This Article Abstract Reprint (PDF) Submit a response to this article Alert me when this article is cited Alert me when Comments are posted Alert me if a correction is posted Services E-mail this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Add to My Personal Archive Download to Citation Manager Reprints & Permissions Citing Articles Citing Articles via HighWire Citing Articles via Web of Science (10) Citing Articles via Google Scholar Google Scholar Articles by Holahan, J. Search for Related Content PubMed PubMed Citation Articles by Holahan, J. Related Collections Access To Care Insurance - Business Coalitions

DataWatch

Health Status And The Cost Of Expanding Insurance Coverage

Abstract

 
This paper uses data on health spending and health status from the Medical Expenditure Panel Survey (MEPS) to estimate the differences in health spending across different types of insurance and across incomes that are attributable solely to health status differences. The results show that the uninsured are less costly than those on Medicaid, based on health status alone, but are more costly than those with employer-sponsored insurance. Adults and children with private nongroup coverage are also less expensive than average, because of better-than-average health. Finally, the data show that expenditures fall (health status improves) with income, regardless of type of coverage.

Policymakers also are often confronted with questions about the relative costs of persons served by different insurance arrangements. For example, average premiums in 1998 for U.S. firms that offer health insurance averaged $2,174 for singles and $5,590 for families.² In 1998 Medicaid costs per nondisabled adult enrollee were $2,706.³ Assuming that the cost for a family is about 2.5 times that for an adult, Medicaid costs for a family would be well over $6,000.

Are premiums in the private market lower than Medicaid costs because of the use of deductibles and coinsurance, or because of less generous benefit packages? Or is it because those with employer coverage tend to be relatively healthy? Is Medicaid more costly because of its broad set of benefits or because of the health problems of those that it serves? It is also believed that nongroup premiums vary greatly but tend to be relatively high.⁴ This is often attributed to high administrative costs, but it could be due to the health status of those who seek coverage in the nongroup market; that is, only those with poor health problems are willing to pay the high premiums. Finally, we know that the uninsured incur only about 50–60 percent of the costs that the insured incur.⁵ Is this because they use fewer services or because they are healthier?

In this paper we describe a method for summarizing differences in health status between the uninsured, Medicaid beneficiaries, and the privately insured. We then use this method to address two key policy questions: (1) To what extent do differences in health explain differences in cost across insurance groups? (2) How useful are private insurance or Medicaid data for estimating the cost of expanding insurance to the uninsured?

Methods: Developing A Health Status Cost Index

Top Methods: Developing A Health... Study Findings Discussion And Policy... NOTES

 
We use data from the Medical Expenditure Panel Survey (MEPS) to estimate the relationship between health spending per person and health status, using a regression equation to control for the effects of age, sex, income, education, race/ethnicity, geographic region, and type of insurance. The regression approach eliminates bias resulting from failure to control for factors correlated with health. For example, if those in fair or poor health are more likely to have lower education, which also affects spending, then the effect of education on spending will not bias the relationship between health status and spending. Furthermore, we only use data on those with public or private insurance for a full year, to avoid any distortions resulting from the effect of partial coverage or lack of coverage on use and spending. By including people with all types of insurance in the regression model, we average out the effects of variations in benefits, cost sharing, and provider payment rates across types of insurance.

From the regression model, we generate predicted spending for each type of health status, holding constant all other variables at their means. Thus, we are predicting spending by health status for a "standardized" person. In presenting the results, we choose as a benchmark children in excellent health, who have the lowest average spending per person. We divide other groups’ predicted average spending by the benchmark to create an index of how much more costly a particular group is relative to the benchmark.

Expenditures increase sharply as health status worsens (Exhibit 1). The benchmark group, children in excellent health, have the lowest average expenditures and are assigned an index of 1.00. Children in very good health have an index of 1.52; in other words, their expenditures are 52 percent higher than the benchmark.

Study Findings

Top Methods: Developing A Health... Study Findings Discussion And Policy... NOTES

 
Population health status. We used Current Population Survey (CPS) data to distribute children and adults by health status, type of insurance, and income. We used the CPS rather than MEPS because of the larger sample; the distribution of the population by health status (when both the insured and the uninsured are included) for both children and adults is very similar in both surveys. We did this for those with all incomes as well as for several income categories (under 100 percent of the federal poverty level, 100–200 percent, 200–400 percent, and over 400 percent). We then multiplied the number of people in each cell by the appropriate health status cost index. Thus, a person with a weight of 4.0 is four times more costly than a benchmark person. Summing these products (number of persons times health status weights) across health status groups within each type of insurance coverage yields the number of persons in terms of benchmark health status equivalents (Exhibit 2). The higher the per person cost associated with fair or poor health and the greater number of persons in fair or poor health, the larger the weighted total population.

Health status scores by income within insurance groups. Exhibit 3 provides summary measures indicating the group health status scores for persons at each income level. Group health status scores decline (that is, health improves) with income for both children and adults. Scores decline for those with employer-sponsored coverage, for those on Medicaid, and for the uninsured. Similar results are found for adults.

Health differences and cost differences. To estimate how variations in health status contribute to relative costliness, we divide each of the group health scores by the score for all people (separate calculations are made for adults and children) (Exhibit 4). If we compare children at all income levels, we see that on average uninsured children are 8 percent more expensive than the average child, solely because of differences in health status. Those in Medicaid are far more expensive than average. For example, children in Medicaid through SSI are 63 percent more expensive than average (acute care only). Medicaid/Temporary Assistance for Needy Families (TANF) recipients are 23 percent more expensive than average. The Medicaid/other group for children consists primarily of those eligible through the poverty-related expansions; these children are 15 percent more costly than average but less costly than other children in Medicaid. It is not surprising that Medicaid a costly program, given the health status of the three Medicaid groups.

The total column for children suggests that those below poverty are 14 percent more expensive than average, those at 100–200 percent of poverty are 7 percent more expensive than average, while those above 400 percent of poverty are 12 percent less expensive than average. Moreover, for each type of insurance the health indices generally decline as income increases. In other words, the poor with employer coverage are more costly than those above 400 percent of poverty with such coverage; the uninsured poor are more costly than the higher-income uninsured.

The results are generally similar for adults. Adults in Medicaid through SSI are more than twice as expensive on average as adults in general. (The relative costs of the SSI population may be understated for two reasons. First, there may be costs associated with disability that are not fully captured by the estimates of the effects of health status on relative costs across insurance groups, particularly those with more disabled persons. Second, our expenditure weights do not include long-term care.) The two other Medicaid groups are also much higher cost than the average adult. The "other" group is small but includes persons covered by Medicare (those with disability insurance and thus in state high-risk pools) and thus is also fairly costly. The uninsured are again 7 percent more expensive than average but are not as expensive as Medicaid beneficiaries. The uninsured are much more expensive than those with employer and private nongroup coverage (1.07 versus 0.92 and 0.93, respectively).

Again, spending declines with income. Uninsured adults below poverty are 28 percent more costly than the average adult. Uninsured adults, however, are less costly the higher their incomes.

Expansion of public programs to those with higher incomes would primarily cover the uninsured but would likely displace some employer and private nongroup coverage. It is clear, however, that the pool of new enrollees that would be brought into a new program is much healthier than are those in the existing Medicaid program. It is important to note that those who would enroll in a (voluntary) program expansion would likely be more costly than the average from the pool from which they are drawn—that is, there is likely to be some adverse selection. The health indices cannot simply be used without some adjustment to account for this selection effect.

Alternative expenditure weights. Our results are based on weights derived from spending on all persons who are insured for the full year. It is arguable that the results would be different if weights were based on those with private insurance or Medicaid for the full year. For example, spending on Medicaid beneficiaries compared with the privately insured may be relatively high for those in excellent health (because of the absence of cost sharing) and low for those in poor health (because of low provider payments). To test this, we derived weights based on those with employer coverage, Medicaid, and other insurance (there were not enough observations on children in the "other insurance" categories to create a reliable set of weights). In fact, the gradient of expenditures, between excellent and poor health, was flatter when based on Medicaid data than when based on data for those with private insurance. Nonetheless, the overall results do not change markedly. Exhibit 5 shows that for both adults and children the overall indices are essentially unchanged. The exception is the Medicaid SSI population where the weights (particularly for children) are sensitive to whether they are based on employer coverage or Medicaid data.

Top Methods: Developing A Health... Study Findings Discussion And Policy... NOTES

Third, as incomes increase, people become considerably less costly to insure relative to the poor. Fourth, those who have employer policies tend to be healthier than average, which would at least partially explain why the cost of employer plans is typically less than Medicaid costs per enrollee. Finally, those with private nongroup coverage are surprisingly healthy, suggesting that only those in good health are able to afford or obtain this type of coverage.

Editor's Notes

 
John Holahan directs the Urban Institute’s Health Policy Center in Washington, D.C.

This paper has benefited greatly from comments by Linda Blumberg, Bowen Garrett, Jack Hadley, and Steve Zuckerman and from the research assistance of Mary Pohl.

NOTES

Top Methods: Developing A Health... Study Findings Discussion And Policy... NOTES

 

1. J. Feder, C. Ucello, and E. O’Brien, The Difference Different Approaches Make: Comparing Proposals to Expand Health Insurance (Menlo Park: Calif.: Henry J. Kaiser Family Foundation, October 1999).
2. Agency for Healthcare Research and Quality, Medical Expenditure Panel Survey (MEPS) Insurance Component, Table I.D.1 (1998), <www.meps.ahrq.gov/mepsdata/ic/1998/TID1.pdf>(22 August 2001).
3. Urban Institute tabulations of Centers for Medicare and Medicaid Services (formerly HCFA) 2082 data.
4. K. Pollitz, R. Sovinar, and K. Thomas, How Accessible Is Individual Health Insurance for Consumers in Less than Perfect Health? (Menlo Park, Calif.: Kaiser Family Foundation, June 2001).
5. M.S. Marquis and S.H. Long, "The Uninsured Access Gap: Narrowing the Estimates," Inquiry 31, no. 4 (1994): 405–414.

                      What's this?

This article has been cited by other articles:

				J. R. Gabel, R. McDevitt, R. Lore, J. Pickreign, H. Whitmore, and T. Ding Trends In Underinsurance And The Affordability Of Employer Coverage, 2004-2007 Health Aff., July 1, 2009; 28(4): w595 - w606. [Abstract] [Full Text] [PDF]

				J. Holahan and A. Weil Toward Real Medicaid Reform Health Aff., March 1, 2007; 26(2): w254 - w270. [Abstract] [Full Text] [PDF]

				L. Boulanger, O. Hauch, M. Friedman, T. Foster, D. Dixon, G. Wygant, and J. Menzin Warfarin Exposure and the Risk of Thromboembolic and Major Bleeding Events Among Medicaid Patients with Atrial Fibrillation Ann. Pharmacother., June 1, 2006; 40(6): 1024 - 1029. [Abstract] [Full Text] [PDF]

				T. C. Buchmueller, K. Grumbach, R. Kronick, and J. G. Kahn Book Review: The Effect of Health Insurance on Medical Care Utilization and Implications for Insurance Expansion: A Review of the Literature Med Care Res Rev, February 1, 2005; 62(1): 3 - 30. [Abstract] [PDF]

				M. P. Cote, T. Byczkowski, U. Kotagal, S. Kirk, M. Zeller, and S. Daniels Service quality and attrition: an examination of a pediatric obesity program Int. J. Qual. Health Care, April 1, 2004; 16(2): 165 - 173. [Abstract] [Full Text] [PDF]