|
Use Of Antihypertensive Drugs By Medicare Enrollees: Does Type Of Drug Coverage Matter?
Research has demonstrated that Medicare beneficiaries with drug coverage consume more clinically essential drugs. However, generosity of coverage varies considerably across beneficiaries. This study examines the association between types of drug coverage and the consumption and cost per tablet of essential antihypertensive medications among beneficiaries with hypertension. The findings indicate that while both state- and employer-sponsored drug coverage are associated with greater consumption of antihypertensive drugs and lower out-of-pocket costs per tablet, private supplemental coverage is not associated with greater use and is associated with only slightly lower out-of-pocket costs than among noncovered beneficiaries.
Prescription drugs are a key component of health care for many elderly and disabled Americans. While not all drug use is appropriate, drug therapy is highly effective, and even life saving, in the management of chronic diseases such as hypertension.1 Unfortunately, the benefits of drug therapy may be beyond the reach of those without adequate coverage for outpatient drug costs.2 Recent studies have estimated that two-thirds of Medicare beneficiaries have drug coverage from some source and that having coverage is associated with greater use of essential and nonessential medications.3 However, some programs strictly limit the types and numbers of drugs covered or impose high cost-sharing requirements. 4 In Medicaid such limitations have been shown to increase adverse health outcomes and use of costly institutional services among frail elderly and low-income enrollees.5 This study examines the association between different types of private and public drug coverage, drug costs, and drug use for a national sample of Medicare beneficiaries with hypertension. Jan Blustein recently examined the association between use of antihypertensive drugs by Medicare beneficiaries and drug coverage from any source.6 Given considerable variation in the generosity of different types of drug coverage, it is important to extend Blustein’s work by comparing the effects of different types of coverage on the use of these clinically essential drugs.
Data source and sample. Our data source was the Medicare Current Beneficiary Survey (MCBS) Cost and Use Files for 1995. The MCBS is a national, longitudinal survey of approximately 12,000 Medicare beneficiaries in which respondents are periodically interviewed regarding their health status, health care use, and access. 7 MCBS respondents are asked to keep written records of health service use (such as receipts and prescriptions) to support their memory of events. At the time this study was conducted, the 1995 MCBS contained the most recent and comprehensive data available on drugs used by the Medicare population. Using the MCBS data, we estimated that 86.6 percent of noninstitutionalized beneficiaries used at least one outpatient prescription drug in 1995. Enrollees spent an average of $605 on medications, and the average number of prescriptions filled was 18.7. Of the 12,096 beneficiaries included in the 1995MCBS, 53 percent reported having hypertension. Prevalence of self-reported hypertension was fairly consistent across insurance status in the MCBS: 52 percent of those with Medicare health maintenance organization (HMO), self-purchased, or employer insurance; 58 percent of Medicaid enrollees; and 49 percent of those with Medicare fee-for-service (FFS) only.8 Our study cohort included MCBS respondents with self-reported hypertension. We excluded beneficiaries with any Department of Veterans Affairs (VA) enrollment during the year because these persons were a highly select group with complete outpatient drug coverage. We also excluded Medicare HMO enrollees, who constituted too small a subgroup in 1995 to compare those with and without drug coverage, and beneficiaries who were ever institutionalized or who were not continuously enrolled in Medicare during the past year. After these exclusions, the sample size of hypertensive patients was 4,439. Approximately 93 percent (4,120) provided detailed information regarding drug use and costs in the prescribed medicine component of the MCBS. Definition of drug coverage. Beneficiaries were classified as having private drug coverage if they reported having drug coverage through a private health plan.9 Medicaid enrollment was ascertained using both administrative and self-reported enrollment, or by seeing if any drugs had been paid for by Medicaid during the year. Individuals were defined as having state-based drug coverage if they identified other public programs as having paid for their drugs. We assigned beneficiaries with more than one type of drug coverage to the source that paid the majority. However, those with any drug expenditures paid by Medicaid were automatically classified as Medicaid enrollees. Beneficiaries without drug coverage from any source and without private insurance were classified as Medicare FFS only. Coverage type was indicated by dichotomous variables. Dependent variables. Our first two outcomes of interest were total and antihypertensive drug use. We chose to examine spending as a measure of use rather than tablets, because data on treatment regimens were not available. However, to verify that spending differences represented differences in use rather than solely differences in price, we also estimated the average number of total and antihypertensive tablets used, by insurance and drug coverage type. To illustrate the relative financial burden of drug use by coverage type, we examined beneficiaries’ out-of-pocket payments. Our final outcomes of interest were the total amount and the amount paid out of pocket per tablet for each of four antihypertensive drugs (diuretics, including loop diuretics; calcium channel blockers; angiotensin converting enzyme [ACE] inhibitors; and beta blockers) by drug coverage type, where antihypertensive drugs were identified using the U.S. Pharmacopeia Drug Information (USP DI), 1999.10 Control variables. Control variables included age (beneficiaries under age sixty-five were classified as disabled), sex, marital status, race, education, insurance status, region of residence, income, self-reported health status, functional status, disability status, number of chronic conditions other than hypertension that might be related to antihypertensive drug use (such as diabetes, angina, arthritis, and other heart conditions), and whether a myocardial infarction or a stroke had ever been reported.11 Functional status was based on the number of self-reported limitations in the following activities of daily living (ADLs): bathing/showering, dressing, eating, getting in or out of a chair, walking, and using the toilet. We also controlled for whether the beneficiary had Part coverage, urban or rural residence, or end-stage renal disease (ESRD), factors that could influence the degree of contact with the health care system. To control maximally for confounding, variables were included in the empirical model if they were associated with the study outcomes at the .25 level in univariate analyses. Statistical analysis. The analysis was conducted in three stages. In the first stage we performed univariate comparisons of drug use in dollars, number of tablets, and percentage paid out of pocket by drug coverage type, class, and several demographic characteristics. In the second stage we estimated the effect of type of coverage on drug use in dollars, using multivariate analyses to control for differences in coverage groups that might influence the level of use. Because the vast majority of Medicaid beneficiaries were in a different income category than the Medicare FFS-only group (the natural control group for those with public drug coverage), they were not included in this phase of the analysis. We first estimated the probability of using any antihypertensive drugs with logit regression. 12 We then estimated total and antihypertensive drug use, for those reporting drug use, using weighted ordinary least squares (WOLS), which allowed us to account for the complex sampling design of the MCBS.13 Both the logit and WOLS models were stratified by insurance type (public only versus private).14 We ran both fully specified and more parsimonious models based on a stepwise model building procedure in which covariates were included based on p values of .25. As the primary variables of interest, the appropriate drug coverage variables were included in the public and private insurance models, and their significance was determined at the .05 level for both the full and reduced models to assure the robustness of the findings to the stepwise modeling procedure. In the third stage, using those with Medicare FFS only as a reference group, we tested the differences in average cost per tablet by antihypertensive drug class for each drug coverage type.
Characteristics of the sample. Approximately 65 percent of the study sample had some drug coverage in 1995: 28 percent Medicaid, 26 percent employer sponsored, 6 percent state based, and 6 percent self-purchased (Exhibit 1  ). Those without private insurance were more likely to be disabled and to have poor self-reported health and functional status. Those with Medicaid coverage were considerably more likely to be poor. Those with private insurance, regardless of drug coverage, were more likely to be white, have a high school diploma, and have higher incomes.
Drug use by insurance status and drug coverage. Although they have similar illness profiles, the state drug coverage group and the Medicare FFS-only group represent the high and low extremes in overall drug use (Exhibit 2 ).On average, those with Medicare FFS only spent $508 less on drugs than did those with state drug coverage. Utilization for persons in Medicaid was similar to that of persons in state programs. Those with employer-sponsored coverage spent an average of $956, compared with $836 for those with self- purchased coverage and $672 for those with private insurance but no drug coverage. Differences by age, race, sex, and health status highlight other important determinants of prescription drug use.
Costs. Predictably, the Medicaid and state drug coverage groups paid much less out of pocket for drugs than did the Medicare FFS only group. Those with employer-sponsored coverage had out-of-pocket payments similar to those of the state drug coverage group, while the private insurance group without drug coverage had similar percentages paid out of pocket as the Medicare FFS-only group. Those with self-purchased drug coverage on average paid 64 percent of their drug costs out of pocket. In contrast to total drug costs, the percentage paid out of pocket does not vary much by demographic group within each coverage category. Number of tablets. Examination of the number of tablets used for those with some tablet use (not shown) verified that drug expenditures reflect differences in actual use. Those with Medicaid and state drug coverage exhibited the highest use (1,379 tablets and 1,239 tablets, respectively) and the Medicare FFS-only group, the lowest (915 tablets). For those with private insurance, persons with employer drug coverage purchased an average of 1,212 tablets, compared with 1,143 for the self-purchased drug coverage group and 1,046 for those with no drug coverage.
Antihypertensive drugs.
Similar spending patterns by drug coverage type exist for antihypertensive drugs (Exhibit 3
Those with employer, state, and Medicaid drug coverage purchased a higher average number of antihypertensive tablets during the year than did those with private insurance without drug coverage or with self-purchased drug coverage. The Medicare FFS-only group purchased the fewest tablets during the year. Effect of coverage on antihypertensive drug use. The effects of drug coverage on propensity to use antihypertensives were generally low. Controlling for demographic and health characteristics, enrollment in state drug coverage programs was marginally significantly associated with increased use of any antihypertensive medications (OR = 1.53; p = .065), compared with those with Medicare FFS only. Compared with the private insurance group without drug coverage, employer coverage was significantly associated with increased odds of antihypertensive use (OR = 1.26; p = .043), but self-purchased coverage was not (OR = 1.15; p = .432). Effect of coverage on adjusted total drug use. State drug coverage was associated with average yearly drug spending per drug recipient of $1,067, compared with $583 for those with Medicare FFS only (t = 8.25; p < .01). For the privately insured, employer drug coverage was associated with adjusted yearly drug spending of $908, $272 higher than for those with private insurance and no drug coverage (t = 5.86; p < .01). Spending among those with self-purchased coverage averaged $795, or $159 more than the privately insured without drug coverage spent (t = 2.58; p < .01). Effect of coverage on adjusted antihypertensive drug use. Among those without private health insurance, state drug coverage was associated with adjusted average spending for antihypertensive drugs of $302, compared to $191 for those with Medicare FFS only (t = 1.97; p < .05). Among those with private insurance, employer drug coverage was associated with antihypertensive drug spending of $280, or $55 more than those with private insurance and no drug coverage spent (t = 3.94; p < .01). Those with self-purchased drug coverage spent an average of $26 more ($251) than did those with private insurance and no drug coverage, but this difference was not significant at the .05 level (t = 1.92). Effect of drug coverage on cost of antihypertensives. While overall drug costs per tablet were relatively uniform across coverage groups, the out-of-pocket costs per tablet were much lower for those with Medicaid, state, or employer drug coverage.15 For example, individuals in these three coverage groups spent at least 60 percent less per tablet out of pocket for calcium channel blockers than did those with Medicare FFS only (p < .01). Persons with self-purchased drug coverage did not have significantly lower out-of-pocket spending per tablet for these drugs.
Under- and overestimation. Our reliance on self-reported data in this study may have led to underestimation of drug usage.16 We also may have overestimated out-of-pocket payments for beneficiaries who had not been reimbursed by their drug coverage providers at the time the survey was completed. Both underestimation of use and overestimation of out-of-pocket costs would have led us to underestimate rather than overestimate the effect of coverage. Further, while drug coverage itself may induce physicians to overprescribe and influence patients to demand unnecessary treatment, antihypertensive medications are essential, life-saving treatments that are known to be underused in a variety of settings.17 It is likely that we have underestimated somewhat the prevalence of drug coverage because of our reliance on self-reports.18 Also, we did not distinguish between part-year and full-year coverage.19 Overall, misclassification of drug coverage should have made it more difficult for us to detect an effect of coverage on use. Selection bias. Like all cross-sectional analyses, our study may contain some degree of selection bias in that persons who anticipate high use and who have sufficient resources are more likely to purchase drug coverage.20 Similarly, those unable to afford private insurance who experience high drug costs are more likely to qualify for state coverage programs.21 We attempted to reduce the effect of possible selection bias by stratifying analyses by insurance type and by including important covariates such as health status. Further, we estimated the degree of possible observable bias in the group without private insurance to be approximately 3 percent by calculating drug use for a hypothetical individual with drug coverage but with socioeconomic and health characteristics similar to those of someone without coverage. Nevertheless, because this sensitivity analysis could not account for selection bias due to unobserved factors, the estimates should be interpreted with caution. However, these findings are supported by evidence from well-controlled studies showing that the use of clinically essential drugs is reduced when coverage is limited.22
The evidence provided in this study indicates that for Medicare beneficiaries with hypertension, drug coverage and, more importantly, specific types of drug coverage have large impacts on overall levels of drug use, as well as on the use of clinically essential antihypertensive medications that can prevent adverse outcomes such as myocardial infarction and stroke. These findings demonstrate the importance of distinguishing between different types of coverage in estimating the effect of coverage on use. Moreover, this study provides evidence that those without drug coverage are paying more for essential medications. The magnitude of the association between type of drug coverage and drug use, especially of essential antihypertensive agents, suggests that we should be quite concerned about underuse of clinically important drugs in patients without coverage or with self-purchased coverage. Undoubtedly, chronically ill and near-poor beneficiaries exhibit the greatest need for drug coverage under Medicare.23 While there appears to be an emerging consensus that some drug coverage under Medicare is desirable, our study findings indicate that proposed drug coverage benefit plans with high cost sharing but no catastrophic coverage are unlikely to provide chronically ill and low-income beneficiaries with the protection they need from the high cost of prescription drugs. Ultimately, the costs of a Medicare drug benefit must be measured against decreases in future adverse health events and the increased quality of life that will result if we increase access to essential prescription drugs.
The authors are all members of the Drug Policy Research Group in the Department of Ambulatory Care and Prevention at Harvard Medical School and Harvard Pilgrim Health Care, in Boston, Massachusetts. This research was supported by the National Institute on Aging (Grant no. 5R01AG14474-03) and the Harvard Pilgrim Health Care Foundation. Alyce Adams was supported by a postdoctoral fellowship from the Agency for Healthcare Research and Quality. The authors thank Donald Rubin, Jennifer Hill, Richard Frank, and JosephKalt for their extensive comments on earlier versions of this manuscript. Also, they are indebted to the Health Care Financing Administration for the data, John Poisal for providing information regarding the Medicare Current Beneficiary Survey, and Robert Le Cates for invaluable assistance with the exhibits.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Study Methods
Study Results
