Medicaid, the uninsured, and national health spending: federal policy implications.

Implications are discussed for Federal policy of "gap-filling" initiatives at the State and Federal level to deal with the problem of the uninsured. Measures currently under active consideration that involve expansions of Medicaid and employment-related insurance are considered in the light of recent studies of the uninsured and recent simulations of their cost and coverage impacts. The limited impact of these gap-filling measures on additional national health spending, in contrast to program costs and Federal outlays, is emphasized. Placing greater emphasis on this broader societal perspective could assist Federal policymakers in developing an acceptable strategy for covering the uninsured.

Introduction

States are undertaking new program initiatives to deal with the problem of the uninsured. Merrill (1990) describes these initiatives and argues for the necessity of greater public-private sector cooperation and for policy changes at the Federal level. These State efforts raise the question of what the role of the Federal Government should be. What can the Federal Government do to facilitate these initiatives? How far would this go toward addressing the problem of the uninsured? What are the barriers to this type of Federal action? These issues are discussed in light of our growing body of knowledge about the uninsured and their characteristics.

Characteristics of the uninsured

In the last 5 years, health services research has done much to improve our understanding of the nature and extent of the problems with insurance coverage. Although there is some controversy about the absolute number of uninsured persons (stemming from the specifics of survey questions), there is general agreement that the number is large—on the order of 31 to 37 million at any point in time (between 13 and 16 percent of the population). Furthermore, although the number of uninsured and their proportion (as a percent of the nonelderly population) rose in the first half of the eighties, since then the percentage has remained more or less constant (Brown, 1989). In addition, four stylized facts about the characteristics of the uninsured have become part of the conventional wisdom that shapes the debate:

Nearly three-quarters of the uninsured are workers or dependents of workers.

About one-third of the uninsured are in families with incomes below the poverty level.

Nearly 50 percent of the uninsured are under age 24.

One-fourth of the uninsured have incomes greater than three times the poverty level.

These facts would seem to argue for a combination of gap-filling measures such as mandating or subsidizing employment-related insurance, expanding the Medicaid program, and even direct service programs for school children. The fact that so many uninsured are workers suggests that we should build on our current, largely employment-based system. The fact that only one-third are poor suggests that Medicaid expansions alone are unlikely to reach a majority. The fact that so many are young places a greater moral burden on society as a whole. And the fact that many have substantial incomes, far above the poverty level, suggests the difficulty of covering everyone through a gap-filling strategy. Though such a strategy could conceivably reach the vast majority of the uninsured, short of a mandate that individuals have insurance, these measures would not necessarily cover every person.

Gap-filling strategies: coverage and costs

Through a number of modeling efforts in the last few years, we now have a fairly good notion of the likely impact of such strategies. Three principal and important conclusions emerge from this literature:

A strategy using a combination of gap-filling measures would cover most of the uninsured.

The strategies vary greatly in how the costs would be distributed among individuals, firms, State governments, and the Federal Government.

The net addition to national health spending of these improvements in coverage would not be large in relation to national health spending.

A brief summary of key findings from recent major studies follows.

In 1988, the Congressional Research Service (1988b) conducted a study of several combinations of employer-based plans coupled with a Medicaid expansion. Using the Lewin-ICF Health Benefits Simulation Model (based in part on 1980 National Medical Care Utilization and Expenditure Survey data), they estimated that a Medicaid expansion to cover all of the poor (defined as those below the poverty level) would leave 29 million uncovered. Although such an expansion would increase Medicaid program costs by $13.3 billion, it would increase national health spending by only $4.1 billion (if it had been in effect in 1986). If combined with a “tailored” employer-based plan (i.e., emphasizing preventive services and having smaller deductibles), the number of uninsured would be reduced from 37 million to 6 million. The resulting increase in national health spending would then be $16.9 billion.

Enthoven and Kronick (1989a; 1989b) developed a “consumer-choice” health plan based on a strategy of managed competition among public and private health plans to promote informed, cost-conscious consumer choice. Their proposed system would be universal, covering all of the currently uninsured, and it would rely on a combination of mandated employer coverage for full-time workers and subsidies for the poor (below 150 percent of the poverty level) and others not covered through employment. According to Congressional Budget Office estimates for 1988, their expanded employment-related coverage would cover about 22 million of the currently uninsured. New Federal expenditures would be approximately $12.8 billion, but there would be $12.4 billion in new tax revenues. They project a one-time, $15 billion increase in national health spending in the first full year of implementation.

Thorpe, Siegel, and Dailey (1989) analyzed the impact of Medicaid expansion with a model based on Medicaid program data on age-specific program costs. Similar to the Congressional Research Service (CRS), they found that in 1987-88 an expansion to cover all the poor would leave 26.1 million uninsured. Net public spending would rise by $7.75 billion, and new national health spending would be less than this, though the cost depends on the richness of the package in terms of covered services and the level of provider payments.

Thorpe (1989) extended the preceding analysis to examine the impacts of an employer mandate in addition to a Medicaid expansion. He found that the various employer mandates could extend coverage to 24.6 million of the 37 million uninsured and could affect the coverage of many other workers, depending on the nature of the mandate. Adding Medicaid coverage up to the poverty level to this would result in total coverage of more than 30 million of the previously uninsured. And national health spending is projected to increase by only $12.5 billion under this combination. Also demonstrated in this article are the drastically different distributional impacts of different types of mandates.

In recent work for the Health Care Financing Administration, Lewin-ICF, using the same model as in their work for CRS, projects that in 1989 a Medicaid expansion to cover all the poor plus pregnant women and infants up to 185 percent of the poverty level would leave 24.6 million uninsured (Needleman et al., 1990). Allowing a buy-in to the Medicaid program for others between the poverty level and 185 percent of the poverty level would reduce this to 21.3 million uninsured. An employer mandate requiring coverage of full-time and most part-time workers would, alone, leave only 6.2 million uninsured. And if coupled with a Medicaid expansion, this could reduce the number of uninsured to 2.7 million. The change in national health spending from various combinations of options would range from $11 billion to $19 billion, depending on whether provider reimbursement is increased.

Holohan and Zedlewski (1989) examined a much wider range of Medicaid expansion options, ranging up to 200 percent of the poverty level. Their simulations use the Urban Institute's Transfer Income Model augmented with enrollee cost estimates based on Medicaid Tape-to-Tape data. Coverage (through subsidized Medicaid buy-ins, etc.) would leave 11.3 million uninsured if extended only to those below 200 percent of the poverty level. They find that Medicaid program costs could rise substantially ($47.2 billion) under the highest of these options. However, they do not present an estimate of the change in national health spending.

A detailed review of the preceding studies would reveal that the distribution implications of these alternatives vary dramatically and are quite complex. Medicaid expansions would place the burden on the Federal and State governments (and ultimately the taxpayers), roughly, 60 percent for the Federal Government and 40 percent for the State. They also tend to reduce other public spending on uncompensated care as well as out-of-pocket spending by the uninsured. The impact of employer mandates depends on assumptions about employee cost-sharing, changes in tax subsidies, and the nature of compensation. In addition, who bears the costs may vary in the short versus long run, particularly because firms may bear substantial costs immediately that are ultimately borne mostly by individuals as part of compensation. Also, there may be long-run employment effects (as the minimum wage is effectively raised) and impacts on consumers (if prices rise as a result). The results of these simulations also depend on the following assumptions about:

The set of services in the benefit package.

The treatment of those covered by VA/CHAMPUS.

The incentives for some small firms not to participate in some arrangements, in effect pushing their employees to the public plan.

Differentials in the per unit prices currently paid by different insurers.

Taken as a whole, these studies suggest that national health spending would increase only between $10 and $20 billion as a result of combined implementation of these gap-filling measures, which would reach the vast majority of the uninsured. This amount does not seem large in comparison with total national health spending in 1988 of $540 billion, which was an increase of $51 billion over 1987.

Service use by the uninsured

This range of impact on national health spending is corroborated from another perspective, based on use of services by the uninsured. It is suggested from previous studies that the uninsured receive at least 50 percent and, perhaps, up to 70 percent of the amount of care they would receive if insured. This is demonstrated in Table 1, in which a summary of estimates from several studies of the utilization of hospital and physician services by the uninsured is given. Also cited in the table is a recent paper by Long and Rodgers (1989) in which they review this literature in detail and conduct empirical analyses of data from the Survey of Income and Program Participation (SIPP), a unique longitudinal data source. Their findings indicate that, with appropriate correction for the dynamic nature of insurance status, these estimates of use by the uninsured may be overstated. Thus, the 50-percent end of the range may be a better estimate.

Table 1

Survey results on utilization of care by the uninsured as a percent of insured use, by survey

Survey	Physician visits1	Hospital services2
1977 National Medical Care Expenditures Survey	Percent
Davis and Rowland (1983)	65	52
Berk and Wilensky (1984)	68	54
1980 National Medical Care Utilization and Expenditures Survey
Long and Settle (1985), low-income adults and children	68	33
1984 Survey of Income and Program Participation
Long and Rodgers (1989)	63	31
1986 National Health Interview Survey
Congressional Research Service (1988a)	64	76
Congressional Research Service (1988a) individuals < $15,000 income	53	60
Long and Rodgers (1989)	75	63
1986 Robert Wood Johnson Survey
Freeman, Blendon, Aiken, et al. (1987)	72	80

Physician services are measured by number of visits in past year or past month.

Hospital service use is measured by either total hospital days or admissions per annum (or in the last month).

SOURCE: This table is adapted in large part from Table 1 in Long and Rodgers (1989).

In any case, within this range, the estimates of the impact of covering the uninsured on national spending are plausible for the following reason. In 1987, average health spending for persons under 19 years of age (who were 31 percent of the nonelderly uninsured in 1987) averaged $745, and for persons 19-64 years, it was $1,535 (Waldo et al., 1989). Thus, if the uninsured had received care at these average expense levels, their total expenditures in 1987 would have been $47.1 billion. If only 50 percent of this was actually provided, then expanded coverage would result in additional national health expenditures of $24 billion. Or, if 70 percent is currently provided, then national health expenses would increase by only $14 billion. This crude calculation could be adjusted for health status and other factors, but even at this level, it provides a consistency check on the results of the microsimulation models. The bottom line is that the additional spending to cover the uninsured is unlikely to represent more than 5 percent of what national health spending would be otherwise.

Dynamics of health insurance coverage

One limitation of the modeling efforts discussed previously is that they are based on a count of the number of uninsured at a given point, with little consideration given to the dynamic nature of insurance status. Two recent studies using data from SIPP highlight this issue. Nelson and Short (1989) found that over a 28-month period 28 percent of the population was without health insurance for at least 1 month. This is much higher than the 13-16 percent at a point in time. They also estimate that only 4 percent of all persons were without coverage for the entire period and another 3 percent had coverage for less than 6 months. Second, also using the SIPP, Monheit and Schur (1988) estimate that over a 32-month period 22 percent of those who began with private insurance lost coverage at some point, if only for a short time. Conversely, only 27 percent of those who began the period uninsured remained so the entire period.

These important new findings, highlighting the dynamic nature of insurance coverage, imply that current microsimulation models based on point-in-time estimates of the number of uninsured could have serious shortcomings. In particular, this implies that the administrative burden associated with these changes in status is probably large and would be somewhat larger if employer coverage is mandated. This additional administrative burden would increase national health spending above the levels suggested earlier. It is less clear, however, how accounting for the substantial turnover in the uninsured population should affect estimates of medical care expenses. In their analysis of the SIPP data, Long and Rodgers (1989) found no evidence that these transitions in status affect the average differential in use between the insured and uninsured. It may well be that the current models provide reasonably accurate medical expenditure estimates but fall short primarily in considering the additional administrative costs. Of course, recognition of high rates of turnover in the uninsured population may have important programmatic, social, and political implications, in terms of the viability of reforms.

Implications for Federal policy

In addition to Medicaid expansions, Merrill (1990) places State initiatives into three general categories: employer mandates (such as “play or pay”), insurance subsidies, and risk pools. Although there is much to learn from these State efforts, there is also good reason to believe that they will not adequately address the problem of the uninsured in the near term. At the very least, as Merrill notes, effective Medicaid expansions will require Federal action to decouple the system from welfare status.

With regard to the Massachusetts play-or-pay plan in particular and State plans in general, Enthoven and Kronick (1989b) argue that such innovations are unlikely to succeed for three reasons. First, the pre-emption of the Employment Retirement Income Security Act of 1974 allows firms that are self-insured to operate outside of State control. Second, there are potential adverse employment impacts if States with these provisions become less desirable business sites. Third, such plans do not take full advantage of the Federal tax subsidy to health insurance. Yet, these State initiatives may be most informative about our ability to extend coverage, via direct subsidies as well as reinsurance, to the significant numbers of uninsured workers in small businesses. Nonetheless, given the substantial variability among States in their Medicaid programs, it is difficult to see a way around these problems without, ultimately, the establishment of a broader Federal framework.

As many have noted, the much heralded cost-containment initiatives of the eighties have done little to slow the growth in national health spending. They may well have shifted more of the cost of care for the uninsured to public hospitals and to the uninsured themselves. Although the proportion of the population that is uninsured has not grown in recent years, this persistent problem is perhaps increasingly seen as a national problem, if not a national embarrassment, that must be addressed.

A major implication of the results summarized here is that this debate might be improved by an emphasis on the impact on the change in national health spending rather than on program costs or Federal outlays. Given the Federal deficit situation, the usual focus on program costs is understandable. Nonetheless, the additional costs to society of providing coverage to the uninsured are probably not as substantial as many would imagine. Federal policymakers should be able to take a societal perspective on this issue. In their efforts to forge a workable combination of Medicaid reform and other gap-filling measures, which clearly could cover the vast majority of the uninsured at a reasonable cost, these policymakers would be well-served to adopt and to present this broader perspective.

In summary, in answer to the questions posed at the outset of this discussion, there is much the Federal Government can do to improve coverage of the uninsured both directly and indirectly through State initiatives. In particular, decoupling Medicaid from welfare status and developing a buy-in option are steps that could lead to coverage of over one-half of the uninsured. Helping the remainder (of mostly working uninsured) is more difficult through State initiatives alone, though State initiatives may represent our best opportunity for learning what innovations are feasible. The distributional consequences of alternative gap-filling measures vary greatly, which is one of the reasons it has been so difficult to construct a political solution to this problem. Although the political attention given to these distributional consequences is understandable, the prospects for a viable solution could be improved if Federal policymakers would keep the costs of these initiatives in view from a broader societal perspective.