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 ISI Web of Science (10) Citing Articles via Google Scholar Google Scholar Articles by Martin, A. Articles by Hinman, L. Search for Related Content PubMed PubMed Citation Articles by Martin, A. Articles by Hinman, L. Related Collections State/Local Issues Health Spending

Health Care Costs

Health Care Spending During 1991–1998: A Fifty-State Review

Abstract

 
Health care spending varies considerably across states. Spending per person ranged from $2,731 in Utah to $4,810 in Massachusetts in 1998, with Medicaid’s share of total health care spending ranging from 9.1 percent in Nevada to 31.5 percent in New York. Research has suggested many reasons for such differences, including socioeconomic and demographic factors, market forces, and diversity in practice patterns. By using consistent methodologies among states, these 1991–1998 estimates, last produced for 1991 alone, will further the understanding of these differences.

In this paper we update earlier per capita personal health spending estimates by state for 1991 and present new estimates for 1992 through 1998.² These estimates are useful at the state level in defining spending, growth trends, the mix of services purchased, and the role of public programs such as Medicare and Medicaid in funding health care. Because these estimates use consistent definitions across states, researchers can analyze health spending patterns against those in other states and use them as baselines to measure the impact of proposed policy changes. While they reflect the economic impact of factors such as utilization, practice patterns, public policy, and socioeconomic and demographic differences on health spending, the analysis raises questions of fairness and equity of health spending that cannot be explained by these factors alone.

Study Methods

Top Study Methods Analysis Of State Spending Factors That Influence States'... NOTES

 
Per capita health spending estimates cannot be directly constructed from available data on state-of-provider spending.³ State-of-provider estimates reflect spending for services delivered in that state to residents and nonresidents, while the population estimates used to construct per capita estimates are based on residency. Using interstate border-crossing expenditure flow patterns, we adjusted the provider-based data to estimate health spending based on state of residence.

Provider-based estimates. First, we estimated state health care spending by location of provider.⁴ We relied heavily on economic census information (available by state once every five years), Internal Revenue Service state tabulations of receipts of for-profit health care businesses, the American Hospital Association’s Annual Survey of Hospitals, and prescription drug sales collected by IMS, as well as other data on population, wages and salaries, and payers’ expenditures.

Beneficiary-based estimates. Next, we adjusted state expenditures from a provider to a beneficiary-residence basis. We separated our provider-based estimates into three payer components: Medicaid, Medicare, and all other payers. Medicaid spending estimates were based on state data provided by the agencies that pay health care costs for eligible residents. Because states do not pay Medicaid benefits for nonresidents and because almost all care paid for by Medicaid is provided by in-state providers, we assume that Medicaid spending by state is identical on a residence and provider basis.

For Medicare, we adjusted spending from a provider to a beneficiary-residence basis using Medicare claims data. Medicare is the only nationwide insurer with publicly available claims files containing a large pool of service-specific records upon which to base interstate flows of spending between providers’ and beneficiaries’ residence locations.

Generally, Medicare fee-for-service claims data are also used to adjust non-Medicare, non-Medicaid expenditures, although the specific procedures vary depending on the service category. For example, for inpatient hospital and physician services, we know that travel patterns differ between Medicare and non-Medicare, non-Medicaid populations. These differences primarily reflect differences in the age distributions of each population and the fact that different age cohorts tend to consume varying mixes of specific procedures and services.⁵ Thus, Medicare expenditure flows are reweighted, using private hospital discharge information and physician claims records, to account for the distinct bundle of specific inpatient hospital and physician procedures purchased by the privately insured population under age sixty-five.⁶ Then, non-Medicare, non-Medicaid spending based on location of provider is adjusted to a state-of-residence basis using the reweighted Medicare expenditure flow patterns.

For most other services, no other private sources are available to adjust the Medicare data for service-mix. For services where Medicare pays benefits, Medicare interstate spending flows are used as a proxy for non-Medicare spending flows. The analysis of Medicare spending flows shows that interstate travel for other services is small. For some services (such as prescription drugs and dental services) Medicare data are unavailable, and thus no adjustment is made.

Caveats. Because of data limitations, state estimates presented in this paper do not adjust for international flows of health care spending. Health purchases in the United States by residents of other countries have the potential to overstate health spending in certain states, while purchases by U.S. residents in other countries would understate spending.⁷ In addition, resident population estimates from the U.S. census do not adjust for the population undercount, which could overstate per capita spending to varying degrees in different states.

Analysis Of State Spending

Top Study Methods Analysis Of State Spending Factors That Influence States'... NOTES

 
Spending levels. In 1998 we spent $1.0 trillion on personal health care expenditures in the United States, or $3,759 per resident (Exhibit 1). Spending in Massachusetts (the highest spender) was 76 percent higher than in Utah (the lowest).⁸ In general, spending tended to be higher in the Northeast and lower in the West. Four states had average spending of 15 percent or more above the U.S. average in 1998, and five states had average spending of at least 15 percent below average (Exhibit 2).

Medicare. The proportion of a state’s population enrolled in Medicare because of disability or age and the distribution by age of enrolled beneficiaries help to explain the variation in Medicare spending among states.⁹ Nationwide, Medicare enrollees represented 14 percent of the total population in 1998; Medicare accounted for 20.6 percent of all personal health care spending that year (Exhibit 3).¹⁰ Florida had both the largest share of enrollees and the largest share of total personal health spending funded by Medicare. Medicare enrollees accounted for the smallest share of the population in Alaska, where Medicare paid for only 9.1 percent of total personal health care in 1998, the lowest share nationwide.

In 1998 Medicare expenditures per enrollee were the lowest in South Dakota. Low per enrollee spending can be attributed to low utilization rates of short-stay hospital, physician and supplier, and home health agency services, which translates into some of the lowest average Medicare payments per enrollee for these services nationwide. Per enrollee Medicare expenditures were the highest in Louisiana, where Medicare short-stay hospital spending per enrollee was higher than average. However, extremely high utilization and payments per enrollee for home health agency services in 1998 boosted Louisiana’s total per enrollee spending by more than 10 percent over what it would have been had its Medicare home health expenditures been at the U.S. average. The federal government’s efforts to combat fraud and abuse in the delivery of Medicare home health services in the late 1980s through the mid-1990s, combined with the Balanced Budget Act (BBA) of 1997, which enacted new home health payment mechanisms, helped to curb growth in home health spending in this and other states.

Medicaid. Whereas eligibility requirements, benefits, and payment policies (after taking into account geographic location) are governed nationwide for Medicare, state Medicaid programs have more flexibility. Combined with different federal matching rates and state-specific budget constraints, this flexibility leads to variation in Medicaid spending by state.¹² Some states cover a large portion of their population and have broader coverage of services and, therefore, tend to incur higher expenses. Furthermore, the inappropriate use of disproportionate-share hospital (DSH) payments, which are intended to increase payments to hospitals that serve disproportionate numbers of low-income persons, introduced additional variation in state spending during 1991–1998 (although legislative action in 1993 and 1997 placed limits on the use of these payments).¹³

New York Medicaid funds the highest proportion of personal health care expenditures of any state, with per enrollee spending almost twice Medicaid’s nationwide spending per enrollee (Exhibit 4). Underlying causes of New York’s high spending include a higher-than-average percentage of the population with incomes at or below the poverty level, large numbers of people with serious health complications such as HIV infection, and high medical prices.¹⁴ This manifests itself in New York’s Medicaid program through higher hospital admission rates and longer lengths-of-stay and a broad range of services for the disabled population, including extensive services through home and community-based waivers and personal care services. In addition, generous eligibility standards, extensive use of DSH payments, and Medicaid-funded graduate medical education (GME) subsidies further boost New York’s Medicaid spending.¹⁵

Rural areas tend to be net importers of services (net flow ratios greater than 1.0), where more services are used by residents than are produced in the state. Residents of net-importing states such as Wyoming, Idaho, and Mississippi may cross borders seeking health care providers that are not available near their residence in their own state.

Research also has shown that border crossing for health care tends to be more predominant for high-technology procedures, such as advanced imaging, cardiovascular surgery, and oncology procedures, than for more routine evaluative services.¹⁷ Interstate net flow ratios show the greatest variation for inpatient hospital and physician services, as patients seeking nonemergency, highly specialized care are more likely to travel greater distances. Urgent care (such as in emergency departments) as well as home health care, nursing home care, and dental services typically are provided locally and show smaller variation in interstate flow ratios.¹⁸

Factors That Influence States’ Spending

Top Study Methods Analysis Of State Spending Factors That Influence States'... NOTES

 
Socioeconomic and demographic factors. Researchers have identified various socioeconomic and demographic characteristics that influence per capita health care spending. For example, the age distribution of a state’s population is a key determinant. This results in part from the disproportionate distribution of the population age sixty-five and older among states (ranging from under 6 percent of the state’s total population in Alaska to more than 18 percent in Florida) combined with higher per person health care spending by this age cohort. For example, persons in this age group use, on average, six times the health care of persons under age eighteen, and they use increasingly more medical technology than do other age cohorts.¹⁹

Income also affects health care spending, although the sensitivity of health spending to differences in income depends on the level (individual, state/region, or national) at which these variables are analyzed.²⁰ For individuals, health spending’s sensitivity to income is small or negative.²¹ This reflects the tendency of third-party insurance to insulate individual consumption decisions from income constraints. At the other extreme, international comparisons report greater sensitivity to income differences, largely reflecting variation in national resources measured as gross domestic product. Studies of regional and state health spending patterns report sensitivity to income that fall midway between those of individuals and of nations.²² In the case of states, federal expenditures through public programs such as Medicaid dampen but do not eliminate the effect of state-specific income differences on health spending through the transfer of federal tax revenues from high- to low-income states.²³ As a result, average income differences explain a small portion of the variation in health care spending among states.

Finally, notable differences in access to and use of medical care by race and ethnicity exist within disease categories and types of health services.²⁴ These differences are evident even after other factors such as socioeconomic status and insurance coverage are controlled for, and they may reflect differences in health status and in cultural norms and preferences.

Market factors. Market factors influence spending differences among geographic areas. One theory suggests that higher concentrations of physicians (especially specialists) to population will generate higher levels of health care spending because physicians deliver or order most of the health care that is used. Although this is subject to debate, it is borne out in studies of the Medicare population.²⁵ Our estimates are also consistent with that theory: Physician-to-population ratios were highest in the Mideast and New England regions, and lowest in the Southwest and the Rocky Mountains, paralleling the highs and lows in regional health spending.²⁶

Researchers also theorize that for many conditions the decision to hospitalize (and, therefore, differences in spending) depends in part on the availability of local hospital inpatient beds. This is because consensus does not exist on treatment protocols for some common conditions. Consideration of local hospital capacity may sway a physician’s judgment in such cases and raise (or lower) the overall rate of hospitalization.²⁷

In the past decade the rapid expansion of managed care has been cited for slowing the growth in health spending. The shift of enrollees to managed care lowered the cost of premiums and tended to slow the growth of health care spending during the 1990s. Studies of health maintenance organizations (HMOs), one form of managed care, suggest that large HMO market penetration lowers overall employer-sponsored health care premiums in that market. Both lower premium rates charged by HMOs and spillover effects of competition on non-HMO premiums contribute to this finding.²⁸

The United States spends more per person on health care than does any other developed country.²⁹ Within our high average spending level, there is wide regional variation. Researchers attribute some of the variation among states to factors associated with differences in patient demographics and socioeconomic characteristics—factors that may be difficult to change in the near term. On the other hand, spending variation resulting from the concentration of health care resources, government financing policies, and other marketplace factors; practice pattern variations, including health care decisions at the end of life; state mandates; and other factors may more likely be influenced by research developments and policy changes. Since 1998 health care cost growth has accelerated, and recent projections for the near term suggest that pressure on both the public and private sectors’ ability to pay for care will likely increase.³⁰ Measuring the impact of factors accounting for spending across states would help leaders to design policies that are better able to limit cost growth without loss of quality of care.

Editor's Notes

 
Anne Martin, Lekha Whittle, Katharine Levit, and Greg Won are economists in the National Health Statistics Group, Office of the Actuary, at the Centers for Medicare and Medicaid Services (CMS) in Baltimore. Lindy Hinman, formerly with the CMS Office of Research, Development, and Information, is a program examiner in the Executive Office of the President, Office of Management and Budget.

The authors thank Jonathan Smith and Mark Zezza of the Office of the Actuary at the Centers for Medicare and Medicaid Services (CMS) for their assistance in computer programming and in analyzing factors associated with differences in spending levels among states. The opinions expressed are the authors’ and do not necessarily represent those of the CMS or of the Office of Management and Budget.

NOTES

Top Study Methods Analysis Of State Spending Factors That Influence States'... NOTES

 

1. Dartmouth Medical School, Center for the Evaluative Clinical Sciences, The Dartmouth Atlas of Health Care 1998, www.dartmouthatlas.org/98US/tocl.php (28 February 2002); and L. Newman, "New Dartmouth Atlas: Improving U.S. Cardiac Care?" Lancet 356, no. 9230 (2000): 660.[Medline]
2. J. Basu, H. Lazenby, and K. Levit, "Medicare Spending by State: The Border-Crossing Adjustment," Health Care Financing Review 17, no. 2 (1995): 219–241[Medline]; and J. Basu, "Border-Crossing Adjustment and Personal Health Care Spending by State," Health Care Financing Review 18, no. 1 (1996): 215–236. [Medline]Personal health care expenditures by state are controlled to estimates presented in K. Levit et al., "Health Spending in 1998: Signals of Change," Health Affairs (Jan/Feb 2000): 124–132.
3. A. Martin, L. Whittle, and K. Levit, "Trends in State Health Care Expenditures and Funding: 1980–1998," Health Care Financing Review 22, no. 4 (2001): 111–140. The data are also available online at www.hcfa.gov/stats/nhe-oact/stateestimates.[Medline]
4. Ibid.
5. Variations in travel patterns are attributable to differences in the mix of specific procedures and services purchased by various age cohorts within broader inpatient hospital and physician service categories, rather than to differences in travel patterns exhibited by each age cohort for the same procedure. Fu Associates, "Interstate Flows of Health Spending: Updates for 1992 and 1993" (Memorandums dated 8 October 1996, 7 November 1996, and 29 January 1997, Contract no. HCFA 500-95-0036, prepared for the Centers for Medicare and Medicaid Services, Baltimore, 1996).
6. Inpatient hospital discharge information from Codman Research Group and claims records for private non-Medicare physician services from a large employer database were used to reweight expenditure flows to reflect the procedure-specific bundle of services used by the non-Medicare, non-Medicaid population. See Basu, "Border-Crossing Adjustment," for further information on service-mix adjustments and using Medicare flows as proxies.
7. A small percentage of U.S. citizens who are Medicare beneficiaries receive health care services (paid for by Medicare) outside of the fifty states, and a small percentage of U.S. citizens living in the U.S. territories and in other countries return to the United States for health care. The net flow of these travel patterns amounts to a net reduction of 0.1 percent in overall Medicare spending. Similar adjustments for the non-Medicare, non-Medicaid populations are not possible.
8. Per capita spending in the District of Columbia was higher than spending in any state in 1998, at $6,656.
9. B. Gage, M. Moon, and S. Chi, "State-Level Variation in Medicare Spending," Health Care Financing Review 21, no. 2 (1999): 85–98.[Medline]
10. CMS, "Medicare and Medicaid Statistical Supplement, 2000," Health Care Financing Review (June 2001): 104. Medicare does not cover all spending for beneficiaries. Certain categories of spending, such as prescription drugs, are not covered, and beneficiaries are responsible for certain copayments and deductibles.
11. Dartmouth Center for Evaluative Clinical Sciences, The Dartmouth Atlas of Health Care 1998.
12. J. Holahan and D. Liska, Variations in Medicaid Spending among States, New Federalism: Issues and Options for States no. A-3 (Washington: Urban Institute, 1997).
13. T.A. Coughlin and D. Liska, The Medicaid Disproportionate Share Hospital Payment Program: Background and Issues, New Federalism: Issues and Options for States no. A-14 (Washington: Urban Institute, 1997).
14. U.S. Bureau of the Census, "Table 184: AIDS, Syphilis, Tuberculosis, and Measles Cases Reported by States: 1999," Statistical Abstract of the United States: 2001, www.census.gov/prod/2002pubs/01statab/health.pdf (24 April 2002); and "Table 684: Persons below Poverty Level by State: 1980 to 1999," Statistical Abstract of the United States: 2001, www.census.gov/prod/2002pubs/01statab/income.pdf (24 April 2002). Also, Thomas Fanning, New York State Department of Health, Office of Medicaid Management, personal communication, 11 April 2002.
15. Public Policy Institute of New York State, "Medicaid: Wreaking Havoc in Health Care," www.ppinys.org/reports/medicd99.htm (18 March 2002); and conversations with CMS Northeast Consortium managers, 20 March 2002.
16. J. Holahan and S. Zuckerman, "Geographic Border Crossing: Implications for Volume Performance Standards," Cooperative Agreement no. 17-C-99473/3-01 (Washington: Urban Institute, 1991).
17. Ibid.
18. Fu Associates, Expenditure Flows Related to Interstate Migration for Health Care Services, Technical Appendix, Contract no. HCFA 500-92-0044 (Prepared for the CMS, 1993).
19. U.S. Bureau of the Census, "Population Estimates for the U.S., Regions, and States by Selected Age Groups and Sex, Annual Time Series," www.census.gov/population/estimates/state/st-98-09.txt (27 April 2000); authors’ tabulations of information for the noninstitutional population from the 1996 Medical Expenditure Panel Survey, meps.ahrq.gov/Data_Pub/HC_FYData96.htm#hc014 (8 April 2002); and D. Mendelson and W. Schwartz, "The Effects of Aging and Population Growth on Health Care Costs," Health Affairs (Spring 1993): 119–125.
20. G. Anderson and P.S. Hussey, "Comparing Health System Performance in OECD Countries," Health Affairs (May/June 2001): 219–232; and T. Getzen, "Health Care Is an Individual Necessity and a National Luxury: Applying Multilevel Decision Models to the Analysis of Health Care Expenditures," Journal of Health Economics 19, no. 2 (2000): 259–270.[Medline]
21. Getzen, "Health Care Is an Individual Necessity."
22. Ibid.
23. The federal share of state Medicaid expenditures is inversely related to per capita personal income in that state.
24. M. Gornick et al., "Effects of Race and Income on Mortality and Use of Services among Medicare Beneficiaries," New England Journal of Medicine 335, no. 11 (1996): 791–799[Abstract/Free Full Text]; R. Mayberry, F. Mili, and E. Ofili, "Racial and Ethnic Differences in Access to Medical Care," Medical Care Research and Review 57, Supp. 1 (2000): 108–145[Abstract/Free Full Text]; and R. Weinick, S. Zuvekas, and J. Cohen, "Racial and Ethnic Differences in Access and Use of Health Care Services, 1977 to 1996," Medical Care Research and Review 57, Supp. 1 (2000): 36–54.[Abstract/Free Full Text]
25. Dartmouth Medical School, Center for the Evaluative Clinical Sciences, Dartmouth Atlas of Health Care (1996), www.dartmouthatlas.org/pdffiles/ATLAS96.PDF (8 April 2002). Analysis of the supplier-induced-demand hypothesis is complicated by the fact that the relationship between physician concentration and per capita expenditures may be endogenous: Physicians may induce demand, or high prices may attract physicians. J. Rizzo and D. Blumenthal, "Is the Target Income Hypothesis an Economic Heresy?" Medical Care Research and Review 53, no. 3 (1996): 243–266.[Abstract/Free Full Text]
26. Physician-to-population ratios are based on data from the American Medical Association and the American Osteopathic Association and population estimates from the U.S. Bureau of the Census.
27. Dartmouth Medical School, Center for the Evaluative Clinical Sciences, The Dartmouth Atlas of Health Care 1999, www.dartmouthatlas.org/99US/chap_3_sec_1.php (8 April 2002).
28. L.C. Baker et al., "HMO Market Penetration and Costs of Employer-Sponsored Health Plans," Health Affairs (Sep/Oct 2000): 121–128.
29. Anderson and Hussey, "Comparing Health System Performance in OECD Countries."
30. S. Heffler et al., "Health Spending Projections for 2001–2011: The Latest Outlook," Health Affairs (Mar/Apr 2002): 207–218.

                      What's this?