5
Risk Selection, Risk Sharing, and Policy

Some of the . . . controversy reduces to a mundane debate about who will pay for whom and how much.

Douglas M. MacIntyre, 1962

Biased risk selection is always possible when individuals, employers, or other groups can choose whether or not to buy health coverage or whether to select one health plan instead of another. Unfortunately, this continual exercise of choice—a valued feature of individual liberty and of markets generally—can create both philosophical and practical problems in the health care arena. In particular, it can seriously diminish the degree to which the burden of health care expenses is shared among the well and the ill.

In addition, risk selection can lead health plans to compete for lower-cost enrollees and to avoid higher-cost individuals because the price and profit advantages from such tactics can outweigh the gains to be achieved by cost-effective management of health care and administrative services. Health care costs are highly skewed in their distribution. In any year, perhaps 5 percent of a population will account for 50 percent of expenditures, and 20 percent will account for 80 percent. Thus, any health plan (or employer) that is more successful at avoiding those who are likely to be in the small-group of higher utilizers will have a significant competitive advantage of a particular sort, but not the principal one desired by advocates of market-based strategies for health care reform.

The evolution and current structure of employment-based health benefits and proposals for change are difficult to understand and evaluate without a careful analysis of risk selection. This chapter reviews basic concepts, examines factors contributing to risk selection, considers evidence about its extent, and discusses key policy issues and proposals for managing the unwanted effects of risk selection. The focus is on employer groups,



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 167
Employment and Health Benefits: A Connection at Risk 5 Risk Selection, Risk Sharing, and Policy Some of the . . . controversy reduces to a mundane debate about who will pay for whom and how much. Douglas M. MacIntyre, 1962 Biased risk selection is always possible when individuals, employers, or other groups can choose whether or not to buy health coverage or whether to select one health plan instead of another. Unfortunately, this continual exercise of choice—a valued feature of individual liberty and of markets generally—can create both philosophical and practical problems in the health care arena. In particular, it can seriously diminish the degree to which the burden of health care expenses is shared among the well and the ill. In addition, risk selection can lead health plans to compete for lower-cost enrollees and to avoid higher-cost individuals because the price and profit advantages from such tactics can outweigh the gains to be achieved by cost-effective management of health care and administrative services. Health care costs are highly skewed in their distribution. In any year, perhaps 5 percent of a population will account for 50 percent of expenditures, and 20 percent will account for 80 percent. Thus, any health plan (or employer) that is more successful at avoiding those who are likely to be in the small-group of higher utilizers will have a significant competitive advantage of a particular sort, but not the principal one desired by advocates of market-based strategies for health care reform. The evolution and current structure of employment-based health benefits and proposals for change are difficult to understand and evaluate without a careful analysis of risk selection. This chapter reviews basic concepts, examines factors contributing to risk selection, considers evidence about its extent, and discusses key policy issues and proposals for managing the unwanted effects of risk selection. The focus is on employer groups,

OCR for page 167
Employment and Health Benefits: A Connection at Risk not individual purchasers of health coverage, and on choices among health plans, not the choice to purchase or not purchase coverage. BASIC CONCEPTS As defined in Chapter 1, biased risk selection exists (1) when individuals or groups that purchase health coverage differ from nonpurchasers in their likelihood of incurring health care expenses or (2) when those who enroll in competing health plans differ in the level of risk they present to specific plans. The first kind of risk selection primarily involves the market for individual and small-group coverage in which those with higher-risks are thought to be more likely than those with lower risks to purchase insurance.1 The individual and small-group market also suffers from the second form of biased risk selection when higher-risk purchasers seek more generous or flexible coverage than lower-risk purchasers. In general, because large employers almost universally provide health benefits and have more predictable costs, large groups present fewer problems with risk selection than either individuals or small groups. However, problems can arise in larger groups when they create an internal market by offering employees a choice among health plans and their higher-risk and lower-risk individuals select different plans. One result of risk selection is risk segmentation, the clustering of individuals at higher and lower risk of incurring health care expenses in different health plans or insurance pools. As noted in Chapter 1, risk segmentation can also occur when insurers experience rate groups on the basis of their claims (cost) history and when larger, less risky employers depart the group insurance market—as most have—in favor of self-insurance.2 1   Some suggestive information on differences between enrollees with individual coverage (who must seek out coverage for themselves) and those with group coverage (who have it offered to them as a matter of routine) is available from an insurer who has not used medical underwriting to screen individual purchasers. Independence Blue Cross (Philadelphia), which offers individual coverage without medical underwriting on an open enrollment basis throughout the calendar year for its five-county service area, reports information on both its individual and its group enrollment. For 1987, its individual subscribers were 6 years older on average than its group subscribers, had a 46 percent higher hospital admission rate, and incurred costs that were 60 percent higher (Independence Blue Cross and Pennsylvania Blue Shield, 1988). Fifty-five percent of the individual subscribers were age 50 or over, compared with 35 percent of the group subscribers. 2   Risk segmentation also can occur when individual choice is quite limited. As noted in Chapter 1, risk segmentation among the German sickness funds occurs because plans draw or are assigned their membership from occupational and other groups that differ in age, income, and other risk factors (Wysong and Abel, 1990).

OCR for page 167
Employment and Health Benefits: A Connection at Risk When a health plan, insurer, or employer attracts a less risky and less costly group than the average (or than its competitors), it has experienced favorable selection. A plan that attracts a more risky and costly membership has unfavorable or adverse selection.  If the cost of health coverage is linked to the risk level of the pool of covered individuals, premiums for those who find themselves in groups with unfavorable selection will be higher, even if their own risk of medical care expense is low. The financial incentive then is for these lower-risk individuals to exit, making the remaining pool even more expensive, perhaps so expensive that even those most in need of the coverage cannot pay the premium and the plan fails. Some argue that the solution to adverse selection is better information, so that everyone pays his or her exact risk-rated premium based on detailed personal information about health status, health behavior, work environments, and other factors affecting the risk of medical care expenses. This approach reflects a value judgment that the young and the healthy should not have to subsidize the old and the unhealthy (i.e., that risk segmentation is fair and desirable), a judgment with which the majority of this committee disagrees. For technical and practical reasons, perfect risk rating is unlikely if not impossible, so adverse selection related to information imperfections would still exist. (As described later in this chapter, technical problems also affect strategies to compensate for or discourage selection by risk adjusting employers' or governments' [not individuals'] payments to health plans.) The policy problem with risk selection is not that it can put adversely affected health plans out of business. Rather, risk selection is a concern because it encourages socially unproductive competition based on risk selection rather than on cost-effective management of care for the ill and injured (GAO, 1991e; Hall, 1992; Light, 1992).3 Any strategy of health care reform that is based on competition and choices about health coverage should address these problems, and several options are discussed later in this chapter. Design of an appropriate strategy depends on an understanding of some of the factors that produce selection and the degree to which the insurers, the insured, and policymakers can manipulate them to exacerbate or control risk selection (Feldman and Dowd, 1991; GAO, 1991e; Light, 1992). 3   The charge that competition in accident and health insurance tends to focus on marketing rather than product quality is hardly new. For example, the following statement dates back to 1928: "[The] outstanding characteristics [of commercial accident and health insurance] are the heterogeneity of its policy forms and the non-scientific nature of its premiums . . . [Each is] a direct consequence of competition, which unlike competition in life and many casualty covers, devotes itself to the devising of new forms rather than to the emphasis of security and service on standard, or practically standard policies" (Kulp, cited in Faulkner, 1940, p. 1).

OCR for page 167
Employment and Health Benefits: A Connection at Risk FACTORS CONTRIBUTING TO RISK SELECTION Many factors have been identified by anecdote, theory, survey data, and other research as sources of favorable or unfavorable selection (see, generally, Pauly, 1974; Berki and Ashcraft, 1980; Berki et al., 1980; Luft et al., 1985; Neipp and Zeckhauser, 1985; Wilensky and Rossiter, 1986; Luft, 1987, 1991; Luft and Miller, 1988; Mechanic et al., 1990; Anderson, 1991b). These factors relate to the characteristics and choices of individuals, employers, and health plans. Individual Factors Individuals' choices about health plans are affected by a variety of characteristics that may not be easy to measure directly but that are correlated to more easily measured characteristics, such as age, gender, occupation, and income. The underlying characteristics that affect choice include individuals' actual and perceived health status (and that of their spouses and children);  the knowledge and preferences individuals have about using and paying for health services and their willingness and ability to accept risk; and  their willingness and ability to engage in informed decisionmaking about joining, leaving, or continuing in a health plan. The dynamics of individual choice among multiple health plans have not been studied in much detail, but one pilot effort suggests that individuals (1) examine only a few options, (2) are aware of coverage differences for their special needs (e.g., mental health care and pregnancy care), and (3) tend to understand traditional plans better than newer plans with their gatekeeper and other managed care features (Mechanic et al., 1990). Freedom to choose one's physician appears to be particularly important to those choosing conventional plans. Individuals with a significant ongoing relationship with a physician will tend to stay with the health plan that includes or covers that physician rather than switch to another plan. Unpublished research on the Federal Employees Health Benefits Plan (discussed below) indicates that expected out-of-pocket costs (employee-paid premium and other cost sharing) is the single most important factor in decisionmaking for low-risk individuals. High-risk individuals focus first on plan benefits. Employer Factors Among small employers, some of the same factors that influence individuals' choices may likewise influence the decisions made by employers to

OCR for page 167
Employment and Health Benefits: A Connection at Risk purchase health coverage or to seek a particular kind of insurance. Virtually all large employers provide broad health benefits, but smaller, more economically marginal employers may be particularly motivated to seek insurance if they have employees with health problems and may be especially sensitive to price in making the buy or no-buy decision. As noted earlier, the major factor affecting risk selection for larger employers is the offering to employees of a choice of health plans (Luft, 1991). A multiple-choice health benefit program allows the individual risk factors identified above to operate within the employment group. In attempting to make individuals more cost conscious or to accommodate the varied personal circumstances of employees, employers may unintentionally—but sometimes deliberately—influence risk selection in a number of other ways, such as requiring that individuals pay part of the premium;  limiting the employer contribution to the premium to a portion of the premium for the least expensive plan or otherwise structuring the premium contribution to favor a particular kind of plan;  offering a choice of more and less generous (high and low option) health plans and requiring employees to pay more for the more generous coverage;  including preexisting condition limitations or requiring specific kinds of coverage in some but not all health plans;  offering a flexible benefit program that encourages employees to select benefits on the basis of individuals' needs and preferences;  allowing employees to opt out of coverage entirely;  combining active employees and non-Medicare-eligible retirees in the same risk pool for purposes of setting the active employees' contribution to premium; and  frequently adding new health plans or major new features to existing plans. The first seven of these employer actions make it attractive for healthier individuals to minimize their purchase of health coverage and to segregate themselves in different risk pools from less healthy individuals (Luft et al., 1985; Luft, 1991). The eighth action sets up a situation in which the sponsors of a new health plan with moderate benefits can ''low ball" premiums (set them lower than the benefits and employee population might seem to warrant) to encourage enrollment by low-risk individuals who are willing to "take a chance" because they are not currently using health services and do not anticipate the need for any. The selection advantage a health plan gains by this strategy may take years to fade. Employers may also deliberately attempt to make specific types of health plans more or less attractive to high-risk individuals. For example, to en-

OCR for page 167
Employment and Health Benefits: A Connection at Risk courage higher-risk individuals to enroll in plans thought to manage health care more efficiently, coverage may be restricted or cost sharing greatly increased for specific services in conventional self-insured plans but not in health maintenance organizations (HMOs). Alternatively, employers may treat their conventional plan as the insurer of last resort for individuals with costly problems, in particular, mental illness. To the extent that some employers offer health benefits and others do not or some offer more generous or flexible coverage, these employers may attract higher-risk employees (either the workers themselves or their families) and also thus suffer from unfavorable risk selection. Employers—like insurers—may respond by reducing coverage for specific medical conditions, strengthening exclusions for preexisting conditions, making higher-risk individuals pay more for health benefits, and other practices. One major airline now permanently excludes coverage for preexisting conditions for new employees (Seeman, 1992). Although evidence of employers' efforts to protect themselves in these ways is limited, the rationale for protective strategies is strong enough to raise concerns not only about further segmentation of access to health benefits but also about the resulting distortions in labor markets. The most drastic actions that employers may take to avoid high-risk employees come in the form of the explicit hiring bans and dismissal policies involving smokers, social drinkers, mountain climbers, and others that were described in Chapter 3. The current extent of such overt policies and similar covert practices is unknown. Health Plan and Insurer Factors Of much current interest to policymakers are the characteristics and strategies of health plans and insurers that may affect individuals' choices. These include underwriting and pricing practices;  plan benefit design;  incentives for use of a network of practitioners and providers;  administrative procedures;  marketing strategies; and  general reputation. As discussed below, health plans can manipulate their features and practices to varying degrees. Much of the controversy over risk selection relates to charges that some health plans deliberately "skim," "cream," or "cherry pick'' good risks or "blacklist," "redline," "churn," or otherwise avoid poor risks (GAO, 1991e; Chollet, 1992a; Hall, 1992; Zellers et al., 1992). This behavior is particularly associated with the small-group insurance market.

OCR for page 167
Employment and Health Benefits: A Connection at Risk The most powerful tool that health plans have to limit or channel risk selection is medical underwriting, which allows them to classify risks, price them, and accept, reject, or limit coverage for individuals and groups. The most extreme form of underwriting is to exclude altogether from coverage individuals with certain characteristics, such as HIV infection, and entire categories of employers or occupations, such as hairdressers and lawyers—the latter for higher perceived risk of litigation. Coverage for a preexisting condition can also be barred temporarily or permanently. In addition, premiums for individual and small-group coverage are commonly based on individual risk factors, including both demographic characteristics and health status. Figure 5.1 shows how age and gender can affect premiums (CRS, 1988b). A health plan that does not use medical underwriting makes itself vulnerable to unfavorable risk selection if its competitors do engage in risk rating and similar practices. Benefit design can crucially affect a health plan's potential for unfavorable or favorable risk selection. The range of relevant design features is broad and includes the types of services, providers, and sites of care cov- FIGURE 5.1 Variation in average annual plan premiums for the typical health plan, by age and gender, 1986. SOURCE: Adapted from CRS, 1988b, Vol. 1, p. 41; original data from Hay/Huggins.

OCR for page 167
Employment and Health Benefits: A Connection at Risk ered; the requirements for patient cost sharing; and caps on the volume, frequency, or duration of a covered service. Some benefits, in particular more extensive inpatient and outpatient mental health benefits, are thought to attract higher-cost patients. Other benefits, for example, fitness and other health promotion programs, may attract healthier individuals. A health plan that chooses (or is required by courts or legislatures) to cover expensive services, such as transplants, hospice care, or certain experimental treatments, when other plans can exclude such services will likely have both a more costly benefit package and a more costly membership. Another characteristic that may contribute to risk selection is whether a health plan limits coverage to a specified network of physicians, hospitals, and other health care practitioners and providers. As defined in Chapter 3, conventional health plans place no or few limits on choice of provider, whereas network plans (e.g., HMOs, preferred provider organizations [PPOs], and point-of-service [POS] plans) reduce or exclude coverage for nonnetwork providers. Plans that limit coverage to a defined panel of practitioners and providers may be unattractive to individuals with chronic health problems who want flexibility in choosing medical specialists and who have established a relationship with a physician that they do not want to disrupt. In addition, individuals such as retirees who travel a lot outside the plan's service area may find a network plan difficult. Network plans may also be vulnerable to risk selection. Independent practice associations (IPAs), PPOs, and other health plans whose physicians also see patients in other plans may attract some higher-risk individuals who would have to switch physicians if they changed to a staff model HMO. Similarly, mature HMOs will likely have a core of older patients with established physician relationships who are not interested in switching to a new HMO. The latter thereby gains a selection advantage. In addition, because plans with maximum choice of provider generally cost more, network plans with nominal copayments may incur adverse selection for some conditions, such as routine pregnancies. The composition of a particular limited or closed provider panel is also relevant. For example, a network plan that includes a university hospital known for its care of a particularly costly medical problem is more likely to attract patients with this problem than a plan that excludes this hospital. A plan that includes more subspecialists is similarly vulnerable. In general, the potential for risk selection is one more factor (in addition to cost, quality, reputation, and geographic coverage) to be weighed when network plans consider the composition of their provider panels. The lore of health plan efforts to avoid poorer risks also includes a variety of imaginative administrative practices (Luft and Miller, 1988). These include requiring individuals to visit health plan offices in order to enroll, permitting long appointment queues to develop for certain kinds of services,

OCR for page 167
Employment and Health Benefits: A Connection at Risk requiring new female enrollees to change gynecologists and obstetricians regardless of whether their existing physician is part of the plan, and counseling sicker members to switch to other plans where they will have better coverage. Marketing and sales strategies are particularly sensitive to risk selection issues. A health plan is more likely to seek or accept an advertising spot during a sports telecast than a spot during a program on the latest strategies for treating HIV infection. A health plan for the elderly that markets in senior centers is less vulnerable to unfavorable risk selection than one that markets in nursing homes. A Case in Point To illuminate the impact of individual, employer, and health plan factors, the experience of the Federal Employees Health Benefits Program (FEHBP), the oldest consumer-choice employer health program in the country, is instructive (Towers, Perrin, Forster & Crosby, Inc., 1988; CRS, 1989; Enthoven, 1989; Jones, 1989; U.S. House of Representatives, 1989; Welch, 1989; Wiener, 1990; GAO, 1992c; Washington Post, 1992; see also Chapter 2 of this report). Established in 1959 and operational in 1960, FEHBP has always offered numerous choices among fee-for-service plans and among HMOs. There is an annual "open season" during which any subscriber can switch from any plan to any other plan available in its geographic area. FEHBP has never allowed medical underwriting. It includes retirees as well as active employees and thus covers a very wide age range. It has always required premium cost sharing by the subscribers. A statutory formula sets the government contribution to any plan at 60 percent of the average premium of the six largest plans' high-option offerings. This contribution cannot, however, exceed 75 percent of the chosen plan's premium costs. According to internal studies conducted by the Blue Cross and Blue Shield Association, a large majority of the 4 million contract-holders (or subscribers) in this program annually review their own health care plan and two or three others and consider making a change (though usually fewer than 10 percent actually make a change). The primary factors considered are out-of-pocket costs, perceived need for specific benefits, and sometimes the quality of plan services, such as claims processing. The weight assigned by individuals to each factor varies from year to year, depending on changes in individual circumstances. The FEHBP has in fact suffered from serious risk selection problems, even for the plans with very large enrollments. Table 5.1 illustrates the impact of risk selection on premiums for three conventional health benefit plans in the federal program. An analysis of high-option and low- (or

OCR for page 167
Employment and Health Benefits: A Connection at Risk TABLE 5.1 Estimated Impact of Biased Risk Selection on Premiums in a Multiple-Choice Program, Individual (self-only) Coverage, Federal Employees Health Benefits Program, 1989   Individual (self-only) Premium Plan Standardized Premium (actuarial value) Total Actual Premium Low-coverage plan $1,422 $ 833 Median-coverage plan 1,700 1,502 High-coverage plan 2,016 3,032 % Difference between low and high plans 42% 264% Discussion: The first column reports the standardized premium or actuarial value of the benefits in three FEHBP fee-for-service plans assuming a standardized or comparable group of enrollees in each. The actuarial value is based on the differences in covered services (for example, whether hospice services are covered) and patient cost sharing (for example, level of deductible and coinsurance). The plan with the most extensive benefits is worth 42% more in coverage than the low-coverage plan. The second column reports the actual premium for each plan, which reflects past utilization of health services. The premium for the high-coverage plan is 264% higher. The difference in the range of actual premiums is substantially greater than the difference in actuarial value. This shows how characteristics of enrollees choosing different plans (risk selection) affect premiums. SOURCE: Adapted from CRS, 1989, p. 123. standard) option plans managed by one insurer that were estimated to have the same actuarial value (despite their labels) found the actual (experience-related) premium for the high-option plan was 79 percent above that for the low-option plan (Welch, 1989). Adverse selection problems contributed significantly to the decision by several fee-for-service plans and dozens of HMOs to withdraw from the program during the late 1980s. For example, it was the major factor in the departure of Aetna, which was one of the original FEHBP participants and which had nearly 200,000 contracts in 1988 before it withdrew from the market. In general, there has been a market "shake-out," due in large part to risk selection. Some efforts have been made to control or limit the extent of risk selection in FEHBP, and some changes made for other reasons may also have been helpful. First, benefit packages among the competing carriers have become more similar (for example, through changes in mental health benefits and dental benefits), and all are moving to incorporate more managed care features. Such correspondence reduces the differences in health plans that might attract or deflect high-risk individuals and encourage fre-

OCR for page 167
Employment and Health Benefits: A Connection at Risk quent switching among health plans. Second, the government's Office of Personnel Management has not allowed new fee-for-service carriers to enter the program and has set community rating rules for newly participating HMOs that discourage "low ball" pricing to attract low-risk individuals. Third, because the government caps its contribution at 75 percent of any health plan's premium rather than paying a fixed dollar amount (which might cover all the premium for a low-cost plan), it mitigates somewhat the reward for enrolling in plans with particularly favorable selection.4 Finally, legislation made federal retirees over age 65 who retired after 1982 eligible for Medicare. When Medicare became the primary payer for this group, it significantly reduced the cost to FEHBP carriers and HMOs of their highest-cost enrollees. EVIDENCE OF RISK SELECTION Several researchers have presented or reviewed the evidence of risk selection (Berki and Ashcraft, 1980; Jackson-Beeck and Kleinman, 1983; Neipp and Zeckhauser, 1985; Price and Mays, 1985; Welch, 1985; Wilensky and Rossiter, 1986; Hellinger, 1987; Luft, 1987; Luft and Miller, 1988; Newhouse et al., 1989; Lichtenstein et al., 1991; Robinson et al., 1991). Although this evidence is not conclusive and the dynamics of the health insurance market are changing rapidly, the evidence taken as a whole does suggest that selection is fairly common and is related to the factors identified above. Nonetheless, it is important to note that (1) the health plans studied are not random or even representative of all plans and communities; (2) most studies have small "sample" sizes; (3) variations related to local delivery systems and cultures may be insufficiently identified; and (4) measures of risk (e.g., average age or average functional status) may inadequately capture important differences among groups. Despite the shortcomings of the data, several policy-relevant findings are suggested by accumulated evidence, the direct experience of the committee members, and the perspective of experts consulted by the committee. First, risk selection is not confined to one type of health plan or benefit design. Both indemnity health plans and HMOs have been found to suffer from unfavorable selection. However, in competition with indemnity plans, HMOs overall are more likely to have experienced favorable selection than indemnity plans. This difference between HMOs and conventional plans is clearest in studies comparing the utilization of those enrolling in HMOs ("leavers") 4   However, it could also reduce somewhat the financial incentive for individuals to enroll in health plans that have gained their premium advantage through efficient management of medical care and medical resources.

OCR for page 167
Employment and Health Benefits: A Connection at Risk 1991). In particular, they have reduced the number of HMOs they offer. This strategy also may simplify the analysis of risk selection among the remaining plans as well as reduce the complexity of plan administration and employee decisionmaking. Others have sought to preserve choice by bringing risk under a single umbrella through a "total replacement" package. One approach under this option is to find a single insurer that will underwrite (in full or part) two or more health plan choices, for example, an HMO and an indemnity plan, or those options plus a PPO. Under such dual-or triple-option arrangements, employees still chose among the two or three plan options on a periodic basis and face separate premiums for each option. The theory is that this approach will eliminate the incentive for health plan competition based on risk selection within the employer group because the same insurer fully or partially underwrites all options within the employer group. It would leave untouched the incentives for insurers to compete to attract low-risk employer groups. A simpler approach—at least descriptively—is the point-of-service health plan, which (1) does not require that a yearly choice be made between network and nonnetwork enrollment and (2) does not establish different premiums for those choosing in-network versus out-of-network services. Instead, employees may choose at the point of service whether they want to use nonnetwork services and pay more for that choice (usually up to some limit). POS plans may be part of a replacement strategy or a consolidation strategy, as in the core case study presented in Chapter 4. Depending on the payment arrangements for network providers, these plans may shift some of the selection problem to the network providers. If the provider network is not attractive to individuals with chronic illnesses, then the plans may shift the burden to these individuals. Replacement and consolidation strategies have been criticized for restricting employee choice among health plans and potentially discouraging innovation. In addition, if an employer becomes dissatisfied with the POS plan or the "umbrella" carrier for dual-or triple-option programs and switches to another plan or carrier, the switch may disrupt patient-physician relationships and interfere with continuity of care. Overall, because consolidation strategies are relatively new and subject to little or no independent evaluation, their impact on employer or employee costs, quality of care, and employee relations has yet to be demonstrated. Regulating the Terms of Competition at the Community Level Other approaches to consolidation would operate at the community level, not at the level of the individual employer. They would create a system of health plan certification and purchasing cooperatives to streamline the purchase of health benefits on a state or regional basis. Some proposals are

OCR for page 167
Employment and Health Benefits: A Connection at Risk aimed at the small-group market; others would cover large groups as well. Proposals that have the control of risk selection as a major objective generally would not allow multiple cooperatives to operate in the same area and would discourage or prohibit coverage through noncertified health plans. Purchasing cooperatives would consolidate the role of purchasing agent for small employers and would operate somewhat like large employers now do when they screen, select, and monitor a choice of health plans for their employees. They could also administer a policy of risk adjusting employer (or employer and government) contributions to health plans. Depending on the certification criteria and their enforcement, the certification mechanism could shrink the number of insurers and health plans and thus result in further consolidation of the risk pool. Still, for the individual who had previously been offered a single health plan by a small employer, some choice of plans would become available. The consolidation of the purchaser role is part of a broader concept of "managed competition." Some comprehensive managed competition strategies propose to regulate various kinds of health plan practices that encourage risk selection and complicate comparisons of health plan performance (Enthoven, 1988a,b; Ellwood, 1991; Enthoven and Kronick, 1991; Kronick, 1991a,b, 1992). Table 5.4 lists some of the targets of these proposals. TABLE 5.4 Some Steps Proposed to Manage or Limit Health Plan Competition Based on Risk Selection Health Plan Features • Definition of standard benefit package(s) to limit features intended to discourage Enrollment by high-risk individuals. • Requirements for health plans to contract with certain types of providers (e.g., tertiary Care centers) so as not to discourage membership by high-risk individuals. Consumer Protection Processes • Regulation and monitoring of enrollment and disenrollment processes and results and Surveys of membership satisfaction to limit and detect deliberate risk selection and Discrimination. • Regulation and monitoring of marketing practices to limit selective or discriminatory marketing. • Specification of information to be made available to consumers and regulators and of Complaint-handling mechanisms. • Limits on medical underwriting. • Risk adjustment of employer or government payments to health plans. Oversight and Management Structures • Creation of one or more official bodies to implement the program and certify health Plan compliance. • Creation of regional purchasing cooperatives to select and oversee health plans that small (or all) employers would be required to offer.

OCR for page 167
Employment and Health Benefits: A Connection at Risk Some proposals include a variety of other features that are not aimed at risk selection but are instead concerned with providing coverage for low-income individuals (e.g., through some kind of subsidy) or increasing cost-conscious behavior. The first of the elements in Table 5.4—standardized benefit packages— is included in a wide array of proposals for health care reform (including those that would deal with risk selection by establishing a single national health insurance system). The definition of a standard benefit package poses a number of practical and theoretical challenges (Chollet, 1992b; IOM, 1992a). What principles will be applied in defining basic benefits? How will they reflect consumer, provider, payer, and other perspectives? Will they be sensitive to differences among patients with the same medical condition? What procedures will be followed to determine and update coverage features? There are a variety of ways of answering these questions, and considerable disagreement and uncertainty have been provoked by the approaches that have actually been tried. (Chapter 7 suggests these issues should be part of the government's research agenda.) The debate over the Oregon strategy for defining basic benefits, which was part of its Medicaid reform strategy, is a vivid case in point.15 Standardizing benefits will limit the choices available to individuals and possibly constrain desirable innovations in benefit design. However, the strategic use of benefit design to attract good risks and discourage bad risks appears to be so appealing that it will—if not limited—almost certainly undermine other steps to control risk selection and limit the advantages to be gained from such control. Monitoring health plan enrollment, disenrollment, and marketing strategies has been tried in the Medicare program to control abuses that might arise as the government encouraged beneficiaries to enroll in HMOs. These efforts have had some successes, but reviews have, in general, been mixed (GAO, 1991d; Welch, 1991). Obviously, the larger the number of insurers that choose to or are able to persist in a highly regulated market, the larger the number of separate entities that would have to be monitored. Most proposals that go under the "managed competition" label envision the creation of a quasi-public body or bodies to oversee the creation and maintenance of a regulated market and to serve as a purchasing agent for the self-employed, small employers, and perhaps others (see, for example, 15   A major objective of Oregon's proposed restructuring of its Medicaid program was to extend basic health care services to all needy citizens and to deemphasize expensive services of minimal benefit. The state initiated an extensive process to define priorities for coverage of treatments for various medical conditions based on their social importance, clinical effectiveness, and other factors. Of the 709 condition-treatment pairs initially ranked, the state appropriation would have covered the top 587 (Wiener, 1992). Implementation of the program will require a waiver of certain federal regulations (Firshein, 1992a).

OCR for page 167
Employment and Health Benefits: A Connection at Risk Enthoven, 1988a; Etheredge, 1990; Resnick, 1992). Depending on the specifics of the proposal, there might be (1) one or more purchasing agents; (2) relatively open or highly selective procedures to select or qualify participating health plans; (3) voluntary or mandatory participation by employers and others; and (4) requirements that employers offer at least one qualified plan, only qualified plans, or all qualified plans. To the extent that proposals provide for multiple purchasing agencies, voluntary participation by employers, and limited standardization of coverage, they would almost certainly create their own problems of risk selection. One critical question is whether the oversight and management entity (or entities) called for by proposals for managed competition can be made sufficiently accountable for its exercise of power. The converse question is whether any governmental or quasi-governmental entity can withstand the pressures that historically have led to constant expansions in coverage without regard to cost-effectiveness and budget constraints. Another issue is how vulnerable the protections offered by such a body would be to changing partisan tides regarding regulation and deregulation. Risk Adjusting Payments to Health Plans Fundamental to a number of proposals for health care reform, especially those based on ''managed competition," are methods to adjust how employers, governments, purchasing cooperatives, or other entities pay health plans based on the risk presented by their enrollees. Unlike medical underwriting, the idea is not to adjust the premium paid by the individual health plan member but rather to adjust the premium contribution from the employer, government, or other entity.16 The focus of risk-adjusted payment is on group insurance rather than on individually purchased coverage. The objectives are to reduce the financial advantage obtained from strategies to attract low-risk individuals and avoid high-risk individuals and to reduce the extent to which individuals are penalized for being members of a plan that has attracted higher-risk individuals. Policy and Strategic Issues Several criteria for risk adjustment schemes have been suggested (Welch, 1985; Newhouse et al., 1989; Anderson et al., 1990; Anderson, 1991b; Bowen 16   For example, a 30-year-old employee, a 50-year-old employee, an employee with diabetes, and an employee with no significant health problem would pay the same individual premium to any given health plan—unless the individual's contribution was linked to his or her income. The employer (perhaps with some kind of government contribution or subsidy) would contribute more to the health plan for the older employee and the employee with diabetes and less for the others.

OCR for page 167
Employment and Health Benefits: A Connection at Risk and Slavin, 1991; Luft, 1991; Robinson et al., 1991). Commonly mentioned criteria specify that a method should be based on characteristics of plan enrollees, not characteristics of the delivery system (e.g., inefficient management of health problems); related to individual need for health care rather than taste for more or less care; resistant to individual or organizational manipulation or "gaming" (e.g., misreporting health status); feasible to administer for many types of employers and health plans (e.g., HMOs, fee-for-service plans, and large and small groups); and compatible with other policy objectives (e.g., encouraging responsible use of medical care). In general, the strategies for risk adjustment attempted to date have not met one or more of the above criteria. A strategy need not be perfect to be helpful, but the most feasible of existing methods have not yet demonstrated enough power and practicality to serve their intended policy purposes. Several types of risk adjustment strategies are discussed below. Beyond the question of what kind of data should be and can be used to assess and adjust for risk is the question whether the adjustment should be prospective or retrospective or some combination of the two. Prospective adjustments permit health plans to budget and manage expected revenue and allow the adjusted premium contributions for individuals to be used during open enrollment periods. Retrospective adjustments are less administratively demanding and allow collection of some information (and therefore permit adjustments) that would not otherwise be feasible. Retrospective adjustments may also be helpful in monitoring health plans for selectively encouraging disenrollment of high-cost individuals. However, they may create problems should a health plan have to "refund" payments after its fiscal year has ended. This problem would be lessened by a system that was primarily prospective in administration with retrospective adjustments for specifically defined situations. Another issue related to risk adjustment strategies involves the definition and stability of the entire population to which the adjustments would be applied. If groups and individuals can easily join or leave the pool or if multiple, competing pools (e.g., competing purchasing cooperatives) are possible, then the problem of risk selection is unlikely to be resolved. Some kind of required pooling arrangement, such as the mandatory purchasing entities proposed in some "managed competition" legislation, may be necessary as part of an effective strategy to control risk selection. Improved risk assessment methods have value beyond the applications discussed above. For example, one concern about Medicare's method of paying for care provided to patients with end-stage renal disease is whether

OCR for page 167
Employment and Health Benefits: A Connection at Risk it adequately adjusts for differences in the severity of illness of patients across time or different providers (IOM, 1991). The same concern has been raised about Medicare's prospective payment system for hospitals (Gertman and Lowenstein, 1984; Iezzoni, 1989; Iezzoni et al., 1991; McGuire, 1991; Schwartz et al., 1991). In addition, efforts to compare the performance of health care providers are complicated by uncertainties about differences in their patient populations, especially when claims-based administrative data are used (Chassin et al., 1987; Greenfield, 1988; IOM, 1990b; Park et al., 1990; PPRC, 1992b). Similarly, although the IOM and others have argued that assessments of the effectiveness of different medical treatments need to move beyond highly controlled and artificial clinical trials to real-world settings, this move is made more difficult by the absence of inexpensive, practical methods of relating differences in outcomes to differences in patient risk (IOM, 1992a). Thus, improved risk adjustment methodologies may have multiple uses. Techniques for Risk Adjusting Payments to Health Plans Assessments of risk adjustment techniques involve both policy and technical challenges (Welch, 1985; Newhouse et al., 1989; Anderson et al., 1990; Anderson, 1991b; Bowen and Slavin, 1991; Luft, 1991; Robinson et al., 1991). The policy challenge is to determine whether adjusting government or employer contributions for a particular individual risk factor (e.g., use of over $50,000 in medical services) is consistent with other policy objectives (e.g., avoiding incentives for inefficient use of resources). The technical challenge is to devise appropriate measures of health risk that use readily available or easily collectible data and that are valid predictors. Specific risk adjustment techniques may use (1) demographic information, (2) data on health service utilization, (3) measures of health status, or some combination of these. Demographic Measures In some respects, the demographic approach to risk adjusting the health plan contributions of employers and governments parallels insurers' use of such variables to establish premiums for individually purchased insurance. The most familiar demographic risk adjusters are age and gender. The Medicare formula for paying HMOs (the average adjusted per capita cost, or AAPCC) includes these variables plus Social Security disability status, welfare status, and institutional status (e.g., nursing home residence). (The complete calculation is more complex and takes other factors such as geographic location into account.) Risk adjusters have also included education,

OCR for page 167
Employment and Health Benefits: A Connection at Risk income, occupation or job classification, length of employment, and marital status (Robinson et al., 1991). A major advantage of demographic data such as age and gender is that they tend to be more available than data on health care utilization or health status. The disadvantage is that demographic factors are relatively crude indicators of risk; for example, variations in health status, utilization, and cost within demographic categories can be quite substantial. One study of the Medicare AAPCC found that it could explain less than 1 percent of the variation in Medicare spending for the elderly (Lubitz et al., 1985). Few observers regard demographic data as sufficient elements of a successful risk adjustment strategy. Prior Use and Cost Measures Some risk assessment strategies have turned to measures of health care utilization or expense to improve the strength of their estimates. In doing so, they again parallel to some degree a strategy long used by insurers for employer groups, that is, experience rating. Adjustment based on prior utilization of health services may use past claims expense alone or also factor in information about types of utilization (e.g., days of hospital care, surgical care, and number of physician visits). Prior use and cost data are a much stronger predictor of expenditures than demographic factors (Newhouse et al., 1989). Even so, such data have been criticized as a basis for risk adjustment on several grounds (Luft and Miller, 1988; Anderson, 1991b). First, they may reflect inefficiencies in health care provision rather than differences in individual risk. Second, they may reflect individual tastes for consumption of health services. Third, they may reflect individual decisions to defer or advance medical care in conjunction with a planned switch in health plans. Fourth, certain measures may be susceptible to provider or patient manipulation. These problems have prompted efforts to develop utilization measures based on relatively nondiscretionary services or diagnoses. The objective of most of this work has been to improve the way Medicare pays HMOs that enroll Medicare beneficiaries. In one approach (Ellis and Ash, 1988; Ash et al., 1989), researchers developed diagnostic cost groups (DCGs) that involve diagnoses thought to entail less patient or physician discretion in the use of services. These less-discretionary diagnoses are weighted and grouped according to expenses projected for the year following hospitalization. In pilot projects testing the use of DCGs to risk adjust payments to HMOs for enrolled Medicare beneficiaries, "many of the HMOs have found that the current . . . method [the AAPCC] provides higher payment rates than does the [DCG]" (Anderson, 1991b, p. 22). Not surprisingly, these projects have suffered considerable attrition of participating HMOs.

OCR for page 167
Employment and Health Benefits: A Connection at Risk Another approach developed for Medicare (Anderson et al., 1990) also uses information about previous hospital admissions (e.g., major diagnostic category as used in Medicare prospective payment and chronicity of the condition) but adds information on whether the individual exceeded the Part B deductible. The rationale for adding this latter factor is that it will identify individuals who may not have been hospitalized but have used outpatient services. The criticism is that providers might be prompted to encourage patients to use enough care to meet the relatively low ($100) Part B deductible. Under this adjustment strategy, the utilization measures are then factored into a larger payment adjustment formula (the payment amount for capitated systems, or PACS) that also includes data on three other individual characteristics—age, gender, and Social Security disability status—as well as adjustments for provider input costs and urban/rural location. Health Status Measures The third major approach to risk adjusting payments attempts to measure health status using indicators such as mortality data, measures of functional status, self-reported health status, clinical diagnosis, and physiological indicators (Anderson, 1991b). Each measure or index of health status or health-related quality of life has strengths and weaknesses (see, generally, McDowell and Newell, 1987; Lohr, 1989, 1992; Spilker, 1990). Mortality data have been widely criticized as a limited indicator of health status, whether the purpose is to evaluate the performance of a health care system, practitioner, or procedure or to adjust per case or capitated payments for care to reflect the severity of illness of a patient population. More valid, stable, and direct are various multifactor measures of health status such as the Quality of Well-being Scale (Kaplan et al., 1989), the Sickness Impact Profile (Bergner et al., 1981; Temkin et al., 1989), and the Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36) (Stewart and Ware, 1992; Ware and Sherbourne, 1992). Although health status measures are becoming more useful for clinicians and quality improvement programs, they have not yet proved practically useful in predicting health care expenses for purposes of risk adjusting health plan payments. One problem is that health status measures generally involve data not easily available to those who would use them to risk adjust contributions to health plans. They require either direct access to medical records, direct questioning of individuals about their perceptions, activities of daily living, and other matters, or both. In addition, health status measures may be poor predictors of cost if they do not differentiate between people whose health status is good because a health problem is being successfully treated (at a cost) and people whose health status is less good

OCR for page 167
Employment and Health Benefits: A Connection at Risk because they are not being treated but who are not suffering costly adverse effects in the short term. Some studies have suggested that subjective generic measures of health status may be less useful than measures related to specific physical conditions in explaining variations in health care expenditures, although both may improve on explanations that rely only on demographic and prior use measures (Beebe et al., 1985; Howland et al., 1987; Newhouse et al., 1989). Although progress is being made in developing less costly and more useful measures of health status, such measurement will still be expensive. Its introduction outside the research setting is best justified as part of a broader strategy aimed at assessing the quality of health care, the performance of specific health care providers, and the effectiveness of alternative medical services. Reinsuring, Allocating, and Pooling High Risk Individuals Many advocates of underwriting reforms, risk-adjusted premiums, and managed competition concede that very high risk or high cost individuals may pose adverse selection problems beyond the reach of their approaches (Blue Cross and Blue Shield Association, 1991a; HIAA, 1991a; National Association of Insurance Commissioners, 1991). Thus, they have offered several strategies for spreading risk for these individuals, including reinsurance mechanisms to cover all or a portion of medical care expenses above a certain level for individuals or entire groups; assignment of high-risk individuals or groups to individual insurers on an unbiased basis; grouping of high-risk individuals in a single "pool" subsidized by contributions from private insurers, taxpayers, or others; and channeling of high-risk individuals to case management programs. Some of these strategies could be used together. For example, reinsurance arrangements might focus on groups with higher than average costs because of a higher incidence of relatively common, relatively expensive problems such as coronary artery disease or mental illness. The separate risk pool might be restricted to extraordinarily high cost individuals. One major question about all these mechanisms is whether the funding arrangements would be sufficient and fair, particularly in the context of a broader effort to extend coverage to the currently underinsured and uninsured. Costs for state risk pools are notorious for being underestimated, for reflecting severe adverse selection, and for burdening state budgets with higher-than-expected expenses (Bovbjerg, 1992). Risk pools also are criticized for relieving primary insurers of too much of the responsibility for managing costs and sharing risk. Current proposals for reform in the small-

OCR for page 167
Employment and Health Benefits: A Connection at Risk group market that include reinsurance provisions leave open the possibility of state funding if the reinsurance premiums paid by insured plans are insufficient, a likely prospect as long as self-insured groups remain outside the funding stream for any reinsurance or similar mechanism. Whether such funding would be forthcoming in amounts adequate to achieve policy objectives is far from clear. Another question is whether high-risk pools or other options would encourage continuity of care and effective management of high-risk individuals by mainstream health plans. Would they instead segregate these individuals in plans that are ill-equipped to provide state-of-the-art management of complex and difficult medical problems? Although not intended as a high-risk pool per se, the public plan component of "play or pay" reform proposals also appears vulnerable to selection by higher-risks. Employers with higher-than-average risk groups and premiums above the level required for the pay option (typically 7 to 9 percent) might find it financially attractive to drop their health plan so that their employees would be covered by the public plan. The likely impact would be higher-than-anticipated costs for the public program. CONCLUSION To a considerable extent, the history of health insurance in the United States is a history of efforts to overcome, exploit, or manage risk selection. Today, the debate over risk selection has many dimensions—empirical, normative, strategic, and technical. Opinions vary about how much selection occurs and why, which kinds of health benefit plans tend to have favorable or unfavorable selection, what the consequences generally are, whether these consequences create inequities or other problems in need of correction, and what corrections are feasible. This committee concludes that risk selection does occur, that it harms individuals and distorts competition, and that it can be difficult to detect and overcome. The fundamental issues for policymakers are these: First, can the deterioration in the insurance market be reversed and the market be made to work? Second, can high-risk individuals be reasonably protected against overt or covert discrimination in insurance or employment? Third, in a highly fragmented insurance market, can effective monitoring of marketing and underwriting be reasonably expected? Fourth, how good does a risk adjustment need to be in predicting or compensating for risk selection to be a useful policy tool? Managing risk selection has proved a difficult task in part because of technical difficulties (e.g., availability of information) and in part because the solutions may conflict with other objectives such as promoting a competitive health care market. There are several strategies to improve market

OCR for page 167
Employment and Health Benefits: A Connection at Risk functioning by reducing or compensating for risk selection. These involve (1) narrowing or eliminating differences in individual or group premiums based on age, gender, health status, or other risk factors, (2) limiting health plan discretion in benefit design, regulating marketing practices, and otherwise closely managing the terms of competition among health plans, (3) risk adjusting employers' or governments' (not individuals') payments to health plans to reflect differences in risk level of their membership, and (4) establishing special mechanisms for handling high-risk individuals. The committee recognizes that these policies and techniques will not directly attack the problems of increasing health care costs and may in fact increase costs for some while lowering them for others. However, some improvements in techniques for risk adjustment may help researchers better assess the performance of health care providers and better evaluate the value of specific medical services. The next chapter examines health care costs and considers further the question of value.