Older Adults and Their Health Insurance

#9903 March 1999 Older Adults and Their Health Insurance by Gail A. Jensen Derek A. Weycker Wayne State University Jon R. Gabel KPMG Peat Marwick LLP

#9903 March 1999 Older Adults and Their Health Insurance by Gail A. Jensen Wayne State University Derek A. Weycker Wayne State University Jon R. Gabel KPMG Peat Marwick LLP The Public Policy Institute, formed in 1985, is part of the Research Group at AARP. One of the missions of the Institute is to foster research and analysis on public policy issues of interest to older Americans. This paper represents part of that effort. The views expressed herein are for information, debate and discussion, and do not necessarily represent formal policies of the Association. 1999, AARP. Reprinting with permission only. AARP, 601 E Street, N.W., Washington, DC 20049 www.aarp.org

TABLE OF CONTENTS Table of Tables. Foreword... Executive Summary..... i ii iii Introduction... 1 Data Sources.. 2 Types of Coverage Reported in the HRS.. 3 Findings. 4 Sources of Insurance Among Older Adults.. 4 Who Has What Coverage?.. 5 Employer Coverage for Workers Versus Non-Workers.. 7 Transitions in Health Coverage 10 Why the Changes in Coverage?. 12 A Closer Look at the Uninsured.. 13 Summary and Conclusions 15 References. 17 Tables 20

TABLE OF TABLES Table 1: Health Insurance Coverage of Older Adults, 1992 and 1994 20 Table 2: Sources of Health Insurance Among Older Adults by Characteristics of the Individual, 1992. 21 Table 3: Composition of Older Populations Defined on the Basis of Health Insurance, 1992 23 Table 4: Characteristics of Employer-Sponsored Coverage Among Older Adults, Worker Coverage Versus Non-Worker Coverage, 1994.. 25 Table 5: Health Plan Enrollment by Age Among Older Adults with Employer- Sponsored Coverage, 1994.. 26 Table 6: Health Plan Enrollment by Whether the Individual was Offered a Choice of Plans Among Older Adults with Employer-Sponsored Coverage, 1994. 27 Table 7: Percentage of Older Adults with Employer-Sponsored Coverage Who Were Offered a Choice of Plans, by Type of Plan Enrolled In, 1994.. 28 Table 8: Out-of-Pocket Premium Costs Among Older Adults with Employer Coverage by Type of Insurance, 1994. 29 Table 9: Changes in Health Insurance Among Older Adults Between 1992 and 1994, by Source of Coverage in 1992.... 30 Table 10: Transitions in Health Coverage Among Older Adults between 1992 and 1994, By Whether the Individual Newly Retired Over the Period... 31 Table 11: Events Occurring Between 1992 and 1994 Among Older Adults Whose Source of Health Insurance Changed Over the Period. 32 Table 12: Events Occurring Between 1992 and 1994 Among Older Adults Whose Source of Health Insurance Did Not Change Over the Period.... 33 Table 13: Self-Reported Health and Disability Status of Uninsured Older Adults Compared to Older Adults with Health Insurance... 34 Table 14: Medical Conditions of Uninsured Older Adults Compared to Older Adults with Health Insurance... 35 Table 15: Health Care Utilization of Uninsured Older Adults Compared to Older Adults with Health Insurance.. 36 Table 16: Demographics of Uninsured Older Adults Compared to Older Adults with Health Insurance.. 37 i

Foreword The health insurance coverage of older adults is an increasingly important public policy issue, as private employers cut back on their coverage of retired workers by a variety of methods. The issue has received heightened attention because of a 1998 proposal by the Clinton Administration to provide a Medicare buy-in for older adults. The desire to retire early is still strong among older workers, despite changes in Social Security eligibility which begin to take effect next year. Voluntary early retirement generally demands that an individual have health insurance coverage through his or her employer or through a spouse s employer. Evidence from a number of sources suggests that such coverage is becoming scarcer and more expensive to obtain. Increased cost may mean that individuals must choose less complete health coverage than they might wish, or perhaps forgo early retirement. Trends in the health insurance coverage of older adults is the focus of this paper by Professor Gail Jensen and two colleagues. Using the first two waves of the Health and Retirement Study, Jensen at al. examine the health insurance coverage of older adults in 1992 and 1994, how their coverage changed over the two-year period, and how their coverage varied with the decision to retire and with various health characteristics. The authors find that, among retirees aged 51 to 61 in 1992, nearly three-quarters relied on an employer plan for their health coverage. The typical non-worker with employer coverage in this age range reported health status that was decidedly worse than among working persons in that age group with either employer or individually-purchased private coverage, and was far more likely to be disabled. The authors show that employer coverage serves a needed and important function. Lack of health insurance coverage among all 51 to 61-year-olds appears to be increasing, and the health status of those without insurance is poorer and deteriorating faster than among those with insurance. Those who have active worker coverage are more likely to be in managed care than those without active worker coverage, and enrollment in managed care under Medicare, currently at about 16 percent (HCFA, 1998a), is likely to grow substantially over the next several years. These trends raise important public policy concerns about the cost of retirement, the ability of older workers to retain health coverage in their retirement years, and the adequacy of health insurance coverage for those who retire prior to eligibility for Medicare. Proposals to increase the age of Medicare eligibility to 67 only serve to increase the concern over adequacy of health insurance for workers reaching retirement age. John R. Gist Associate Director for Economic Policy Research Public Policy Institute AARP Research Group ii

Executive Summary Background Evidence from a number of sources suggests that employers have reduced health insurance coverage for their retired workers, and the costs of retaining health insurance for retired workers who still have employer coverage has increased. The percentage of older adults without health insurance has also risen. Purpose of the Report This paper examines the health insurance coverage of Americans aged 51 to 61 in 1992, as well as the coverage of the same cohort two years later, using the first two waves of the Health and Retirement Study. It reports on: changes in health insurance over the 1992-94 period; the detailed features of employer-sponsored coverage; and the relationship among changes in an individual's source of insurance, a decision to retire, and other major life events that occurred over this period. It also examines the size and characteristics of the uninsured population in this age range, and compares the characteristics of this subgroup to older adults who had coverage. Methodology The study uses the 1992 and 1994 waves of the longitudinal Health and Retirement Study. Most of the analysis consists of tabulations of type of health insurance coverage by key demographic and economic variables. The study also examines transitions over time in several ways: (1) by comparing the insurance status of respondents to the HRS in 1992 and again 24 months later; (2) by examining the health coverage status of various subgroups in 1992 and comparing it with their status 24 months later; and (3) by examining health status and health status transitions between 1992 and 1994 and relating them to changes in health insurance coverage. Empirical Results Five major findings emerge from the analysis: (1) Employer-sponsored health insurance is an important safety net for persons in the 51-63 age range who are not working. For many persons who retired between 1992 and 1994, their decision to do so was coupled with declining health or the emergence of a disabling condition. (2) The percentage of older adults with no source of health insurance increased from 12 to 14 percent between 1992 and 1994. This contrasts with the trend for the non-elderly population over this period; among all non-elderly Americans, the percentage without health insurance in 1992 and 1994 did not change, but rather held steady, according to the Current Population Survey. iii

(3) Receipt of employer-sponsored health insurance as a dependent declined among older adults between 1992 and 1994. This decrease was only partly offset by an increase in employer coverage in one's own name. (4) Uninsured older adults are in difficult straits. Among those without any source of health insurance in 1992, most were still uninsured two years later. As a group, they reported more disabilities than did persons of the same age with insurance, their health status was not as good, and it was deteriorating more rapidly than among other groups. (5) About half of all older adults (aged 53 to 63 in 1994) with employer-sponsored coverage are now under managed care. Conclusion Employer-sponsored health insurance is the primary source of health insurance coverage for early retirees. However, nearly nine percent of older workers lose such coverage when they retire, and an increasing percentage of older adults have no health insurance coverage at all. Those lacking health insurance report more health conditions and disabilities than those having health insurance. Greater enrollment in managed care plans among older adults with active-worker coverage than among those with non-worker coverage suggests that, over time, an increasing percentage of Medicare beneficiaries will be enrolled in managed care rather than in conventional Medicare. iv

1. Introduction 1 For the growing number of uninsured Americans, any serious illness can be financially devastating. This is particularly true for older adults, by which we generally refer to persons aged 55 to 64 in this paper. By almost any measure, the risk of illness increases as a person ages. Compared to younger adults, for example, more persons aged 55 to 64 report their overall health status as being fair or poor (rather than good or better than good), they suffer more often from debilitating chronic illness, their average medical expenses are substantially higher, they are more likely to receive hospital care and incur catastrophic health care expenses, and they experience a far higher death rate due to disease (National Center for Health Statistics, 1995). Yet, for many persons past age 50 but not yet 65 when coverage under Medicare normally begins, a lack of comprehensive health insurance may be the result of major transitions that have just occurred in their lives. Workers who retire before age 65, for example, may lose employersponsored coverage, the primary source of health insurance for most Americans. Older women not participating in the labor force may lose dependent coverage when their marriages end in divorce or their husbands die, or become disabled and are unable to work. For persons in their pre-retirement years, individual coverage is typically available through insurers, but for those with existing health problems, premiums may be prohibitively expensive. This paper provides an analysis of health insurance coverage in 1992 and 1994 among a cohort of older adults surveyed in both of these years. Our source of data is the longitudinal Health and Retirement Survey (HRS), Waves 1 and 2, which is representative of Americans who were aged 51 to 61 in 1992, and aged 53 to 63 in 1994. We report on: sources of health insurance within this population in these years; the relationship between changes in their health coverage, retirement status, and other major life events that occurred over this period; and the size and characteristics of the uninsured population in this age range. Other factors besides their financial vulnerability cause older adults health insurance coverage to become an increasing national concern. As baby boomers begin to enter their 50s and 60s, sheer numbers will magnify the health insurance problems they will encounter. The U.S. Census Bureau estimates that by 2005, there will be 29.6 million persons (or 10 percent of the population) who are aged 55 to 64, compared to 21.2 million (or 8 percent) currently. By 2010, their count will rise to 35.4 million (or 12 percent), almost as many as the number who will be aged 65 or older (39.7 million persons). In addition, growing numbers of men are withdrawing from the labor force before they are eligible for Medicare. Between 1970 and 1995, the percentage of men aged 55 to 64 who reported they were not working for pay nor looking for employment rose from 17 to 34 percent. The same has not been true of older women, however. During this period, their labor force participation rate actually rose modestly, from 43 to 46 percent. Still, the net result is that access 1 The authors thank AARP for its support, and John Gist, Amy Pienta, Gerry Smolka, Jules Lichtenstein, and two anonymous reviewers for helpful comments on an earlier version of this paper. The views expressed are those of the authors and do not necessarily reflect the opinions of AARP. 1

to active worker insurance is declining among older adults, while other sources of coverage, particularly employer-sponsored retiree coverage, are growing in importance. But according to surveys, employers have been cutting back recently on their provision of retiree health benefits. Data from the Bureau of Labor Statistics (BLS) on medium and large private business establishments, for example, document that between 1991 and 1995 the percentage of full-time insured workers employed by such firms who were promised retiree health benefits in early retirement (i.e., if they retired before reaching age 65) declined from 43 to 38 percent (BLS, 1994, 1997). In addition, BLS data indicate that the percentage who were promised such benefits at age 65 or older also declined, from 40 to 35 percent over this period. Other employer surveys conducted by KPMG Peat Marwick show a similar decline in the 1990s in the provision of retiree benefits (Gabel et al., 1997). KPMG data further show that, among firms that have continued to sponsor retiree benefits, many have effectively reduced their obligations in other ways, e.g., by tightening coverage eligibility rules or by limiting their contributions towards premiums. Thus, as time goes by, fewer retirees may have this source of coverage to help protect them from the costs of health care. The most recent analyses of health insurance among older adults were based on data from the mid- to late-1980s (Short and Monheit, 1988; Custer, 1990; Jensen, 1992), before the changes in Financial Accounting Standards Board standards occurred (in 1989). While we know that coverage for workers of all ages who have health insurance has changed dramatically since then (Gabel et al., 1994; Jensen et al., 1997), we know less about whether coverage among older adults has changed, and if so, how. While many in their 50s and early 60s, indeed, are actively working, a substantial number of persons in this age range are not, yet they rely on employersponsored health insurance. To report on the health insurance of this age cohort, we have organized the paper as follows. The next section describes in greater detail the Health and Retirement Survey (HRS) and the sources of health insurance discernible through this survey. Following this, we discuss the prevalence of different types of coverage observed within the HRS cohort, changes in coverage that occurred between 1992 and 1994, and the nature of employer-sponsored health insurance, both for those who were working in these years and for those who were not. We next examine the frequency with which individuals changed their source of health insurance over the period, and how coverage changes related to major life events, such as a change in marital status, retirement or change in level of workforce participation, onset of a disability, etc. We then take a closer look at individuals who were uninsured in 1992, and the characteristics of those within this subgroup who were (or were not) able to obtain some source of coverage by the time they were interviewed again in 1994. The concluding section summarizes our main findings and some of their implications for public policy. 2. Data Sources The longitudinal Health and Retirement Survey (HRS) is an ongoing nationally representative panel survey of community-dwelling adults in the U.S. who were born between the 2

years 1931 and 1941. In all, the HRS interviewed 12,652 persons in 1992, and this same group was then re-interviewed in 1994. The sample includes both the "age-eligible" survey participants, i.e., those persons who were born between 1931 and 1941, as well as their spouses or live-in partners, regardless of their date of birth. We restrict our analysis to survey respondents, both primary and secondary, 2 who are part of the "age-eligible" HRS cohort, meaning they were born between 1931 and 1941. For 1992 (Wave 1), our sample consists of 9,672 respondents aged 51-61, and in 1994 (Wave 2), 8,814 persons aged 53-63. 3 We use these two samples to examine the health insurance of this age cohort in 1992 and 1994, respectively. Our analysis of transitions in coverage over this period is based on the second sample because, for these individuals, we have data from both their 1992 and 1994 interviews. All estimates we report are population-weighted, and are representative of the entire non-institutionalized U.S. population born between 1931 and 1941. 3. Types of Coverage Reported in the HRS For each respondent, the survey gathered information on whether he or she had: (1) public health insurance coverage, i.e., Medicare, Medicaid, CHAMPUS, or another type, (2) employersponsored coverage, whether directly or through a spouse, (3) private individually-purchased coverage, or (4) no health insurance. Individually-purchased policies (those in the third set) include both comprehensive health insurance policies (not sponsored by an employer) as well as secondary policies, such as a Medicare supplemental ("Medigap") policy or a long-term care insurance plan. In 1994, the HRS gathered additional information on the nature of employer coverage -- whether it was a conventional plan, a health maintenance organization (HMO), or a preferred provider organization (PPO); whether prior authorization was required (from a physician gatekeeper or the insurer) to access specialist care; whether the individual had been offered a choice of plans by the employer or former employer; and whether the individual had been required to make out-of-pocket premium contributions. This special set of questions in 1994 was administered to all primary survey respondents 4 who had employer coverage, but only to some of the secondary respondents with such coverage. Specifically, of the 3,083 secondary respondents in our 1994 sample, only 907 of them were 2 Each household included in the HRS contains a primary respondent. For single-person households, the householder is the primary respondent. For couple households, one member is designated as the primary respondent, and the other as the secondary respondent. The primary respondent for these households is the individual who indicated (in 1992) that he or she is the most knowledgeable about financial matters in the household. The primary respondent may be either an original survey participant (i.e., an individual aged 51-61 in 1992 selected for inclusion in the HRS sample) or a person outside this age range who is the spouse or partner of an original survey participant. 3 The difference between these two counts consists of 858 persons who did not provide a Wave 2 interview. By far, the most common reason for HRS sample attrition in 1994 was simply refusal to be interviewed. Other reasons, which occurred with much less frequency, were institutionalization or death. 4 See footnote 1 above for the distinction between primary and secondary respondents. 3

asked these questions. 5 Because we lack these particular data for a majority of secondary respondents, we restrict analysis of the content of employer plans to primary respondents only, since only for the latter are the data complete. Throughout the paper, we divide persons with employer coverage into two groups: (1) those who indicated that they were actively working and covered by an employer plan, and (2) those who were not actively working although they were covered by such a plan. 6 We refer to the former as "employer coverage for workers," and to the latter as "employer coverage for nonworkers." This latter group includes employer coverage provided to early retirees and their spouses, non-working spouses of older workers (e.g., full-time homemakers), spouses who may have been working but who nonetheless had employer coverage as a dependent rather than employer coverage in their own name, and individuals who were not working but who were receiving COBRA continuation coverage. 4. Findings A. Sources of Insurance Among Older Adults Public and private sources covered 88 percent of Americans aged 51 to 61 in 1992. (See Table 1.) 7 Although some persons had government-sponsored coverage (12 percent), most relied on private health insurance. By far, persons in this age range were covered by employer- or union-sponsored plans; 72 percent had such coverage. Usually the coverage was in their own name (45 percent) rather than in their spouse's name (27 percent). Sixteen percent carried individually purchased health insurance, often as supplemental coverage, although for eight percent, individually purchased health insurance was their only source of coverage. Twelve percent were uninsured, according to these data. Among adults who were re-surveyed in 1994, the percentage with either a public or private source of coverage declined to 86 percent, with a corresponding increase in the percent uninsured to 14 percent. Significantly fewer persons reported employer coverage as a dependent (i.e., through their spouse). Specifically, 24 percent were receiving dependent employer coverage, which was down from 27 percent two years earlier. Other data from employer surveys has shown that in the early 1990s, many employers increased their required premium contributions for dependent benefits much more than they increased premium contributions for individual benefits (Jensen et al., 1997). This may be one of the reasons why fewer persons elected dependent benefits over this period. 5 In particular, those secondary respondents who indicated that "their employer coverage was unchanged from what it was in 1992" were not asked these new questions to the 1994 survey. 6 Individuals were coded as "working" if they said they were working at the time of the survey. Throughout the paper we use work status categories that were self-assigned by respondents. In each wave of the HRS, respondents were asked whether they considered themselves to be: working full-time, working part-time, unemployed, retired, or not in the labor force but not retired. 7 Because some people have multiple sources of coverage, the sources discussed here will sum to more than 100 percent. 4

At the same time, however, this decrease in coverage as a dependent was partly offset by an increase in employer coverage in one's own name, from 45 to 47 percent. The net result was a one percent decrease in the receipt of employer-sponsored health insurance, from 72 to 71 percent. Also in 1994, fewer persons in this cohort said they had only an individual policy which they had purchased on their own. The decreases in these two sources of private coverage account for the increase in the fraction who were uninsured. 8 B. Who Has What Coverage? We address this question in two ways. First, we examine how the distribution of insurance coverage varied for different socio-demographic subgroups of the age 51-61 cohort. We then divide our sample into subpopulations defined by their type of health insurance and describe the composition of each of these groups. Of special interest is the receipt of non-worker employer coverage, because many persons in this age group were not working, yet they depended on retiree health benefits, active-worker spousal benefits, or COBRA coverage (all in this category) for their health insurance. When asked in 1992, a full third of all adults in this age range said they were working neither full- nor part-time. Nine percent reported they had retired, 24 percent said they were not in the labor force, although not retired (e.g., a full-time homemaker), and five percent said they were unemployed. The single most important correlate of whether respondents were receiving non-worker employer coverage was whether they considered themselves retired (Table 2). Among individuals who said they had retired, 64 percent were receiving employer coverage for non-workers. Most of this is undoubtedly retiree health insurance, sponsored by their (or their spouse's) former employer. Those who reported they were not in the labor force but also not retired and those who were unemployed also had high rates of non-worker employer coverage -- 45 and 43 percent, respectively. For the latter two groups (especially the unemployed), some of this coverage is likely COBRA continuation coverage rather than retiree health insurance. Unfortunately, the HRS does not distinguish between the two. Rather, all we know regarding employer-sponsored coverage is whether it was in an individual s own name or someone else's name and whether the respondent was working at the time of the survey. If the respondent had 8 Recall that our sample in 1994 excludes 858 of the persons we report on in 1992 (most often because they refused to be re-interviewed in 1994). Consequently, our estimate of the percent uninsured in 1994 may be lower or higher than the true percent, depending on the uninsured status of these "unobservables." We can bound the size of this potential measurement problem, however. If all attritional respondents actually had health insurance in 1994 (the best case scenario), our estimate of the percent uninsured in that year would drop to 13 percent, which is still higher than the rate in 1992 (12 percent). However, if all of them were actually without health insurance in 1994 (the worst case scenario), our estimate of the percent uninsured in 1994 would rise to 22 percent. Thus, it is clear that the true rate rose over the period. 5

employer coverage but was not actively working at the time of the survey, then he or she falls under our column, "employer-sponsored coverage for non-workers." There is a pronounced pattern by age regarding who had non-worker employer coverage (Table 4). The youngest in this cohort (aged 53-56) were less likely to have such coverage and more likely to have active worker employer coverage instead. The oldest (aged 61-63), however, were most likely to have non-worker coverage and least likely to have active worker coverage. Married persons had higher rates of non-worker employer coverage, as did whites and persons with household incomes (in 1991) between $10,000 and $60,000. In contrast, those with incomes under $10,000 more often were either on public coverage or uninsured, while those in the highest income group (more than $60,000) tended to have active worker coverage. There is a definite correlation between (self-assessed) health status and type of insurance. The worse the respondents health, the more likely they were to have non-worker employer coverage or public coverage or be uninsured. Likewise, persons with better health were much more likely to have active worker coverage. Poor health has been shown to increase the probability of early retirement (Gordon and Binder, 1980; Burtless and Moffitt, 1985; Karoly and Rogowski, 1994), which is consistent with the pattern of insurance coverage in Table 2. Further bolstering this inference are findings that disabled individuals were twice as likely to have non-worker employer coverage as persons who were not disabled, and that persons who had experienced a recent hospital stay were 50 percent more likely to have non-worker employer coverage, compared to those who had not had a hospital stay. Table 3 examines the issue of who has what coverage in another way -- by comparing the demographic profile of persons insured through different sources within this age range: (1) employer coverage for non-workers, (2) employer coverage for workers, (3) public insurance, (4) an individual policy only, or (5) uninsured. As a group, older adults aged 51-61 with non-worker employer coverage have a more upscale economic profile than either those on public coverage or the uninsured (Table 3). They have slightly higher incomes than persons who purchase insurance individually, but are not quite as well off as their age-peers with active worker coverage. For example, about 38 percent of those with non-worker employer coverage had an annual income (in 1991) greater than $50,000, as compared to 53 percent of those with active worker coverage, 26 percent of those with public coverage, 30 percent of persons with individually purchased coverage, and 16 percent of persons who were uninsured. More than other groups, older adults with non-worker employer coverage tended to be married (88 percent). Eighty-seven percent of persons with non-worker employer coverage were white, as compared with 71 percent of older adults on public coverage, and 65 percent of the uninsured. About a third (34 percent) of all persons with non-worker employer coverage reported that they were retired, 62 percent said they were not in the labor force but not retired, and 12 percent said they were unemployed. Retirees account for a far larger share of non-worker employer coverage than they do any other type of coverage. The vast majority of persons with 6

active-worker coverage were working (not surprisingly), and most who had individually purchased coverage were likewise working, typically part- rather than full-time. Among the uninsured, half were working, 15 percent were unemployed, six percent were retired, and a third were not in the labor force, but not retired. Self-reported health status among those with non-worker employer coverage was decidedly worse than it was among those with either active worker coverage or only an individually purchased policy. In fact, in terms of health status, older adults with non-worker employer coverage are similar to the uninsured; the only group with worse health status are persons on public coverage. Over a quarter (27 percent) of those with non-worker employer coverage indicated they were in poor or fair health, as opposed to 10 percent of those with active worker coverage. Similarly, only 45 percent of those with non-worker health coverage indicated their health was better than good, as opposed to 63 percent of those with active worker coverage and 58 percent of those with an individually purchased policy. More had disabilities, as well. Thirty-four percent of the non-worker employer coverage group reported they were disabled, compared to eight percent of the active worker coverage group, 17 percent of the individually insured, and 28 percent of the uninsured. Only those with public coverage had a worse profile (59 percent disabled). Likewise, 15 percent of those in the non-worker coverage group had been hospitalized within the last 12 months, as opposed to eight percent of those with active worker coverage and seven percent of those with individual coverage. As with the other health measures, only persons with public coverage were worse off; 24 percent reported they had been recently hospitalized. C. Employer Coverage for Workers Versus Non-Workers In 1994, the HRS asked primary respondents who had employer-sponsored coverage about several aspects of their insurance, e.g., what type of plan they were in, the choices for coverage they had been offered, and their out-of-pocket premium expenses. Table 4 contrasts these characteristics for respondents who had non-worker employer coverage with respondents who had active worker coverage. The survey distinguished between three types of employer plans: conventional health insurance (also called indemnity insurance or a fee-for-service plan), a health maintenance organization (HMO), and a preferred provider organization (PPO). Respondents were asked to identify the type of plan in which they were enrolled, and whether they had public coverage or an individual supplemental policy in addition to their employersponsored coverage. As shown in Table 4, there was more dual coverage among persons receiving non-worker employer coverage. Sixteen percent had additional public coverage (e.g., Medicare or Medicaid), compared to five percent among persons with active worker coverage. Their rate of purchasing individual supplemental policies, however, was the same -- about 12 percent. Half (49 percent) of older adults with non-worker employer coverage belonged to a managed care plan (i.e., an HMO or PPO). Specifically, 26 percent were in HMOs, and 23 7

percent were in PPOs. These rates of enrollment were slightly less than among people of the same age with active worker coverage. Among those with active worker insurance, 30 percent were members of HMOs, and 26 percent were enrolled in PPOs. On the one hand, the high rate of managed care among those with non-worker coverage is not particularly surprising. Employers that provide retiree health coverage, which likely accounts for at least half of all non-worker coverage in this age range, generally offer benefits identical to those of active workers (Jensen and Morrisey, 1992), and managed care is now the dominant form of health insurance among active workers (Jensen et al., 1997). On the other hand, managed care plans have long had a reputation for attracting younger workers and families with young children (Hellinger, 1987), and clearly persons in this age cohort do not fit these categories. These data show, however, that managed care plans have a substantial presence, indeed, among older adults with employer coverage. While the rate of managed care enrollment among older adults is lower than that among workers of all ages with employer-sponsored coverage (62 percent in 1994) (Jensen et al., 1997), it is much higher than among Medicare beneficiaries, where managed care plans still have only about a 10 percent market share (Enthoven and Singer, 1996). Among persons with active worker coverage, the preference for fee-for-service (FFS) coverage increases mildly with an individual's age (Table 5), but even among those at the top end of the cohort (aged 61 to 63 at the time of this 1994 interview), most (54 percent) are in managed care plans. Persons with non-worker employer coverage exhibited no particular enrollment pattern as a function of age. Their total managed care enrollment was 46 percent among those aged 55 to 58, 52 percent among those aged 59 to 61, and 49 percent among those aged 62 to 64. These findings suggest that over the next several years, new Medicare beneficiaries may be much more receptive to managed care, since many will be transitioning from managed care plans to Medicare. They will have experience with managed care, unlike many of those already under Medicare who, for the most part, are accustomed to fee-for-service delivery and complete freedom of choice in their providers. In 1994, many older adults were in employer plans where they had to pay extra to see a specialist on their own without being referred. Close to half in both the non-worker and worker groups with employer coverage (42 and 47 percent, respectively) said they face such terms for specialist care (Table 4). Since many managed care plans stipulate that a primary care gatekeeper control access to specialist and hospital services, responses to this question are consistent with the enrollment data by type of plan, discussed above. The survey also inquired about the insurance choices made available by the employer (or former employer) providing their coverage. Forty-four percent of those with non-worker employer coverage said they had been offered a choice of plans (Table 4). While this is fewer than among workers of all ages with employer coverage (59 percent had a choice in 1994) (Jensen et al. 1997), this is still substantial. It is also slightly higher than among persons with active worker coverage in this age range (42 percent had a choice). The latter finding may primarily be 8

an artifact of firm size. Since retiree health insurance, which is likely most of this non-worker coverage, is definitely a mid- to large-firm phenomenon, we may be seeing more plan choice for retirees simply because larger firms are much more likely to sponsor multiple plans (Jensen et al., 1997). What types of plans did people select when offered a choice among conventional coverage or different forms of managed care? This question is addressed in Table 6. Enrollment in a managed care plan, especially an HMO, was more likely when offered a choice. Among those who faced a choice for non-worker employer coverage, one third (34 percent) selected an HMO and 21 percent, a PPO. Among active workers with a choice, 40 percent selected an HMO and 26 percent, a PPO. The survey did not inquire as to the nature of the choice, i.e., the types of plans on the menu. We know from other surveys, however, that a choice of plans more often than not includes a conventional (i.e., fee-for-service) option, as well as managed care plans (Gabel et al., 1997a). Thus, it is probably safe to infer that many older adults turned down a conventional plan to select their HMO or PPO. It is also apparent from the data that many firms offering only one plan in 1994 were offering a managed care product. When only one health plan was offered, half the time it was either an HMO or a PPO (Table 6). This is true for both worker coverage and non-worker coverage. This contrasts sharply with the late 1980s when solo offerings were almost always conventional fee-for-service plans (Gabel et al., 1997a). The flip side of this finding is that about half of all managed care enrollees among older adults freely chose their plan over another option offered by the firm (Table 7). Thus, for many of them, managed care was apparently their choice. On average, non-worker employer coverage cost only a little more in out-of-pocket premiums than did employer coverage for workers (See Table 4). Although the same proportion paid nothing for their benefits (32 to 33 percent), when a contribution was required, it tended to be higher for those with non-worker benefits. On a monthly basis, 12 percent of persons with non-worker coverage paid $1 to $40, 10 percent paid $41 to $80, and 45 percent paid more, whereas among persons with active worker coverage, 19 percent paid $1 to $40, 13 percent paid $41 to $80, and 35 percent paid more. The difference likely stems from employer contribution policies for retirees receiving post-retirement health benefits. During this time period the mid- 1990s close to one-out-of-five employers that offered retiree health coverage required the retiree to pay the full premium, i.e., they made no employer contribution toward the coverage (BLS, 1997). This practice is more common for retiree health coverage than it is for active worker coverage. Out-of-pocket costs varied only slightly by type of plan, i.e., whether it was fee-for-service (FFS), an HMO, or a PPO (Table 8). PPO coverage appears to cost a little less, on average, than either FFS or HMO coverage. 9

D. Transitions in Health Coverage So far, we have presented a mostly static picture of the health insurance held by this age cohort. Point-in-time descriptions, however, tell us nothing about the dynamics of coverage. How stable are different types of insurance? Do individuals retain the same coverage or are they likely to switch sources? How does insurance change upon retirement? What do those who lack access to active worker health benefits do for coverage once they retire? Because the HRS is a panel survey, these questions can be answered for the two-year period, 1992-1994. We analyze transitions in coverage among the 8,814 respondents to the HRS who were 51 to 61 years of age in 1992, and who were still participating in the study in 1994. We examine coverage transitions for three groups: (1) for the entire cohort (all 8,814 respondents); (2) for persons who newly retired over the period (438 respondents); and (3) for persons who did not newly retire over the period (8,376 respondents). This second group, clearly the smallest of the three, made up six percent of the total cohort, and eight percent of those individuals who were working in 1992. Table 9 provides a general description of the changes in coverage that occurred between 1992 and 1994, while Table 10 provides detail on the nature of these transitions. For persons with each type of coverage in 1992 (represented by table rows), Table 10 reports the full array of coverage sources observed for them in 1994 (represented by columns). For example, the entry 71.0 percent in the upper corner of the first panel of numbers indicates that among all older adults who had employer-sponsored coverage for non-workers in 1992, 71.0 percent still had that coverage when they were re-interviewed in 1994. For completeness in both tables, "no health insurance" is coded as a separate category. Persons least likely to retain their same source of health insurance were those with active worker coverage in 1992 who newly retired over the period. Only 16 percent of this subgroup retained active worker coverage; the rest switched from such coverage to another insurance category with their move into retirement (Table 9). For most (76 percent), the switch was to a different source of coverage (either non-worker employer coverage, 9 public coverage, or an individually purchased policy). Eight percent, however, became uninsured upon retiring from their job that had provided active worker benefits. These calculations ignore changes in coverage where the same general source was retained, e.g., a switch from an individually-purchased conventional plan to an individually-purchased HMO, or a from Medicaid to Medicare (both are public coverage). However, they do reflect changes from worker to non-worker coverage where the exact same insurance plan was retained (e.g., "Group Health Cooperative of Puget Sound" as an active worker to "Group Health Cooperative of Puget Sound" as a retiree). 10 One of the most disturbing findings from these data is the high percentage of the uninsured 9 An instance where an individual kept the exact same employer plan in their own name, but merely switched from active worker to retiree status is considered a movement onto non-worker employer coverage. 10 Unfortunately, the HRS does not provide enough information to ascertain whether an individual kept the exact same plan when he or she moved from active worker to non-worker coverage. 10

in 1992 who were still uninsured two years later. Of the 12 percent of older adults who were uninsured in 1992, less than half (44 percent) managed to acquire some type of coverage by 1994 (Table 9). The majority, 56 percent, were still uninsured. Among those who were able to acquire coverage by 1994, 17 percent went on to public coverage, 14 percent went on to active worker coverage, another 12 percent purchased coverage on their own, and a tiny fraction (four percent) acquired non-worker employer coverage (Table 10). Although not shown in either of these tables, persons who retired over the period showed the following sources for their health insurance in 1992: 74 percent had active worker coverage, six percent were on public coverage, 16 percent had an individually purchased policy, and four percent were uninsured. Among those in the first group (persons who newly retired from jobs which had carried active worker coverage), 73 percent switched to non-worker employer coverage with their move into retirement, 16 percent were still insured with active worker health insurance, eight percent were on public coverage, and 17 percent had purchased individual coverage (Table 10). These percentages sum to more than 92 percent -- the percentage who were insured in both years within this subgroup -- because some of these new retirees had multiple sources of insurance -- e.g., employer coverage and public coverage. Persons most likely to retain their source of insurance were those who in 1992 had active worker coverage and who continued working over the period (87 percent retained their employer coverage), persons on public coverage (85 percent retention rate), and persons with non-worker employer coverage (71 percent) (Table 9). Persons who began the period with individually purchased insurance were less likely to have retained it. Forty-two percent of them were without individual coverage two years later. Among those who no longer had individual coverage, most acquired other insurance; however, a fifth (20 percent or 8.6/42.0) became uninsured (Table 10). The retention rate for non-worker employer coverage was 71 percent, which is well below the retention rate for active-worker employer coverage (81 percent). There are multiple reasons for this. First, some of the non-worker employer coverage is undoubtedly COBRA continuation coverage, which under current federal rules, ceases some 18 to 36 months after it begins. Second, not all of those who are receiving employer coverage as a dependent are entitled to receive it if their spouse (whose name the coverage is in) dies. For example, about a fifth of all individuals who are receiving spousal retiree health benefits (included within non-worker employer coverage) face rules that terminate their coverage in the event that their covered spouse dies (Jensen and Morrisey, 1992). Finally, some of these individuals began working over the period, in which case, their employer coverage as non-worker would likely switch to employer coverage as a worker. It is clear that, with one exception (namely, persons with active worker coverage electing early retirement), people in this age range were more likely than not to remain with the same source of coverage they held in 1992. Early retirees, however, moved onto non-worker employer coverage. Also, if coverage changed, the path of change was dependent on the type of insurance held in 1992. For example, persons who already had non-worker employer coverage moved more often to public coverage (if they moved at all). Persons with private individually purchased 11

insurance, in contrast, were more likely to switch to active worker coverage, if at all, and the uninsured, like those receiving employer-sponsored non-worker coverage, most often moved to public coverage (if they moved at all). E. Why the Changes in Coverage? What events precipitated these transitions? We used the HRS data to uncover recent changes in marital status, employment, and health status that may have precipitated these changes in coverage. Table 11 reports our findings for older adults who switched to non-worker employer coverage, active worker employer coverage, public coverage, private individually purchased insurance, or who became uninsured by 1994. The five columns in the table report the percentage of newcomers to that type of coverage who experienced the event described by each row. Thus, the entry 1.1 in the first row indicates that about one percent of persons who newly acquired nonworker employer coverage between 1992 and 1994 experienced a death, divorce, or separation from their spouse over the period. (Clearly, very few of the persons entering this insurance group experienced this triggering event.) 11 To put the frequency of these events in perspective, Table 12 reports their incidence among persons who did not change coverage over the period. Only if "movers" (Table 11) experienced an event more often than "stayers" (Table 12) should we infer that it may have precipitated the move. For each individual, if multiple events occurred, all were flagged and recorded. Thus, there was no hierarchical assignment system. Changes in employment and health status appear to have triggered a substantial portion of the movement onto various coverages. In contrast, changes in marital status, including death of spouse, divorce, separation, or a recent marriage, did not. The results reinforce and confirm what Tables 2 and 3 suggested: early retirement, and to a lesser extent, job loss led persons aged 51-61 onto non-worker employer coverage. Equally important, a switch to non-worker employer coverage between 1992 and 1994 was also associated with health status that had recently been worsening, the onset of a disability since 1992, and/or a hospital stay over the period. Persons newly acquiring non-worker employer coverage clearly experienced much more "change for the worse," as measured by each of these three health indices, compared to persons already on such coverage. In fact, their deteriorating physical state is rivaled only by recipients of public coverage, whether new or continuing, and the so-called "long-term" uninsured. This last group (see the last column of Table 12) are persons who had no health insurance in either 1992 or 1994. While declining health for newcomers to non-worker employer coverage (Table 11) is not quite as dramatic as it is among those with public coverage, they did experience a high rate of recent hospitalization. In fact, their rate of hospitalization was higher than for any other group except new beneficiaries of public coverage. They were also, by far, the most likely group to have developed a disability since 1992. About a quarter (23 percent) reported that between 1992 and 11 Under the federal COBRA provisions, persons who are separated from an employer plan due to the death of a spouse, or divorce or separation from their spouse, are eligible to continue participating in that plan for 36 months, provided they pay 102 percent of the full premium themselves. Thus, eligibility under COBRA is one reason why we might expect these events to trigger entry into this category. 12