Health Affairs, 23, no. 5 (2004): 27-38 doi: 10.1377/hlthaff.23.5.27 © 2004 by Project HOPE	New Online   Obama's Success & Next Steps   3 Health Policy Options for President Obama   A Hybrid Proposal   Bipartisan Compromise

This Article Abstract Figures Only 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 ISI Web of Science 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 (7) Citing Articles via Google Scholar Google Scholar Articles by Cunningham, P. Articles by Kirby, J. Search for Related Content PubMed PubMed Citation Articles by Cunningham, P. Articles by Kirby, J. Related Collections Access To Care Health Reform Insurance Coverage - Children Maternal And Child Health Medicaid

Children's Coverage

Children’s Health Coverage: A Quarter-Century Of Change

Abstract

 
Data from a series of medical spending surveys over the past twenty-five years show that uninsurance rates for children have fallen to levels not seen since the late 1970s (about 8 percent). Rates of uninsurance have fallen in particular for poor children. Deteriorating family economic circumstances, along with a weakened safety net, contributed to the decrease of private insurance coverage and rising uninsurance rates between 1977 and 1987. Although family circumstances have stabilized since the late 1980s, high annual increases in health insurance costs have continued to erode private coverage. Public coverage expansions have reversed the surge in uninsurance rates during 1977–1987.

Public coverage expansions were necessitated by increases in the number and percentage of children without health insurance coverage—especially poor and low-income children during the early 1980s—as well as a long, steady erosion of private insurance coverage over the past twenty-five years. Earlier analyses attributed the decline in private insurance and increase in children’s uninsurance rates between 1977 and 1987 to changing social and economic circumstances of children, specifically stagnating incomes, the increase in child poverty, and the increase in single-parent families.¹ Since that time, however, many of these key correlates of children’s health insurance coverage have stabilized and even improved, as exemplified by the sharp drop in poverty among children during the late 1990s.² Nevertheless, private insurance coverage continues to erode, while public programs have greatly expanded, thereby offsetting the erosion in private insurance coverage and contributing to decreases in children’s uninsurance rates.³

The main objective of this paper is to describe trends in children’s health insurance coverage over the past quarter-century. Such an analysis puts the more recent gains in children’s coverage in a broader historical context and cumulatively shows how changes in key characteristics of children, public policy, and affordability of private insurance have all contributed to changes in coverage.

Study Data And Methods

Top Study Data And Methods Study Findings Impact Of Changes In... Discussion And Policy... Editor's Notes NOTES

 
Data. The data for this study come from three national household surveys sponsored by the Agency for Healthcare Research and Quality (AHRQ): the 1977 National Medical Care Expenditure Survey (NMCES), the 1987 National Medical Expenditure Survey (NMES), and the 1997 and 2001 Medical Expenditure Panel Survey (MEPS). Each of these surveys collects information that provides nationally representative estimates of health care use, insurance coverage, and sociodemographic characteristics for the U.S. civilian, noninstitutionalized population.⁴ As such, they are the only data sources capable of providing consistent estimates of children’s health insurance coverage over the past quarter-century.⁵ Our analysis is restricted to children under age eighteen.

Analysis. Our main analytic variable is a three-category, hierarchical variable indicating the type of insurance coverage a child had during the reference years. Children were coded as having private insurance if they had coverage from a private source at any time during the year. These children were further classified as having employer-sponsored health insurance if their parents obtained coverage from their jobs. Children who had coverage through a public plan only were coded as having public insurance. Finally, children who had neither private nor public health insurance at any time during the year were coded as being uninsured.⁶

Our analysis is primarily descriptive, with all estimates weighted to be nationally representative and standard errors adjusted for clustering in the sample. Unless otherwise noted, all differences discussed in the text are significant at the .05 level or better.

Study Findings

Top Study Data And Methods Study Findings Impact Of Changes In... Discussion And Policy... Editor's Notes NOTES

 
Trends in children’s coverage, 1977–2001. Exhibit 1 illustrates the trend in the percentage of children who had no health insurance coverage between 1977 and 2001. Regardless of how they are measured, uninsurance rates increased sharply between 1977 and 1987, before steadily declining back to near their 1977 levels by 2001.

View larger version (18K): [in this window] [in a new window]   EXHIBIT 1 Percentage Of Children Who Were Uninsured, Selected Years 1977–2001

Initially, much of the decrease in private insurance coverage was attributable to increased poverty among children. Between 1977 and 1987 the percentage of children in families with incomes below poverty increased from 13.7 percent in 1977 to 20.4 percent in 1987, because of a severe economic recession during the late 1970s and early 1980s, high unemployment, stagnating incomes, and a rise in single-parent families. Since 1987, however, changes in family income or childhood poverty were probably less of a factor in the decline of private insurance. Between 1987 and 1997 private coverage rates continued to fall, despite little change in child poverty rates. And while child poverty rates fell to 16.3 percent by 2001—because of strong economic growth and low unemployment during this period—rates of private insurance coverage did not change significantly.

Decreases in private insurance coverage since 1987 were likely related to the high annual increases in private health insurance premiums from the 1980s through the mid-1990s.⁷ In addition, employees assumed a much higher share of the premium for family coverage during the 1980s and 1990s, which likely contributed to a decrease in take-up rates among people who were offered coverage at work.⁸ While structural economic changes, such as the shift from manufacturing to service-sector jobs, have played some role in private coverage erosion, research has shown that increasing health care costs were the dominant factor.⁹

Public coverage expansions. Despite the large increase in child poverty rates between 1977 and 1987, public coverage did not immediately offset the decline in private insurance coverage. In fact, the 1981 Omnibus Budget Reconciliation Act (OBRA 1981) reduced Medicaid coverage for many poor children by limiting the allowable earned income for determining eligibility for Aid to Families with Dependent Children (AFDC). This led to a large decline in the percentage of poor children with public coverage—from 52 percent in 1977 to 39 percent in 1987—while uninsurance rates for poor children rose from 16 percent to 24 percent (Exhibit 2).¹⁰

Since 1987, increases in public coverage have more than offset private coverage declines and have also contributed to declining uninsurance rates for low-income children. A series of Medicaid expansions between 1984 and 1990 extended eligibility to many poor and near-poor children and began to sever the previously strong link between AFDC and Medicaid eligibility (which restricted eligibility mostly to children in single-parent, nonworking families with incomes well below poverty).¹¹ As a result, the percentage of poor children with public coverage rose dramatically, while uninsurance rates among poor children declined.

In the meantime, uninsurance rates continued to rise for other low-income children (those with family incomes between 100 and 200 percent of poverty), primarily because of decreases in private insurance coverage. But unlike poor children, many low-income children above the poverty level were not eligible for the Medicaid expansions of the 1980s and 1990s. As a result, uninsurance rates for this income group rose to 19.5 percent by 1997 (from 13 percent in 1977) and had become considerably higher than for poor children (11 percent), many of whom had already benefited from the earlier Medicaid expansions.

SCHIP was enacted in 1997 to address the coverage gap for low-income children who were not poor enough to qualify for Medicaid. As a result, the percentage of publicly insured children in the 100–200 percent of poverty group increased from 1997 to 2001, while uninsurance rates declined.¹² Public coverage expansions have also resulted in some substitution of public for private insurance coverage, whereby some children with private insurance or access to private insurance enrolled in free or lower-cost public coverage. Thus, not all of the large increases in public coverage for poor and low-income children came from the ranks of the uninsured. Other research confirms that both Medicaid and SCHIP expansions led to some substitution of public for private insurance, although estimates of the extent of this substitution vary greatly.¹³

Changes in the characteristics of children and families. Dramatic changes in family composition and the workforce participation of women also contributed to changes in children’s health insurance coverage during the past quarter-century.¹⁴ As a result of rising divorce rates and out-of-wedlock births, the percentage of children living in single-parent families increased from about 14 percent in 1977 to 27 percent in 1987 (Exhibit 3). Children in single-parent families (most of which are headed by mothers, whose workforce participation and earning potential have typically been lower than those of fathers) are much more likely than children in two-parent families to be poor or low-income, much less likely to have private insurance coverage, and more likely to have public coverage.¹⁵

But while the change in family composition was an important factor in the initial surge in uninsurance rates, it has likely been a much less prominent factor in coverage changes since 1987. After 1987, the percentage of children in single-parent families actually decreased slightly, and the Medicaid and SCHIP expansions of the late 1980s and 1990s reversed the earlier decreases in public coverage.

The increased workforce participation of women has also helped to offset some of the decreases in private coverage and the increase in uninsurance rates and is also probably related to the decrease in child poverty since 1987. Among children in single-parent families (the vast majority of which are mother-only), the percentage with an employed parent rose from 65.1 percent in 1977 to 85.2 percent in 2001, with most of the increase coming after 1987. Although rates of private coverage among children in single-parent families (including those where the parent is employed) have been falling since 1987, the decline would have been much greater if not for the rise in employment among single parents.¹⁶

Mothers’ increased labor-force participation raised the percentage of children in two-parent families with both parents working between 1977 and 2001. Having two workers in the family increases the chances that at least one parent will be offered employer coverage, and the generally higher incomes of two-worker families enable them to afford the rising out-of-pocket premiums for family coverage.

Private coverage has declined much more rapidly for children in two-parent families with a single worker than for children in two-worker families, although public coverage expansions have minimized the impact on uninsurance rates for both groups of children. But the growing divergence between two-worker and single-worker families regarding coverage type indicates that the number of workers in the family—rather than the number of parents per se—is becoming more important in determining the type of coverage children have.

New concerns: uninsurance rates among Hispanic children. Other demographic changes have also affected the prevalence of insurance coverage among children during the study period. In particular, the increase in immigration from Latin America combined with a higher birth rate among Hispanics has altered the ethnic composition of children in the United States.¹⁷ Consequently, the percentage of children who are Hispanic has more than doubled during the past quarter-century, making Hispanics the largest U.S. minority group (Exhibit 4). Moreover, Hispanics have the highest and fastest-rising uninsurance rates among all of the major racial/ethnic groups. High uninsurance rates among Hispanic children reflect lower incomes and lack of access to employer-sponsored coverage, in large part because their parents are more likely to be in the types of low-wage, low-skill jobs that typically do not offer generous health benefits.¹⁸ Eligibility barriers for public programs—such as restrictions for many noncitizens and most undocumented immigrants, as well as fear of jeopardizing immigration status—prevent greater enrollment in public programs.¹⁹

View larger version (17K): [in this window] [in a new window]   EXHIBIT 4 Trends In The Health Insurance Coverage Of Hispanics, Selected Years 1977–2001

Impact Of Changes In Children’s Characteristics On Coverage Changes

Top Study Data And Methods Study Findings Impact Of Changes In... Discussion And Policy... Editor's Notes NOTES

 
To estimate the contribution of changes in children’s characteristics to overall changes in health insurance coverage between 1977 and 2001, we simulated coverage levels for 1987, 1997, and 2001, assuming no changes in children’s characteristics since 1977 (Exhibit 5).²⁰ If the simulated and actual coverage changes are similar in magnitude, then one can conclude that changes in children’s characteristics have minimal effects on coverage trends. The larger the divergence between the actual and simulated coverage changes, then the larger the effect of changes in children’s characteristics on coverage trends.

Conversely, public coverage would have decreased between 1977 and 1987 by more than three percentage points (instead of increasing slightly) if there had been no change in children’s characteristics. That is, the 1981 Medicaid cuts likely would have reduced public coverage overall (as it did for poor children) if not for the increase in the number of children who became eligible because of higher poverty, more single-parent families, and higher parental unemployment.

Nevertheless, the effect of changes in children’s characteristics on uninsurance rates between 1977 and 1987 was fairly minimal—uninsurance would have risen by between two and three percentage points, regardless. Thus, the surge in children’s uninsurance between 1977 and 1987 was likely attributable to a sharp increase in child poverty (with the increase in single-parent families as a contributor), along with a weakened safety net that failed to offset declining private coverage.

Over time, the influence of changes in children’s family circumstances since 1977 on private and public coverage trends has diminished, as indicated by the growing convergence between the actual and simulated changes for the periods 1977–1997 and 1977–2001. Had family characteristics not changed since 1977, private coverage would still have decreased by about six percentage points between 1977 and 1997 (compared to the actual decrease of 9.4 percentage points). By 2001 the decrease in private coverage would have been virtually the same regardless of the changes in family circumstances since 1977 (ten to eleven percentage points). Similarly, the effects of changes in family characteristics on public coverage trends have also greatly diminished over time.

The decreasing influence of family characteristics on private and public coverage trends reflects several factors. First, the decline in family income and two-parent families was halted and even reversed after 1987.²¹ Second, the escalating costs of private health insurance likely became the most important factor responsible for the continued erosion in private coverage. Finally, public coverage expansions extended public coverage eligibility to increasingly large numbers of children who did not meet the traditional welfare standards of eligibility based on extreme poverty, family composition, and parental employment.

As the influence of family characteristics on coverage trends has diminished over time, the influence of the increase in the Hispanic population has grown, although the effects on coverage trends are still small. By 2001 the increase in the Hispanic population since 1977 accounted for about two points of the 10.8-percentage-point decrease in private coverage. The impact on public coverage trends was even smaller, with the result that uninsurance rates were higher in 2001 by about one percentage point than they would have been if the proportion of Hispanics was the same as in 1977.

Discussion And Policy Implications

Top Study Data And Methods Study Findings Impact Of Changes In... Discussion And Policy... Editor's Notes NOTES

 
Since 2001 the outlook for children’s health insurance coverage has changed markedly and has some similarities with the early 1980s. The return of high annual increases in health insurance costs, rising unemployment, and increased child poverty contributed to decreases in private insurance coverage between 2000 and 2002, reversing a brief period of modest private coverage increases during the late 1990s.²² Of perhaps even greater concern is that state budget pressures have led virtually all states to make cuts in Medicaid or SCHIP or both, which raises the prospect that, as in the early 1980s, the safety net is weakening at a time when it is needed the most.

So far there is little indication that coverage rates for children have declined since 2001. Despite the recent cuts, eligibility levels are vastly higher than they were twenty years ago and now include almost all children with family incomes below 200 percent of poverty. Most states have avoided making deep, permanent cuts in eligibility levels, in part because these programs receive strong support from the general public, policymakers, and health care providers.²³

But while the safety net may remain largely intact for lower-income children, there is growing concern about coverage for many moderate- and middle-income children (most of whom fall outside the eligibility levels of Medicaid and SCHIP). Despite high rates of private insurance coverage, the return of annual double-digit increases in private health insurance premiums starting in 2000 will make coverage more difficult for many employers, workers, and families to afford.²⁴ As labor markets become less tight than they were in the late 1990s, employers may pass on much of these higher costs to their workers in the form of either higher premiums (especially for family coverage) or less generous coverage (higher copayments and deductibles). If so, it is likely that fewer families (especially low-income families) will elect to enroll in family coverage. With no safety net to fall back on, many moderate-and middle-income children may lose coverage entirely.

Given federal and state budget realities, expanding fully subsidized public coverage to middle-income children may be difficult. Some of these families may be able to take advantage of the proposed tax credits and deductions for the purchase of nongroup coverage or catastrophic plans (along with the recently passed health savings accounts), although it is unclear whether the net out-of-pocket payment for these new types of coverage will be lower than those of the employer coverage they turn down. Since a high proportion of middle-income children have access to employer coverage, increasing the public subsidy for such coverage may help to avoid any erosion in private insurance by offsetting higher premium costs, although such an approach could still require sizable new public expenditures. The unclear direction of policy, the ongoing escalation of health insurance costs, and federal and state budget pressures will continue to make children’s health insurance coverage an important issue that requires regular monitoring.

Editor's Notes

Top Study Data And Methods Study Findings Impact Of Changes In... Discussion And Policy... Editor's Notes NOTES

 
Peter Cunningham (pcunningham{at}hschange.org) is a senior health researcher at the Center for Studying Health System Change in Washington, D.C. Jim Kirby is a service fellow at the Agency for Healthcare Research and Quality in Rockville, Maryland.

The views in this paper are those of the authors, and no official endorsement by the Agency for Healthcare Research and Quality or the Department of Health and Human Services is intended or should be inferred. The Center for Studying Health System Change is supported in full by the Robert Wood Johnson Foundation.

NOTES

Top Study Data And Methods Study Findings Impact Of Changes In... Discussion And Policy... Editor's Notes NOTES

 

1. P.J. Cunningham and A.C. Monheit, "Insuring the Children: A Decade of Change," Health Affairs 9, no. 4 (1990): 78–90[CrossRef][Medline]; and R.M. Weinick and A.C. Monheit, "Children’s Health Insurance Coverage and Family Structure, 1977–1996," Medical Care Research and Review 56, no. 1 (1999): 55–73.[Abstract]
2. U.S. Census Bureau, Current Population Survey, Detailed Poverty Tables 1994–2000, www.census.gov/hhes/poverty/prevdetailtabs.html (7 June 2004).
3. P.W. Newacheck et al., "Trends in Private and Public Health Insurance for Adolescents," Journal of the American Medical Association 291, no. 10 (2004): 1231–1237[Abstract/Free Full Text]; J.S. Banthin and T.M. Selden, "The ABCs of Children’s Health Care: How Medicaid Expansions Affected Access, Burdens, and Coverage between 1987 and 1996," Inquiry 40, no. 2 (2003): 133–137[ISI][Medline]; and P.J. Cunningham, J.D. Reschovsky, and J. Hadley, "SCHIP, Medicaid Expansions Lead to Shifts in Children’s Coverage," Issue Brief no. 59 (Washington: Center for Studying Health System Change, December 2002).
4. G.S. Bonham and L.S. Corder, "National Health Care Expenditure Study, Instruments and Procedures 1: NMCES Household Interview Instruments," Pub. no. (PHS) 81-3280 (Rockville, Md.: U.S. Department of Health and Human Services, 1981); W. Edwards and M. Berlin, "Questionnaires and Data Collection Methods for the Household Survey and the Survey of American Indians and Alaska Natives," National Medical Expenditure Survey Methods 2 (Rockville, Md.: Agency for Healthcare Research and Quality, 1989); and J.W. Cohen et al., "The Medical Expenditure Panel Survey: A National Health Information Resource," Inquiry 33, no. 4 (1996): 373–389.
5. Despite some differences in the content of the three surveys, questions on insurance coverage are comparable (that is, asked in the same way and in the same order across the three surveys). The vast majority of interviews in all three surveys were conducted in person. Recall periods are somewhat longer in MEPS because there are fewer rounds of data collection during the survey year, although the impact on estimates of coverage is likely to be minimal. Trends in coverage for various periods are generally similar to those reported in other surveys, such as the Current Population Survey and the National Health Interview Survey. See, for example, P. Fronstin, "Sources of Health Insurance and Characteristics of the Uninsured: Analysis of the March 2003 Current Population Survey," EBRI Issue Brief no. 264 (Washington: Employee Benefit Research Institute, 2003); and National Center for Health Statistics, Health, United States, 2003 (Hyattsville, Md.: NCHS, 2003), Tables 127–129.
6. This definition of "insurance coverage" differs from two previous studies of trends in children’s health insurance coverage using the same surveys for 1977–1996. These earlier studies measured insurance coverage based on the first part of the year. As such, estimates of children’s uninsurance rates (reflecting being uninsured throughout the year) are lower in this study than in the earlier studies. See Cunningham and Monheit, "Insuring the Children: A Decade of Change"; and Weinick and Monheit, "Children’s Health Insurance Coverage." Children who were uninsured only part of the year are classified as "insured" in this study, although these children may also experience problems getting access to medical care. While estimates of the number and percentage of children who are uninsured differ depending on whether uninsurance is measured as all year, part year, or point in time, trends in uninsurance rates over the study period are very similar regardless of the definition used (see Exhibit 1).
7. J. Gabel et al., "Health Benefits in 2003: Premiums Reach Thirteen-Year High as Employers Adopt New Forms of Cost Sharing," Health Affairs 22, no. 5 (2003): 117–126.[Abstract/Free Full Text]
8. C. Pemberton and D. Holmes, eds., EBRI Databook on Employee Benefits (Washington: EBRI, 1995); and P.F. Cooper and B. Schone, "More Offers, Fewer Takers for Employment-Sponsored Health Insurance: 1987 and 1996," Health Affairs 16, no. 6 (1997): 142–149.[Medline]
9. G. Acs, "Explaining Trends in Health Insurance Coverage between 1988 and 1991," Inquiry 32, no. 1 (1995): 102–110[ISI][Medline]; S.H. Long and J. Rodgers, "Do Shifts toward Service Industries, Part-time Work, and Self-Employment Explain the Rising Uninsured Rate?" Inquiry 32, no. 1 (1995): 111–116[ISI][Medline]; and R. Kronick and T. Gilmer, "Explaining the Decline in Health Insurance Coverage, 1979–1995," Health Affairs 18, no. 2 (1999): 30–47.[Abstract]
10. Although the 1981 Medicaid cuts were reversed and some expansions occurred before 1987, the effects on coverage were probably not immediate and therefore are not fully reflected in coverage rates until after 1987.
11. For a detailed description of the legislative changes affecting children’s health insurance coverage since 1965, see C. Mann, D. Rowland, and R. Garfield, "Historical Overview of Children’s Health Care Coverage," Future of Children 13, no. 1 (2003): 31–54.[Medline]
12. The decrease in uninsurance rates was statistically significant at the .10 level.
13. L. Dubay, Expansions in Public Health Insurance and Crowd-out: What the Evidence Says (Washington: Henry J. Kaiser Family Foundation, 1999); P.J. Cunningham, J. Hadley, and J. Reschovsky, "The Effects of SCHIP on Children’s Health Insurance Coverage: Early Evidence from the Community Tracking Study," Medical Care Research and Review 59, no. 4 (2002): 359–383[Abstract/Free Full Text]; and A.T. LoSasso and T.C. Buchmueller, "The Effect of the State Children’s Health Insurance Program on Health Insurance Coverage," NBER Working Paper no. 9405 (Cambridge, Mass.: National Bureau of Economic Research, 2002).
14. Cunningham and Monheit, "Insuring the Children"; and Weinick and Monheit, "Children’s Health Insurance Coverage."
15. P.W. Newacheck et al., "Adolescent Health Insurance Coverage: Recent Changes and Access to Care," Pediatrics 104, no. 2, Part 1 (1999): 195–202.[Abstract/Free Full Text]
16. If employment rates among single parents had been the same in 2001 as they were in 1987, private insurance coverage for all children in single-parent families would have decreased by about thirteen percentage points between 1987 and 2001, instead of the five-percentage-point decline that actually occurred.
17. NCHS, Health, United States, 2003.
18. C.L. Schur and J. Feldman, Running in Place: How Job Characteristics, Immigrant Status, and Family Structure Keep Hispanics Uninsured (New York: Commonwealth Fund, 2001); and Newacheck et al., "Trends in Private and Public Health Insurance for Adolescents."
19. G. Lessard and L. Ku, "Gaps in Coverage for Children in Immigrant Families," Future of Children 13, no. 1 (2003): 101–115.[Medline]
20. The simulation is based on ordinary least squares (OLS) regression analysis using the survey data for 1987, 1997, and 2001 (separate regressions for each year). Separate regressions were run for private insurance, public coverage, and uninsured. Independent variables included family income (relative to the federal poverty level), family structure (one parent versus two parents), parental employment status (two workers, one worker, no worker), Hispanic ethnicity, and age. Coverage rates for 1987, 1997, and 2001 were simulated using the regression coefficients for each year and the 1977 independent variable means. Simulations were computed holding all factors constant to their 1977 means, as well as separately for individual factors. Although separate variables in the regression, the results for family type, family income, and parent’s employment status are combined in Exhibit 6 because of the close correlations between these factors.
21. In fact, when 2001 coverage is simulated based on the 1987 mean for family income, the results show that private insurance coverage would have declined by an additional three percentage points between 1987 and 2001 if child poverty rates had not declined.
22. J. Holahan and M. Wang, "Changes in Health Insurance Coverage during the Economic Downturn: 2000–2002," Health Affairs, 28 January 2004, content.healthaffairs.org/cgi/content/abstract/hlthaff.w4.31 (7 June 2004).
23. J.F. Hoadley, P. Cunningham, and M. McHugh, "Popular Medicaid Programs Do Battle with State Budget Pressures: Perspectives from Twelve States," Health Affairs 23, no. 2 (2004): 143–154.[Abstract/Free Full Text]
24. B.C. Strunk, P.B. Ginsburg, and J.R. Gabel, "Tracking Health Care Costs: Growth Accelerates Again in 2001," Health Affairs, 25 September 2004, content.healthaffairs.org/cgi/content/abstract/hlthaff.w2.299 (7 June 2004).

                      What's this?