Objective:To examine the association between generosity of drug coverage and essential cardiovascular medication use among retired seniors. Study Design: Retrospective analysis of the 1997 to 2000 Medicare Current Beneficiary Survey, a nationally representative survey of the Medicare population.
Methods:The study examined community-dwelling fee-for-service Medicare beneficiaries aged 65 years or older with retiree health insurance and with coronary heart disease and hyper-lipidemia (n = 1220) or congestive heart failure (n = 1147). Generosity of drug coverage was defined as the percentage of the beneficiary’s annual drug expenditures paid by the employer. Dependent variables were any statin use for the group with coronary heart disease and hyperlipi-demia and any angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use in the group with congestive heart failure. Logistic regression analyses estimated the adjusted odds of essential medication use by generosity category in each disease group. We estimated the extent to which medication use would change if generosity levels moved to those under standard Medicare Part D levels.
Results: The overall prevalence of statin use was 64.1%, and that of angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use was 50.0%. In both disease groups, retirees in the less generous drug coverage categories had significantly lower adjusted odds of use than retirees with the most generous drug benefits (ie, covering >76% of annual drug expenditures). Overall, the shift to a standard Medicare Part D structure would result in mean declines of 8.4% in statin use and 5.2% in angiotensin-converting enzyme inhibitor or angiotensin receptor blocker use. Retirees with the most generous drug coverage face twice the mean decline in drug use.
Conclusion: Retirees who already have generous drug benefits from their employers may be placed at risk for decreased utilization of effective medications due to any future scaling back of retiree drug coverage.
(Am J Manag Care. 2007;13:425-431)
One third of the Medicare beneficiaries now have drug coverage through employer-sponsored health plans, and most of these are offered by their own or their spouse's former employer or union. Most of these beneficiaries receive more generous coverage under these plans than they would under the new Medicare prescription drug coverage (Part D) which began January 1, 2006.1 Competitive labor markets have provided sufficient incentive in the past few decades for employers to voluntarily offer retiree health benefits with generous drug coverage to workers and their families.2 However, because an accounting rule change in the early 1990s requiring employers to include retiree health benefit liabilities on their financial statements, coupled with a double-digit growth in healthcare costs (particularly for pharmaceuticals), many employers have been overhauling their retiree health and drug benefits to control, reduce, or eliminate their retiree health benefit liabilities.3 With the enactment of the Medicare Prescription Drug Improvement and Modernization Act, additional (albeit unintended) incentives for terminating or scaling back retiree drug coverage have been created. First, the existence of Medicare Part D as an alternative source of drug coverage for retirees would reduce political resistance for employers considering dropping coverage altogether. In fact, the Congressional Budget Office4 estimated that 17% of retirees were likely to lose their employer-sponsored drug benefits; more recent estimates from the Department of Health and Human Services suggest that it might be double that.5 Second, in light of the ongoing trends of reductions in generosity of retiree drug coverage due to business, accounting, and cost factors,3 employers planning to continue offering benefits may now have Medicare Part D as a concrete example of a "floor" or minimum level of generosity for retiree drug benefits. This is because the US Congress has built in financial incentives in terms of direct subsidies of up to $88 billion to employers for maintaining their retiree drug coverage. To receive a tax-free subsidy (equal to 28% of total drug costs between $250 and $5000 for each covered retiree), the employer would need to offer employment-based retiree drug coverage that is at least actuarially equivalent to the standard Medicare Part D coverage. Hence, employers planning to continue offering drug coverage could over time reduce the generosity of their benefits all the way to Medicare Part D levels and still receive the benefit of the federal subsidies.
Regardless of whether employers will drop coverage to standard Medicare Part D levels or not, recent surveys suggest that employers are scaling back prescription coverage through increased drug cost-sharing for retirees and are likely to continue doing so during the long term.3,6,7 The question of how substantial increases in out-of-pocket shares for prescription expenditures could potentially affect medication use, particularly among Medicare beneficiaries with retiree health insurance, has not been studied in detail. This study takes advantage of the existing variation in retiree drug benefits to examine the potential effect of significant changes in prescription cost-sharing on use of essential chronic medications among retired seniors. To reduce the complexity in testing this relationship, we specifically examine the association between the generosity of drug coverage and use of 2 effective cardiovascular medications selected from well-established clinical guidelines,8-10 namely, (1) statin use among retirees with coronary heart disease (CHD) and hyperlipidemia and (2) angiotensin-converting enzyme (ACE) inhibitor or angiotensin receptor blocker (ARB) use among retirees with congestive heart failure (CHF).
The data source used for the study was the 1997 to 2000 Medicare Current Beneficiary Survey (MCBS) linked with Medicare claims. The MCBS is a nationally representative survey of Medicare beneficiaries conducted under the auspices of the Centers for Medicare & Medicaid Services. It includes detailed information on beneficiaries' demographics, health and functional status, health insurance coverage, and annual use and spending for all healthcare services, including prescription drugs. The health insurance section of the survey captures the source of supplemental insurance (eg, current employer, former employer, self-purchased, Medicaid, and others) and whether these plans offer prescription drug coverage. The prescription drug use section captures drug names during the thrice-yearly interviews of respondents who are instructed to keep a medication log, save insurance receipts, and show the interviewers all of their medication containers.
The sample frame consists of community-dwelling fee-for-service Medicare beneficiaries aged 65 years or older with supplemental health insurance from their own or their spouse's former employer from 1997 to 2000. We limited the sample to fee-for-service beneficiaries because availability of Medicare claims was necessary to identify disease conditions. From the
1997 to 2000 pooled cross sections, we selected 2 groups for analysis based on 1 or more inpatient or outpatient claims for (1) CHD and hyperlipidemia (n = 1220) and (2) CHF (n = 1147) (an appendix containing the International Classification of Diseases, Ninth Revision, Clinical Modification and Current Procedural Terminology codes is available online at www.ajmc.com). For the primary analysis, we excluded Medicare retirees with additional drug coverage from sources other than an employer (153 retirees from the group with CHD and hyperlipidemia and 167 retirees from the group with CHF). However, secondary analysis without excluding this group indicated that our results are not sensitive to this sample exclusion criterion, and results only for the primary sample are presented.
Our dependent variable was an indicator of the presence of any use of the essential cardiovascular medications. The indicator represented any statin use for the group with CHD and hyperlipidemia and any ACE inhibitor or ARB use in the group with CHF.
The main independent variable of interest was the generosity of drug coverage. This variable was defined as the percentage of the beneficiary's annual drug expenditures paid by the employer. We grouped beneficiaries into the following 4 generosity categories: 0%, 1% to 50%, 51% to 75%, and 76% to 100%. The 0% category primarily represents beneficiaries with no retiree drug benefits but also includes a small group with such limited drug benefits that none of their drug spending was paid by the employer.
Control variables included socioeconomic and demographic factors such as age, sex, education, income, metropolitan status, and geographic region of residence. We also controlled for smoking status and whether the beneficiary had ever visited a cardiologist. In addition, we included the following clinical characteristics: numbers of physician visits and hospitalizations for the specific study condition, presence of contraindication to the essential medication, cardiovascular risk factors likely to increase essential medication use, number of chronic conditions other than cardiovascular disease, and self-reported health status (the details are available in an appendix from the author). Survey year of observation was also used as a control.
The adjusted odds of receiving the essential cardiovascular medication by coverage generosity level were estimated using separate logistic regression analyses in each disease group. Although the logistic regression analyses controlled for differences in available demographic and clinical measures across the generosity of drug coverage categories, it is possible that there remain unmeasured factors predictive of generosity of drug coverage and of essential medication use resulting from the mechanism of drug coverage selection. We test for the presence of this bias using the Durbin-Wu-Hausman test.11 To aid in the policy interpretation of the adjusted odds ratios, we estimated the extent to which essential medication use would change if generosity levels for all retirees moved from existing levels to Medicare Part D levels in 2006. This required estimating generosity levels for all retirees under standard Medicare Part D. We did this in 2 steps. First, given that existing levels of drug expenditures are from the 1997 to 2000 MCBS data and that the standard Medicare Part D benefit limits and thresholds were designed for 2006, we projected annual drug spending from each year of the 1997 to 2000 MCBS to 2006 using historical and projected per capita drug-spending data from the National Health Accounts (and assuming no Medicare Part D implementation).12 These forecasts represent what one would expect retiree drug spending to be in 2006 if Medicare Part D had not been enacted. Second, we imposed the standard Medicare Part D benefit structure on these spending estimates to compute the new levels of generosity of drug coverage. The standard Medicare Part D plan in 2006 required all (except low-income) beneficiaries to pay an annual deductible of $250, 25% coinsurance for drug spending from $250 to $2250, 100% coinsurance for the next $2850 of drug expenses (ie, the "doughnut hole"), and 5% coinsurance after $3600 in out-of-pocket costs (ie, $5100 in total expenses). Hence, for example, if prescription spending in 2006 for a retiree is projected to be $4000, we calculated the mean out-of-pocket spending under a Medicare Part D–like structure as the sum of the $250 deductible, $500 in coinsurance, and $1750 in the doughnut hole, for a total out-of-pocket payment of $2500. Therefore, the generosity of coverage under the Medicare Part D–like structure is computed to be 37.5% (ie, the percentage of drug spending paid by a third party), and the retiree was classified under the 1% to 50% generosity category. We then estimated what essential medication use would be under these new generosity levels. All analyses were performed using STATA software version 8.0 (StataCorp LP, College Station, Tex) and accounted for the complex survey sample design.
Retirees in both disease groups received generous drug benefits from their employer (Table 1). More than three quarters of annual drug expenditures were paid by employer insurance for 54.8% of retirees with CHD and hyperlipidemia and for 44.6% of retirees with CHF. Only 7.1% to 12.2% of retirees in both groups received no retiree coverage for their annual drug expenditures (ie, the 0% group). The differences in the socioeconomic, demographic, and clinical characteristics of the retirees across the generosity of drug coverage categories for the group with CHD and hyperlipidemia and the group with CHF are available in an appendix from the author.
The overall prevalence of statin use was 64.1% among retirees with CHD and hyperlipidemia and that of ACE inhibitor or ARB use was 50.0% among retirees with CHF (Table 1, last 2 columns). Essential cardiovascular medication use was lower for those with less generosity of drug coverage.
For instance, while 72.6% of the most generous drug coverage category (76%-100%) received statins, the prevalence dropped significantly (from 45.6% to 55.5%) in the less generous drug coverage categories. A similar pattern of ACE inhibitor or ARB use by drug coverage generosity was observed among retirees with CHF.
Table 2 gives the odds ratios for essential cardiovascular medication use by drug coverage generosity and disease group. Given the observed differences in characteristics by drug coverage generosity, logistic regression analyses were used to estimate adjusted odds ratios in addition to unadjusted odds ratios. In general, the magnitude and significance of unadjusted odds ratios remained similar even after adjusting for all covariates. The results confirm that less generosity of drug coverage was associated with lower odds of using an essential cardiovascular medication. In both disease groups, retirees in the less generous drug coverage categories (0%, 1%-50%, and 51%-75%) had significantly lower adjusted odds of essential cardiovascular medication use than retirees with the most generous drug benefits (76%-100%). Also, results of the Durbin-Wu-Hausman test were not statistically significant in either disease group, thus ruling out bias due to unmeasured confounders.