| Chapter 3 | Table of Contents | Appendix B |
| Cover Page | Section 1 | Section 3 | Section 4 | Bibliography | Abstracts |
Spending on outpatient prescription drugs by non-institutionalized Medicare beneficiaries reached $25.1 billion in 19961, the latest year for which accurate statistics are available (Poisal, et al, forthcoming). Four years earlier in 1992, beneficiaries spent $16.2 billion (Laschober and Olin, 1996). Not only does this represent a substantial growth in spending (55 percent), but the rate of increase is also rising, from approximately 9 percent annually between 1992 and 1994 to over 14 percent annually between 1994 and 1996 (Laschober, 1997; Olin and Liu, 1998; Poisal, et al., forthcoming). In per capita terms, the growth rate over this period (44 percent) was somewhat lower (rising from $468 to $674), reflecting the fact that the size of the Medicare population has also continued to grow.
More current data are available on prescription drug costs for beneficiaries enrolled in Medicare HMOs. According to industry surveys, the average per member per month (PMPM) expenditure on prescription drugs rose from $37.53 in 1995 to $52.50 in 1998, a 40 percent increase in just three years (CibaGeneva, 1996; Novartis, 1999). The increase would have been even steeper had HMOs not raised patient copays by more than 25 percent over the same period (CibaGeneva, 1996; Novartis, 1999).
The issue of rising drug costs is at the core of the national debate over whether a prescription drug benefit should be added to Medicare coverage. On the one hand, rising costs are cited as a main reason why Medicare should cover drugs. Despite the fact that drug coverage of Medicare beneficiaries expanded during the first half of the 1990s (see section 2.3); the burden of out-of-pocket drug spending has also grown (see section 2.5). On the other hand, the growth in drug spending will make any meaningful Medicare drug benefit expensive to finance. From either perspective, there is a clear need to answer the question: What factors explain drug spending by Medicare beneficiaries?
The answers lie at both the individual level and the market level. The literature on individual determinants of drug spending by Medicare beneficiaries identifies seven factors: (1) health status, (2) functional impairment, (3) age, (4) gender, (5) race, (6) income, and (7) insurance coverage. Health status is an obvious factor. Beneficiaries reporting excellent health spent an average of $363 on prescription medicines in 1996 compared to $1,107 for persons reporting poor health (Poisal, et al., forthcoming). For beneficiaries with no functional impairment, drug spending per person was $582 versus $1,000 for persons with one or more functional impairments (Poisal, et al., forthcoming). Other highlights from the latest report of Medicare beneficiary drug spending (Poisal, et al., forthcoming) include the following:
The finding that spending levels differ by socio-demographic status and insurance coverage are not unexpected, but simple descriptive comparisons may mask the true determinants of drug spending behavior. In one of the very few multivariate studies of prescription spending by aged Medicare beneficiaries, Lillard, et al. (1999) find that private health insurance, urban residence, and ill health are positive predictors of drug spending, while age and education level are negative predictors, when other factors are equal. Some factors that appear to be strong correlates of drug spending in descriptive comparisons lack predictive power in a multivariate framework, including gender, income, wealth, and Medicaid coverage.
The determinants of drug spending described above operate primarily through the drug selection decisions made by Medicare beneficiaries and their physicians. However, drug prices can be an important factor, particularly for those beneficiaries without stable drug coverage. The question most frequently asked in this regard is: Do Medicare beneficiaries pay more for their prescriptions than other Americans ? The underlying issue is whether the prices for drugs commonly used to treat age-related chronic conditions are either higher or increasing at a faster rate than drugs used for other diseases. Research on this question is difficult because drug prices vary greatly depending on the payor, and private payors rarely disclose their pricing practices. For this reason, most attempts to compare prices across age and disease groups are based on standardized price listings such as the Average Wholesale Prices (AWP), the Producer Price Index (PPI), or the average retail price (ARP). Highlights from three recent studies show that:
These studies do not resolve the issue of whether age-related price discrimination exists in the market for drug products because of what happens to prices at different points in the chain of distribution. For example, Berndt et al., (1998) found no systematic age-related difference in price increases at the juncture between manufacturer and wholesaler for three classes of drugs (antibiotics, antidepressants, and calcium channel blockers), but from the wholesaler to the pharmacy one of the three classes (antibiotics) showed more rapid growth for the elderly than the young. Furthermore, at the point where the patient makes the purchase in the pharmacy, relative age-related prices for antidepressants favored the young rather than the aged (Berndt, et al., 1998).
Medicare pays for prescription drugs only if they are administered in institutional settings (e.g., hospital or nursing home) or belong to several special drug categories such as immunosuppressives, erythropoiten, oral anti-cancer drugs, hemophilia clotting factors, and some vaccines. The vast majority of outpatient prescription drugs are not reimbursable under the standard package of Medicare benefits. Supplemental insurance thus plays an important role in understanding how much Medicare beneficiaries spend on prescription medications. What are the characteristics of Medicare beneficiaries with and without drug coverage?
More than 90 percent of Medicare beneficiaries have some form of supplemental health insurance coverage, either by qualifying for public assistance programs or by enrolling in private insurance policies through employee benefits, retiree plans or individually purchased plans. There is no centralized database that records the characteristics of these supplemental policies or of the beneficiaries who obtain them. Instead, our information on coverage rates depends almost entirely on survey data. The best nationally representative survey (and the one used by most researchers in this area) is the Medicare Current Beneficiary Survey (MCBS). Since 1992, longitudinal samples of approximately 12,000 aged and disabled beneficiaries have been queried three times a year about their supplemental insurance (including prescription drug coverage), health status, access to care, and use and cost of all health services. The scientifically rigorous methods used in selecting the respondents and soliciting accurate information make MCBS data the most reliable single source for estimating prescription drug coverage for the Medicare population.
According to 1995 MCBS data, 65.2 percent of community dwelling Medicare beneficiaries had some prescription drug coverage for at least part of the year (Poisal, et al., 1999). Over 70 percent of the disabled Medicare population had drug coverage in that year. Rates were lower among elderly beneficiaries, ranging from 66 percent of those aged 65-69 to just 60 percent of those over age 85 (Poisal, et al., 1999). Males and females have similar coverage rates (70 and 68 percent, respectively in 1995). White beneficiaries were less likely to have coverage (64 percent) than blacks (69 percent) or persons of other races (75 percent). Coverage rates rise with income from 65 percent among beneficiaries with annual incomes below $10,000 to 72 percent among beneficiaries with incomes above $30,000 per year (Poisal, et al., 1999). Beneficiaries who reside in metropolitan areas are much more likely to have prescription coverage (69 percent) compared to those who do not (54 percent) (Poisal et al., 1999). In 1995, 64 percent of beneficiaries reporting excellent health had drug coverage compared to 69 percent of those in poor health (Poisal, et al., 1999). The differences are much greater when health status is measured by counts of chronic conditions. In 1996, 65 percent of Medicare beneficiaries reporting having none of ten common chronic conditions had some prescription coverage compared to 77 percent of beneficiaries with five or more of these conditions (Stuart, et al., 2000).
The reasons for these differences are varied and complex. For beneficiaries above the poverty line, coverage rates rise with income. This could be due to higher purchasing power and the fact that retirees with higher incomes are more likely to qualify for employer-sponsored health insurance. The reason that prescription coverage is higher among nonwhites is probably not because of race per se, but rather because nonwhite retirees have lower incomes and therefore higher Medicaid eligibility rates. The negative correlation of coverage rates and self-reported health status might mean that those who need coverage the most are the most likely to seek it out, but it could also capture the impact of income and Medicaid eligibility status.
To date there has been just a single study of the determinants of prescription drug coverage for Medicare beneficiaries (Lillard, et al., 1997). It showed no evidence of adverse selection into prescription coverage, but the data for this study are quite old (1968-1990). There is a substantial body of research on selection into Medicare supplemental policies independent of benefit coverage. Recent studies of Medicare HMO enrollment (Hellinger, 1995; PPRC, 1996; Riley, 1996; Hamilton, 1999; Call, et al) find strong evidence of favorable rather than adverse selection. Indications of adverse selection in the Medigap market have been found by Ettner, (1997), Wolfe and Goddeeris (1991), and Atherly (1999). Other researchers (Cartwright, Hu, and Huang, 1992; Hurd and McGarry, 1997; Lillard, et al., 1999) find none. Most researchers assume that there is no selection into retiree plans given the nature of plan sponsorship, but Atherly (1999) finds evidence of adverse selection in this market. These thoroughly mixed findings do little to clarify the issue of possible adverse selection into prescription coverage.
Although the determinants of who has prescription benefits remain unclear, there is no question that drug coverage rates among Medicare beneficiaries have been rising. In 1992, 57 percent of beneficiaries had prescription coverage (Poisal, et al., 1999). By 1996 the rate was 68.8 percent (Poisal, et al., forthcoming). The movement of Medicare beneficiaries into HMO plans is one reason for the trend. Enrollment into Medicare risk plans increased 30 percent between 1995 and 1996 and HMO enrollees are more likely to have prescription coverage (94.6 percent) than those in any other private supplemental insurance plan. However, these prescription benefits are often inadequate.
From a policy standpoint the critical unanswered question is: Will the trends in prescription coverage of Medicare beneficiaries witnessed in the early and mid 1990s continue? There are no reliable data on prescription coverage rates since 1996. However, it appears unlikely that the trend maintained its steep upward course in the late 1990s. Recent reports show that large employers are scaling back retiree benefits (Hewitt Associates, 1999) that HMOs are not renewing Medicare risk contracts (Barents Group, 1999), and that some Medicare risk plans are dropping drug benefits, lowering payment caps, or imposing stiff premium surcharges for such coverage (Gold, 1999). This is such a volatile mix of factors, that any forecast of beneficiary drug coverage rates is fraught with uncertainty in both short-term and long-term projections.
This uncertainty has its roots in dynamic patterns of coverage that became evident in the mid 1990s. Profiles of typical Medicare supplemental insurance portfolios present a dynamic picture of beneficiaries gaining and losing prescription coverage, holding multiple plans, and changing plans during the year. Poisal et al., (forthcoming) found that nearly nine percent (over 3 million people) of the total non-institutionalized Medicare population switched plans at least once during 1996. This figure is slightly higher than the previous year’s estimate of eight percent (Davis, et al., 1999). However, both of these estimates understate the true degree of change. A month-to-month examination of supplemental insurance in 1996 by Stuart et al. (2000) shows that prescription coverage is often fragmented and non-continuous. They found that 18.9 percent (about 6.3 million people) of beneficiaries with full-year Medicare enrollment had drug benefits for only part of the year, averaging around six months of coverage. Having coverage at some point in the year may be a valuable benefit, but the continuity of that coverage is equally important.
Surely the most compelling issue for policy makers regarding a possible Medicare drug benefit is the question of cost. For any given proposal, actuaries can readily determine that part of the program cost which represents a shift in financial responsibility from beneficiaries or other third party payors. It is much more difficult to predict the impact that coverage will have on the utilization patterns of previously uninsured beneficiaries. This issue has two important components: Does drug coverage encourage Medicare beneficiaries to use more outpatient prescription medications? Does drug coverage influence the type and cost of the medications prescribed and used? The available literature on these topics is primarily descriptive, representing cross-tabulations of utilization and expenditure by presence and source of coverage. Taken at face value, the reported differences suggest that insurance plays a very significant role in the drug utilization decisions of Medicare beneficiaries and their prescribers. A sampling of these findings tells the story:
The differences in utilization and annual drug spending by insurance status extend to all major population subgroups categorized by age, income, and health status:
The source of coverage also appears to matter:
There have been no systematic studies of the relationship between drug coverage and the composition of drug utilization, i.e., the selection of specific therapeutic products for a given disease. However, there is a widely held belief that insurance affords better access to newer, more expensive therapeutic agents. One recent finding consistent with this view shows that Medicare beneficiaries in 1996 who reported good to excellent health utilized drug products with an average cost of $40.70 if insured, but only $28.74 if they were not covered for prescription drugs (Stuart, et al., 2000).
All of these reported findings are subject to a major caveat. Although the differences in drug cost and use associated with insurance coverage appear both large and consistent, they are purely descriptive and do not necessarily mean that drug coverage causes or induces beneficiaries to use more. Indeed, some analysts argue that the differences are primarily due to the fact that sicker individuals who are heavy users of prescription medication are more likely to purchase drug coverage or enroll in public plans that provide it. While it is true that beneficiaries in poorer health are more likely to have prescription coverage (as noted in section 2.3 above), that does not rule out an independent insurance effect on demand.
There is a small but growing analytical literature that attempts to disentangle the various influences on the demand for prescription drugs by the elderly. Early studies by Long and Gordon (1989) and Long (1989) support the view that observed differences in drug use for persons with drug coverage under Medigap plans are the result of adverse selection rather than an insurance effect. These authors find no insurance effect among Medicare beneficiaries with employer coverage (where there is presumably little adverse selection).
More recent work supports the opposite view that drug coverage does induce additional usage. Based on data from a survey of Pennsylvania elderly, Coulson and Stuart (1995) and Coulson et al. (1995) found that prescription coverage increased drug use by approximately 3 percent for every 10 percent reduction in out-of-pocket cost to beneficiaries, all else being equal. These authors also found that the primary effect of drug coverage is to induce additional persons to use prescription medicine as opposed to increasing prescription use among users. This “hurdle” phenomenon associated with patient cost-sharing was first observed in the famous Rand Health Insurance Experiment of the late 1970’s and early 1980’s (Manning et al., 1989). It has since been observed in a number of recent studies of the insurance effect of drug coverage on elderly Medicare beneficiaries (Stuart and Grana, 1998; Stuart and Zacker, 1999; Ya-chen, 1999; Lillard, Rogowski, and Kingston, 1999). These studies all reach the same basic conclusion—drug coverage increases the probability of drug use but has minimal effect on the number of prescriptions filled by users. This conclusion is strengthened by the fact that the studies focused on different groups of elderly using a variety of databases. Stuart and Zacker (1999) used MCBS data to assess the impact of drug copayments on elderly Medicaid/Medicare dual eligibles. Stuart and Grana (1998) analyzed survey data on medicine use for 23 common health problems reported by a large sample of Pennsylvania elderly. Ya-chen (1999) used pharmacy records from the Dialysis and Morbidity and Mortality Study to study non-Medicare covered drug use by ESRD patients. Lillard, Regowski and Kingston (1999) analyzed survey data on prescription use in the RAND Elderly Health Supplement to the 1990 Panel Study of Income Dynamics.
The empirical estimates of insurance effects produced by these studies vary within a relatively narrow range. The addition of drug coverage is estimated to increase the probability of any prescription being filled by between 4 and 16 percent depending on population subgroup and generosity of drug coverage. The effect of adding a comprehensive drug benefit to Medicare is estimated to increase overall drug spending among elderly beneficiaries by between 20 percent (Lillard, Regowski and Kingston, 1999) to 34 percent (Coulson and Stuart, 1995).
Medicare beneficiaries who have no prescription coverage must, by definition, bear the entire cost of their drug purchases out of pocket. However, the obverse does not follow; no prescription plan currently available to Medicare beneficiaries covers the entire cost of outpatient prescription drugs. Were Medicare to adopt a drug benefit, it would presumably provide the greatest benefit to those who are presently most vulnerable to high out-of-pocket spending for drugs. The obvious question for policy makers is: What factors make Medicare beneficiaries most vulnerable to high out-of-pocket drug costs?
The literature on this issue identifies five principle factors: (1) total spending on prescription drugs, (2) insurance coverage, (3) income, (4) health, and (5) functional status. According to the Health Care Financing Administration, half of all drug spending by Medicare beneficiaries in 1995 was paid for out of pocket (Poisal, et al, 1999). The single greatest predictor of high out-of-pocket spending, is high total spending. The National Academy of Social Insurance estimates that 14 percent of Medicare beneficiaries had drug spending exceeding $2,000 from all payment sources in 1999 (Gluck, 1999). Almost a third of these individuals also had out-of-pocket drug spending greater than $2,000 for the year (Gluck, 1999).
The presence of prescription coverage mitigates against high out-of-pocket drug costs, but does not preclude them. A recent study published by AARP, estimates that Medicare beneficiaries with some drug coverage in 1999 spent just 3 percent of income on out-of-pocket drug purchases compared to 6 percent for beneficiaries without coverage (Gibson, et al, 1999). The income of Medicare beneficiaries is often lower in proportion to other populations, thus the out of pocket expenses are relatively burdensome. However, 42 percent of those with out-of-pocket drug spending exceeding $1,000 also maintained some prescription coverage. The principal reason for this is the limited scope of benefits offered by health plans that supplement Medicare, particularly individual Medigap policies (Gibson, et al, 1999). The three Medigap policies with drug coverage all have a $250 deductible followed by a 50 percent coinsurance up to maximum payment caps of $1,250 or $3,000. Stuart et al. (2000) found that Medicare beneficiaries with continuous coverage spent about half as much out-of-pocket in 1996 on prescription drugs ($219) as persons with part year coverage ($424) or no coverage at all ($468).
There is surprisingly little variation in total or out-of-pocket drug spending by income level (Poisal, et al., 1999; Gibson, et al, 1999; Poisal, et al., 1999 forthcoming). In 1995, for example, beneficiaries with annual incomes below $10,000 spent an average of $707 for prescription drugs of which $206 represented out-of-pocket payments. Those with annual incomes over $30,000 spent an average of $642 of which $226 was out of pocket (Poisal, et al., 1999). However, this similarity in dollar spending means that low-income beneficiaries spend proportionately more of their income on prescription drugs. Gibson, et al., (1999) estimate that beneficiaries below the poverty level in 1999 ($8,760 for individuals and $11,334 for couples) spent nine percent of their incomes on drugs compared to just 2 percent for those with income above 400 percent of the poverty line.
Twenty-seven percent of Medicare beneficiaries report being in fair or poor health. These individuals are the heaviest consumers of prescription medicines of any group and bear the highest average burden of out-of-pocket costs. Gibson, et al. (1999) estimate that they paid $590 per-person in out-of-pocket prescription drug payments in 1999, representing seven percent of annual income. Being uninsured and in poor health magnifies the burden. Stuart, et al. (2000) show that beneficiaries self-reporting poor health and no drug coverage spent $732 out of pocket in 1996 compared to just $318 for those with continuous drug coverage for the year.
As this last statistic demonstrates, the burden of out-of-pocket drug expenses varies widely depending on the constellation of attributes that beneficiaries manifest. Because these attributes tend to be inter-correlated, to understand the independent effect of each predictor requires multivariate analysis. To date there are no published papers in this area. A study by Crystal et al. (in press) examines predictors of aggregate out-of-pocket spending by Medicare beneficiaries for all types of health care combined using standard multivariate techniques. They find that self-reported health status, number of medical conditions reported, privately purchased Medicare supplemental health insurance, and functional impairment all increase the level of predicted out-of-pocket payments, while only HMO participation predicts lower payments (Crystal, et al., in press). Given that outpatient prescription drugs comprise approximately a third of all out-of-pocket health spending by beneficiaries (Crystal, et al., in press), one would expect these multivariate findings to hold true for prescription drug spending by itself.
This section reviews what is known as well as that which remains unanswered in the context of the bolded questions presented throughout the discussion above. There are some common themes. The first crosscutting theme is that the literature focuses on drug use and spending almost as if “prescription drugs” were a single homogeneous product. A few studies have analyzed utilization at the level of the therapeutic class, but there has been virtually no research on the nexus between insurance coverage and appropriate or inappropriate drug use. The presumption is that lack of drug coverage is a barrier to appropriate drug therapy and impacts access to medications. It is also possible that lack of coverage reduces exposure to inappropriate drug use and thereby reduces the incidence of adverse drug events. More research is needed to clarify this issue as adverse drug events contribute to additional health expenditures through hospitalizations to treat these events as well as the costly cycle of prescribing additional drugs to treat the side effects of many previously prescribed drugs. Further research is also warranted to identify the relevant elements in insurance contracts that demonstrate empirical evidence of assuring quality and safety in drug use.
A second common theme in the literature is the focus on aged Medicare beneficiaries to the near exclusion of the disabled and beneficiaries with ESRD. Nearly all of the studies reviewed here focus strictly on the elderly or simply subsume the disabled in with the aged. Twelve percent of Medicare beneficiaries or approximately 5 million individuals receive their program entitlement from disability insurance. These beneficiaries are (by definition) all under age 65 and have very different characteristics and prescription drug needs than their aged peers. For example, there is a much higher prevalence of mental illness among the disabled compared to the elderly, making behavioral medicine a critical issue for this population group. Also, the disabled do not have the same access to private Medigap supplements as those over 65 years of age.
Another common characteristic of the literature on drug use and spending by Medicare beneficiaries is that it is largely descriptive rather than analytical. Making public policy on the basis of means, frequency distributions, and cross-tabulations can be perilous given the interdependence of factors relating to drug coverage, use, and spending in this population. There is a need for additional multivariate research in each of the areas addressed in this review as identified below.
What factors explain drug spending by Medicare beneficiaries? This issue is paid considerable attention in the literature, but the studies are predominantly descriptive. We know that beneficiaries in poor health are high spenders, but we have little knowledge of the drugs they use or the illnesses they treat with them. We know that white beneficiaries spend more in total for prescription medicine than blacks, but we do not know whether the explanation is racial or is due to some other co-varying factors such as income or education.
Do Medicare beneficiaries pay more for their prescriptions than other Americans do? There is scant research on this topic and even fewer answers. The question is important, as limitations in prescription drug policies place many beneficiaries with private insurance at risk of significant out-of-pocket purchases. The “best price” practices reserved for large purchasers of prescription drugs means that Medicare beneficiaries who buy prescriptions without the benefit of group volume discounts will tend to pay higher prices. It would be useful to know whether Medicare beneficiaries with supplemental prescription coverage are reaching maximum expenditure caps earlier in the year because of drug price increases. Research is also needed to examine the extent of generic drug use by age and how this is influenced by rising drug prices and drug benefit design.
What are the characteristics of Medicare beneficiaries with and without drug coverage? We know who has drug coverage but not why. Here is another instance where lack of analytic studies hinders interpretation of the descriptive statistics about which beneficiaries maintain coverage and which do not. A particularly important question yet to be answered is whether beneficiaries with the greatest need for coverage find it easy or difficult to obtain it. An equally important question is how stable the prescription coverage is for those who have it.
Will the trends in prescription coverage of Medicare beneficiaries witnessed in the early and mid 1990s continue? Most commentators think not, but the evidence to date is strictly anecdotal. Reliance on the Medicare Current Beneficiary Survey as the primary data source on Medicare drug coverage, means that the most up to date information is almost four years old before it is published.
Does drug coverage encourage Medicare beneficiaries to use more outpatient prescription medications? The answer is yes, but additional studies are critical to improve the precision of the estimated size of the “insurance effect.” Also needed is research that examines the characteristics of beneficiaries most strongly influenced by prescription coverage as well as drug regimens they use (or would use if given the opportunity). We need to know whether the observed difference in utilization rates between the insured and the uninsured arises because the uninsured fail to fill prescriptions or because the insured are simply prescribed more medications.
Does drug coverage influence the type and cost of the medications used? There is little known about this important issue. Research is needed on the potential influence that drug coverage plays on the patient-physician relationship. For example, do Medicare beneficiaries without drug coverage visit the doctor less often (and thereby avoid the problem of getting prescriptions that they would not fill anyway)? Are doctors less likely to prescribe medications to those without coverage? Are they more likely to prescribe less expensive generic drugs to these clients?
What factors make Medicare beneficiaries most vulnerable to high out-of-pocket drug costs? This is perhaps the best understood of all the major issues raised in this review. Although the research to date is descriptive, the range of comparisons is large enough to give a clear profile of which groups of beneficiaries are at greatest economic peril from out-of-pocket drug costs.
1 - The true expenditure levels may be higher than this figure considering the possibility of under-reporting of prescription drug use in consumer surveys.
| Cover Page | Section 1 | Section 3 | Section 4 | Bibliography | Abstracts |
| Chapter 3 | Table of Contents | Appendix B |