasing capped drug benefits for Medicare + Choice beneficiaries, understanding how caps and coverage gaps affect medication use is critical to designing adequate drug benefits for Medicare beneficiaries. Since many of the medications most contributing to patients' drug expenditures are for the treatment of important chronic health problems such as hypertension, hypercholesterolemia, heart disease, and diabetes,¹⁰ decreases in medication use due to cost after exceeding the cap may also negatively impact health. We carried out a cross-sectional survey of Medicare + Choice beneficiaries with capped drug benefits to determine their cost-cutting strategies, the type of medications involved, and the financial burden of drug costs for patients who exceeded caps and had a gap in coverage compared with those patients who did not exceed their caps.

Participants were enrolled in a Medicare + Choice plan that had 438 802 Medicare enrollees in 1 state who filled at least 1 prescription in 2001 (86% of plan members). We used the different caps offered in different counties as a natural experiment to test the effects of exceeding a cap. Our study is a comparison of survey responses from seniors with different caps on the health plan's share of medication costs. Study participants (n = 665) exceeded a $750 or $1200 yearly cap on the health plan's share of medication costs in 2001 and had coverage gaps of 75 to 180 days. Control participants (n = 643) had $2000 caps, which they did not exceed. Study and control participants were matched by average total drug expenditures per month. The study compared survey responses from participants aged 65 years and older with these different caps. Eligible participants were surveyed by telephone (average time to complete, 20 minutes) and initial nonrespondents were mailed a written survey (average time to complete, 15 minutes). The survey was completed from March through July 2002, with two thirds of participants interviewed in May or earlier.

Plan members had annual caps of $750 to $2000 on the plan's share of prescription costs. If a beneficiary had a $750 per year cap, the plan paid for its share of prescription costs (actual prescription cost minus the beneficiary's co-payment), until the sum of these costs reached $750. Afterwards, beneficiaries paid the entire cost of their prescriptions out of pocket for the remainder of the year. Prescription co-payments ranged from $7 to $30 per 30-day prescription, there were no deductibles, and all beneficiaries had the same drug formulary. Beneficiaries who chose to use the plan's pharmacy mail order service paid twice the monthly co-payment for 3 months of prescriptions. The health plan set cap levels based on where a beneficiary lived because reimbursement rates from Medicare varied geographically according to the average spending by Medicare fee-for-service members in each region. Risk of selection bias was low since beneficiaries could not select higher or lower caps by paying different premiums.

Medicare beneficiaries were eligible if they were aged 65 years or older, enrolled continuously in the plan in 2001, lived in an area where the health plan continued to provide coverage in 2002, and did not have dual coverage with Medicaid. A total of 82% of the plan's members who filled a prescription for at least 1 medication in 2001 (70% of total plan members) met these criteria. Health plan members had drug benefit caps ranging from $750 to $2000 per year applied to the plan's share of prescription costs. Using pharmacy claims, we identified whether health plan members exceeded their cap in 2001. We obtained study group participants (ie, those who exceeded the cap) from the pool of members with $750 and $1200 caps and control group participants (ie, those who did not exceed the cap) from the pool of members with a $2000 cap. Overall, 22%, 11%, and 4% of beneficiaries in the health plan with $750, $1200, and $2000 caps, respectively, exceeded their caps in 2001.¹⁰

To ensure that study group participants had coverage gaps for a significant period of time, we sampled patients who exceeded a $750 or $1200 cap for 75 or more days. This represented 58% and 38% of the plan's members who exceeded the $750 and $1200 caps, respectively. We further restricted study participants to those exceeding a $750 cap for 180 or fewer days or a $1200 cap for 100 or fewer days, in order to obtain control participants with similar average monthly total drug expenditures from the $2000 cap benefit who did not exceed their cap.

To ensure that control participants would be similar to study group participants in age and average monthly total drug expenditures, we used the following stratified sampling strategy. First, we calculated the median age of study participants only. Next, we calculated the average monthly total drug expenditures (costs paid by the patient plus costs paid by the plan) for study group participants for the period prior to exceeding the cap, and determined their 10th, 25th, 50th, 75th, and 90th percentile values. Then we calculated the average monthly total drug expenditures for all eligible control participants based on the entire year, and used these values to categorize all control participants by age (2 groups) and average monthly total drug expenditures (6 groups). Lastly, we sampled our control group participants from each of these 12 categories in the same proportions as study group participants.

We measured strategies for controlling costs after exceeding the cap by asking patients if they lowered the out-of-pocket cost of their medications in 2001 by using any 1 of 6 behaviors: using a current medication less often than prescribed (eg, skipping or decreasing a dose), stopping a medication altogether, not starting a new medication, switching to a less expensive medication, using free medication samples, and/or using others' medications. Patients were asked to answer "yes" only if the cost of the prescription was the primary motivation for adopting a strategy. We defined patients to have "decreased medication use because of cost" if they reported using a current medication less often, stopping a medication, and/or not starting a new medication. We also asked patients about a number of additional strategies to decrease cost that may or may not have led to lower use, such as the ability to obtain discounts on prescriptions (eg, shopping around or buying medications outside of the United States).

For participants who completed the telephone version of the survey (n = 887) and who reported using less medication, stopping, and/or not starting, we asked which drug(s) were involved in these decisions. Interviewers assisted participants with drug identification by using medication lists based on each participant's 2001 pharmacy claims data. Participants who completed written versions of the survey were not included because although they often reported cost-cutting strategies, they often did not write down which drugs were involved. To identify which drug therapeutic classes were most often affected, we classified each unique drug into a therapeutic class. For instance, we divided medications for peptic ulcer and gastroesophageal reflux disease into 3 therapeutic classes: proton pump inhibitors, H₂ blockers, and coating agents. We counted the number of times that each therapeutic class was affected by a strategy that decreased medication use (using less medication, stopping, not starting) and ranked each class from highest to lowest counts.

We measured financial burden resulting from drug costs by patients' self-reported ease or difficulty in paying for their prescriptions, and whether drug costs affected their ability to pay for enjoyable activities, pay rent or bills, or pay for nonprescription medical care in 2001. We asked patients if they had supplemental drug insurance outside of the health plan.

We calculated patients' monthly out-of-pocket costs (costs paid by the patient) by summing their co-payments as recorded in pharmacy claims. We calculated patients' monthly total drug costs (costs paid by patient plus costs paid by the plan) by summing the co-payments and actual costs paid by the plan as recorded in pharmacy claims. For study participants, we determined the average cost for the period before they exceeded the cap; for control participants, we averaged costs across the entire year. Pharmacy claims were incomplete after study group participants exceeded the cap and so we do not know their annual out-of-pocket or total drug costs.

Patients reported their age, education, marital status, race/ethnicity, household size, and annual household income. Patients were classified as white if they selected only the "white/Caucasian" category when asked about race; patients who selected at least 1 race category other than "white/Caucasian" were classified as "nonwhite" in data analyses. We assessed race because previous studies such as that of Laira et al¹¹ have found race to be an independent predictor of whether patients decreased medication use due to cost. We used the median income and household size of those patients who reported actual income (49%) to impute the income of those participants who reported only an income range (38%) or no income information (13%). We measured health status by patients' self-reported global rating of health (poor, fair, good, very good, excellent) and ability to perform instrumental activities of daily living (IADLs), such as preparing meals and managing money.¹² We also asked patients if they had ever been diagnosed with any of 10 common health problems (hypertension, hypercholesterolemia, coronary heart disease, gastroesophageal reflux disease, diabetes mellitus, emphysema/asthma, depression, osteoporosis, cancer [other than skin], and stroke) and asked about pain symptoms and use of prescription pain medications.

We compared baseline demographic and health characteristics for participants in the study and control groups using t tests and χ² analyses. We carried out multivariate logistic regression to determine if exceeding the cap was a significant predictor of adopting strategies to lower out-of-pocket medication costs after controlling for variables that might independently affect medication use, such as age (10-year intervals), sex, education, marital status, income, supplemental drug insurance, health status, functional independence in IADLs, and number of health conditions. Significance tests were based on the regression coefficient for each cost-cutting strategy. The significance level for all statistical analyses was 2-sided P<.05. We expressed effect sizes as predicted percentages of adoption of each strategy in the study and control groups, after adjustment for differences in covariates between the groups.¹³

We began with a random stratified sample of 2344 potentially eligible participants (1172 study, 1172 control) identified from pharmacy claims data. We excluded 345 patients (15%) contacted by telephone or mail because of death, illness, disenrollment from the health plan in 2001, or inability to complete a written or telephone survey in English. Of the remaining 1999 participants, 1308 (65%) completed a survey, 454 (23%) refused, and contact could not be established with 237 (12%). Counting all refusals and noncontacts as potentially eligible, this resulted in a response rate of 65% (1308/1999) for potentially eligible participants. Response rates were similar between study participants in the $750 group (372/551 [68%]), $1200 group (293/443 [66%]), and the $2000 control group (643/1005 [64%]). Two thirds of the participants completed the survey by telephone and one third by written questionnaire.

Respondents were on average 75.9 years old; 64% were women, 88% were white, 54% were currently married, and 83% had a high school degree or higher. The majority of participants (68%) had annual household incomes of $30 000 or less in 2001. Nearly all participants (92%) reported that they did not have supplemental insurance outside of the health plan that covered outpatient medications.

Study group participants were more likely than control group participants to be married (57% vs 51%, respectively; P = .02), have slightly higher average income ($32 818 vs $29 800, P = .02), and were less likely to be women (59% vs 68%, P<.001) (Table 1). For patients who could not be contacted or who refused to participate, only age and sex were available from claims data. On average, nonrespondents were slightly older than respondents (77.1 vs 75.9 years, respectively; P<.001) but were similar in the percentage of women (62% vs 64%, P = .42).

Study and control group participants did not differ significantly in health status, functional limitations in IADLs, prevalence of most medical comorbid conditions, severity of pain, or frequency of use of pain medications (Table 1). Control participants had slightly higher rates of hypertension, pulmonary disease, and mean number of health conditions. As reflected by their large drug expenditures, participants overall had a high incidence of common health problems such as hypertension (73%), hypercholesterolemia (66%), coronary heart disease (45%), gastroesophageal reflux disease (39%), diabetes mellitus (29%), and pulmonary disease (emphysema or asthma) (25%). A substantial percentage of participants rated their health as fair (29%) or poor (11%). Their mean score for Lawton and Brody IADLs was 7.0 out of 8.0 (full functionality), indicating 1 functional limitation on average. A total of 42% of patients reported moderate or severe pain symptoms, and 23% used prescription pain medications on most days or almost every day.

Our matching strategy resulted in very little difference in the average monthly total drug expenditures between study participants prior to exceeding the cap ($209/mo) and control participants throughout the year ($207/mo) (Table 2). For study group participants, out-of-pocket costs averaged $91 per month before exceeding the cap. Control group participants had slightly smaller prescription co-payments and averaged $72 per month in out-of-pocket prescription costs throughout the year.

In multivariate analyses adjusted for age, sex, education, ethnicity, annual household income, general health status, functional limitations in IADLs, number of health problems, written vs telephone survey, and whether or not participants provided income data, exceeding the cap was found to be a significant independent risk factor for several medication strategies (Table 3). Study participants (ie, those exceeding the caps) were significantly more likely than control participants (ie, those not exceeding the caps) to adopt at least 1 strategy that decreased medication use, ie, using less medication, stopping, or not starting (24% vs 16%, P<.001). This was due mainly to their using their current medications less often than prescribed (18% vs 10%, P<.001). Both groups had similar rates of stopping medications completely (8% for both, P = .86) and of not starting medications (6% vs 5%, P = .39). Study participants were also more likely than control participants to switch medications (15% vs 9%, P = .002), and to use free drug samples (34% vs 27%, P = .006) to reduce costs. Very few in either group took someone else's medication because of cost (2% vs 1%, P = .26). Other independent risk factors for adopting at least 1 strategy that decreased medication use included younger age, lower income, and greater number of health problems. For 4 out of 6 strategies (not starting medication, switching, using samples, using someone else's medication), having a greater number of health problems was a significant risk factor for adopting that strategy.

Participants also were asked about strategies to obtain lower prices on medications. Nearly two thirds of participants in both groups mentioned using mail order services, which was highly encouraged by the plan. In adjusted models, patients who exceeded the cap were more likely to call pharmacies to find the best price (46% vs 29%, P<.001) and were more likely to obtain discounts for being 65 years or older (12% vs 7%, P = .003). However, very few patients in either group obtained discounts for being a member of Medicare (10% vs 7%, P = .13) or discounts through pharmaceutical companies (2% vs 1%, P = .11). Only 3% of participants in each group said they had successfully bought medications outside of the United States, although many more participants in both groups reported anecdotally that they had tried or were considering this option.

A total of 197 of 887 telephone participants (22%) decreased their medication use (used less medication, stopped, or did not start) because of cost. The majority of telephone participants (171/197 [87%]) who decreased medication use due to cost (used less medication, stopped, did not start) were able to name at least 1 drug involved. This resulted in 323 drug-reducing events and represented 112 drugs and 56 therapeutic classes. Three fourths of all drug-reducing events belonged to the 20 therapeutic classes most often affected (Table 4). Twelve of these top 20 classes were for chronic illness such as hypertension (diuretics, angiotensin-converting enzyme inhibitors, calcium channel blockers), hyperlipidemia (3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors), and emphysema or asthma (bronchodilators, steroid inhalers, ipratropium inhalers). Several therapeutic classes were for health problems that may have been short-term or chronic, such as pain (cyclooxygenase 2 inhibitors, narcotics, nonsteroidal anti-inflammatory drugs), peptic ulcer or gastroesophageal reflux disease (proton pump inhibitors, H₂ blockers), or allergy (nonsedating antihistamines, nasal steroids). In 2004, generic equivalents or over-the-counter medications were available in 15 of the 20 top therapeutic classes.

Half of all participants said that paying for their medications was somewhat difficult (35%) or very difficult (15%) in 2001, and the percentage was significantly higher for study participants than for control participants (62% vs 37%, P<.001) (Table 5). Paying for prescriptions affected both study and control participants' ability to participate in enjoyable or leisure activities (46% vs 34%, P<.001) and caused difficulty in paying rent or bills (24% vs 18%, P = .02). Similar percentages of study and control participants reported not obtaining nonprescription medical care because of prescription costs (15% vs 13%, P = .20), having to borrow money or ask for help paying for prescriptions (12% vs 10%, P = .23), or both.

In our study, beneficiaries who exceeded the cap 75 to 180 days before the end of the year reported more often that they reduced their use of medication. Beneficiaries exceeding the cap often decreased their use of essential medications and were more likely to experience difficulty paying for medications (62% vs 37%). Caps are an important benefit design feature that allows insurers to provide drug coverage for many beneficiaries at affordable costs. Previously we found that cap generosity had a significant impact on the number of beneficiaries who exceeded their cap, with 22%, 11%, and 4% of patients exceeding $750, $1200, and $2000 caps, respectively.¹⁰ The study herein showed that exceeding the cap for 2.5 to 6 months also affected important medication use and self-reported financial burden from drug costs. The results herein also confirm previous findings that Medicare beneficiaries who are most vulnerable to medication nonuse due to cost include patients with lower incomes,^{2 ,4 - 6 ,8} younger age,^{6 ,14 - 15} and greater number of health problems or poorer health.^{4 ,14 - 15} We refer readers to Craig et al¹⁵ for greater details and a summary of these studies.

Our results provide insight into how beneficiaries' medication use may be affected if they fall into the "donut hole" in the national Medicare drug benefit,¹⁶ ie, the gap in which no coverage is provided for total drug expenditures between $2250 and $5100.¹ The $2250 cap may seem much more generous than the caps ($750, $1200, $2000) found in our Medicare + Choice population. However, the national prescription benefit will apply annual caps to total prescription costs rather than to only the health plan's share of costs. In fact, two thirds of the beneficiaries in our study who exceeded their cap had average monthly total prescription costs of $187.50 or more ($2250/12) prior to exceeding the cap and would be at risk for having a gap in coverage under the national drug benefit. Although beneficiaries with low income and many health conditions are most likely to be affected, other design features such as premiums, deductibles, catastrophic coverage, and low-income subsidies will also affect beneficiaries' out-of-pocket drug costs and medication use.

Our study also shows that when participants aged 65 or older reduced their medication use because of cost, important medications for treating disabling and potentially life-threatening illnesses such as hypertension, hyperlipidemia, depression, osteoporosis, prevention of stroke, asthma/emphysema, and diabetes were affected, as were medications used for treating "less essential" health problems such as allergies. These findings are similar to those from the study by Goldman et al¹⁷ of cost-related medication decreases in younger patients (aged <65 y) in which many of the same therapeutic classes are also affected (nonsteroidal anti-inflammatory drugs, antihistamines, antihyperlipidemics, antiulcerants, antiasthmatics, and antihypertensives). It is also consistent with our previous study that showed that 15 of the top 20 medications accounting for the prescription costs of patients who exceeded the cap were for treating chronic conditions such as hypertension, heart disease, and diabetes.¹⁰ It remains to be determined how patients made these cost-cutting decisions (eg, why they choose to decrease use of 1 medication but not another) and if they involved their clinicians. Many of the therapeutic classes in which patients decreased their medication use have at least 1 generic or over-the-counter drug available. Although most direct generic equivalents are considered to be equally efficacious with the brand-name drugs, (eg, lovastatin/Mevacor), substitutions within a therapeutic class (eg, lovastatin for brand-name atorvastatin) involve more detailed trade-offs in efficacy, adverse effects, and dosage convenience. It will be important to determine how much clinicians could have helped patients decrease their medication costs by switching to less expensive medications or stopping ineffective medications. The actual health impact of exceeding the cap remains unknown. Lowering the doses of some medications (eg, those for stroke prevention and diabetes) may have greater immediate and long-term health consequences than stopping other medications (eg, those for treating allergies).

Of note, patients who exceeded the cap were more industrious about using strategies that reduced the price they paid for a medication, such as calling pharmacies to find the best price, switching to lower-cost medications, and using free samples. However, only 2% to 9% of participants reported success in obtaining discount medications from pharmaceutical companies, buying medications outside of the United States, or obtaining "senior" or "Medicare" discounts from pharmacies. It remains to be seen how well strategies such as the new Medicare discount cards and the debated reimportation of prescription drugs from overseas can be used on an individual basis by Medicare beneficiaries to increase their access to prescriptions. Even with increased efforts to obtain lower drug costs, half of the participants reported difficulty paying for prescriptions. Many states (eg, Maine)^{18 - 19} have followed the lead of health plans, the Department of Defense, and the Department of Veterans Affairs in negotiating lower drug prices for their constituents. These state programs have faced legal challenges from pharmaceutical companies,¹⁸ and whether the national Medicare drug benefit will adopt a similar strategy for Medicare beneficiaries is under intense debate.

Our findings also show the importance of a matched control group in before-and-after studies. A surprisingly high percentage (16%) of control participants in our study, who had coverage throughout the year, adopted at least 1 strategy that decreased medication use because of cost. It may be that these control participants reduced their medication use in order to avoid exceeding their benefit cap. Studies that represent all Medicare beneficiaries have shown much lower rates (2% to 4%),^{5 ,11} but our findings are more consistent with the 8% to 42% of more vulnerable Medicare beneficiaries (eg, those with lower income, less-generous drug coverage, greater number of health problems, higher out-of-pocket costs) who decrease medication use due to cost.^{2 ,4 - 5 ,11 ,13 ,20} Our control group was selected from beneficiaries with higher out-of-pocket costs (mean, $72/mo), which is an independent risk factor for decreased medication use.^{2 ,5}

Important limitations to this study are that we sampled participants from a single Medicare + Choice plan in a single state, and medication strategies were self-reported and subject to recall bias. We also only included those who had high enough prescription expenditures to exceed their cap for 75 to 180 days. However, these are the subgroup of Medicare beneficiaries who will also be affected by the coverage gap in the national Medicare drug benefit. It is reasonable to assume that those participants with shorter gaps in coverage will be at lower risk for decreasing their use of medication due to cost, while the impact for those with longer coverage gaps will be greater. Beneficiaries who had dual coverage with Medicaid were excluded, but this represented only 7% of health plan members.¹⁰ The strength of this study is that we had control group participants who were similar to study participants in terms of demographic characteristics and average monthly total prescription expenditures to isolate the impact of exceeding caps on medication use.

This study emphasizes the tradeoffs involved in setting cap generosity for Medicare beneficiaries. Lower caps may allow drug benefits to be offered to a greater number of beneficiaries by decreasing the cost of providing such a benefit, but exceeding the cap can increase the risk for decreasing essential medication use due to cost and increase financial burden. Even with a generous cap, some beneficiaries will exceed their drug benefits and those who are ill or have many chronic health problems are most likely to have high prescription expenditures and to be at risk for exceeding the cap. Therefore clinicians, insurers, and the public will need to consider how to best balance the cost and benefits of their medications to maximize the benefit from such plans.