|
Tracking Changes In Eligibility And Coverage Among Children, 1996–2002
Data from the 1996 Medical Expenditure Panel Survey (MEPS) reveal that 4.7 million children were eligible for Medicaid but were uninsured. Numerous changes have occurred in the landscape for children’s health insurance since then, including welfare reform and implementation of the State Children’s Health Insurance Program (SCHIP). We use data from the 1996–2002 MEPS to track changes in the eligibility and coverage of children. As of 2002, uninsurance among children remained as much a problem of participation as one of eligibility. Nevertheless, we find evidence of dramatic improvements in program participation, reflecting the success of efforts to improve outreach, simplify enrollment, and increase retention.
Between 1977 and 1987 the percentage of low-income children lacking health insurance climbed from 20.9 percent to 30.8 percent.1 Spurred in part by this decline in coverage, federal and state governments have worked together to increase the availability of free or heavily subsidized public health insurance coverage for children. The poverty-related Medicaid expansions beginning in the late 1980s conferred eligibility to millions of children in families ineligible for welfare, the traditional pathway for children to receive Medicaid coverage. By 1996 the expansions had helped reduce the rate of uninsurance among low-income children to 23.0 percent.2 The pace of reform intensified in the late 1990s. Poverty-related Medicaid eligibility continued to expand, and in 1998 states began implementing the State Children’s Health Insurance Program (SCHIP). By 2002, uninsurance among low-income children fell to 18.6 percent.3 Medicaid covered 21.7 million children at some point during 2000, and SCHIP enrollment during the first quarter of 2002 grew to 3.8 million children.4 Although expansions in eligibility increased public coverage for children and helped reverse the rise in uninsurance, not all eligible children enroll in public coverage. Research using the 1996 Medical Expenditure Panel Survey (MEPS) found that 4.7 million children were eligible for Medicaid but were uninsured.5 Approximately 40 percent of all uninsured children were eligible for public coverage in 1996, making uninsurance among children as much a problem of enrollment as one of eligibility.6 Since 1996 the economy has expanded and contracted, and private insurance premiums have outpaced the general rate of inflation. Both changes may have influenced trends in private and public coverage. Moreover, there have been several major changes in public policy. Welfare reform restricted immigrants’ eligibility for public coverage and may have deterred some eligible immigrant families from enrolling.7 Welfare reform also restricted eligibility for cash payments and severed the link between cash welfare and Medicaid. One consequence of this may have been reduced Medicaid enrollment, if fewer families applied for welfare or if applicants were not informed of their Medicaid eligibility.8 Concerns about Medicaid enrollment also led to concerns about enrollment in SCHIP.9 Many children made eligible through the Medicaid expansions had failed to enroll, and in the absence of substantial outreach efforts, SCHIP enrollment rates seemed likely to be lower still. One reason for this is that enrollment rates decline with age, and SCHIP-eligible children are older on average than Medicaid-eligible children. Also, SCHIP-eligible children typically have working parents who may be unaccustomed to applying for public benefits. Concerns about Medicaid and SCHIP enrollment have led to unprecedented efforts to improve outreach, reduce stigma, simplify enrollment, and retain eligible enrollees since 1996.10 SCHIP may also have had a beneficial spillover effect on Medicaid enrollment, because its legislation requires states to screen SCHIP applicants for Medicaid eligibility.11 Finally, expansions in family coverage under Medicaid may have increased enrollment among children.12 Given the many changes since 1996, it is useful to examine recent trends in eligibility and coverage among children. We rely on data from the 1996, 1998, 2000, and 2002 MEPS, combining a consistent data source with a consistent eligibility simulation methodology over time to estimate trends.
The data for our analysis come from MEPS, a stratified and clustered random sample of households sponsored by the Agency for Healthcare Research and Quality (AHRQ).13 When combined with sample weights, MEPS is designed to yield nationally representative estimates of insurance coverage, medical expenditures, and a wide range of other health-related and socioeconomic characteristics for the civilian, noninstitutionalized population. MEPS has an overlapping panel design, with data collected in five rounds over two and a half years. We focus on eligibility and enrollment of children age eighteen and under in the first part of each calendar year. The number of sampled children varies from a low of 7,446 in 1996 to a high of 13,050 in 2002, for a total of 36,729 observations. All results discussed in the text are statistically significant at the 5 percent level unless otherwise noted. All standard errors and statistical tests are adjusted for the complex design of MEPS. Insurance coverage definitions. We considered children to be covered if they had insurance covering physician and hospital care at any time during the round (typically four to five months). Thus, we classified children as uninsured only if they were continuously without coverage during the entire round. The 1–2 percent of all children who held both public and private coverage were classified as having public coverage. Income measurement. We measured earnings from all jobs held during the interview week. Our earnings measure thereby corresponds as closely as possible to the period during which we measured insurance coverage. To measure unearned income from interest, dividends, Social Security, and pensions, we linked MEPS data to the National Health Interview Survey (NHIS). This survey provides us with indicators for income receipt by type, which we used in constructing cold-deck imputations from MEPS full-year data.14 We obtained our estimate of assets by capitalizing interest and dividend income flows using the average return on six-month certificates of deposit.15 Eligibility simulation. Our eligibility simulation refines the approach previously used with the 1996 MEPS.16 We used data on age, earned and unearned income, marital status, employment status, family structure, and state of residence, combined with detailed program eligibility rules by state and by year.17 In particular, we applied detailed rules regarding income disregards, assistance unit composition, and asset tests. We also simulated eligibility according to immigration status. Legal immigrant children often face more stringent eligibility criteria than citizen children, and undocumented aliens are rarely eligible for public coverage. Linked NHIS data provide information on nativity, length of time in the United States, and citizenship.18 Among children in families under 200 percent of the federal poverty level in 2002, 4.4 percent of all children and 10.0 percent of uninsured children were ineligible for Medicaid and SCHIP solely based on their immigration status.19 We grouped children into three broad classes of eligibility. "Welfare-related" Medicaid eligibles include children in families eligible for welfare or (after welfare reform) Section 1931 family coverage. This group also includes children eligible through medically needy programs, the Ribicoff Children program, free Medicaid waiver programs targeting families, and separate state-funded programs providing similar coverage to immigrant families. "Poverty-related" Medicaid eligibles include children born after 30 September 1983 into families with net incomes below the federal poverty guidelines, children under age six in families below 133 percent of poverty, and those eligible through state expansions covering older children and those in families with higher incomes. This category also includes remaining children eligible for free coverage through Medicaid waivers and through separate state-funded programs to provide similar coverage to immigrant children. Finally, "SCHIP" includes children eligible for Medicaid SCHIP, separate state SCHIP, or separate state-funded programs to provide similar coverage to immigrant children. Also, some states expanded coverage before SCHIP implementation, providing SCHIP-like coverage to children later targeted by SCHIP. We therefore included in this group children eligible for public coverage via other non-SCHIP state programs or Medicaid waivers requiring (subsidized) premiums.
Insurance coverage by poverty status. As of 2002, 61 percent of all children held private coverage, 26 percent held public coverage, and 13 percent were uninsured (Exhibit 1  ). Among poor children (below 100 percent of poverty), the dominant form of coverage was public insurance. Among near-poor children (100–200 percent of poverty), private and public coverage played approximately equal roles. For middle- and higher-income children (above 200 percent of poverty), private insurance was the dominant form of coverage.
Uninsurance rates also varied by poverty level in 2002. Among poor children, nearly 18 percent were uninsured. The uninsurance rate among near-poor children was higher, although this difference is not statistically significant. Middle- and higher-income children, in comparison, had far lower rates of uninsurance. Private coverage among all children rose and fell between 1996 and 2002, mirroring the economic expansion and contraction over the period.20 By 2002, private coverage rates were slightly below 1996 levels (not a significant difference). In contrast, the percentage of children with public coverage rose five percentage points from 1997 to 2002, and the percentage without coverage fell steadily. The largest increase in public coverage was among near-poor children. This twelve-percentage-point increase reflects declines in both uninsurance and private coverage.21 In particular, private coverage among near-poor children declined by 7.9 percentage points between 1996 and 2002. This may reflect "crowd-out," whereby families dropped private coverage for their children when free or highly subsidized public coverage became available. Researchers have found evidence of this phenomenon in both the Medicaid and SCHIP expansions.22 Declines in private coverage may also reflect premium increases. Clearly, the trends we present here are not sufficient to identify separately these two effects on private coverage.
Eligibility for public coverage.
The number of children eligible for free or highly subsidized public coverage expanded from 21.4 million in 1996 to 36.0 million in 2002 (Exhibit 2
Uninsurance by eligibility status. As eligibility for public coverage expanded, so did the number of eligible uninsured children (Exhibit 3 ).23 By 2002, 2.8 million children age eighteen and under were uninsured but eligible for SCHIP. In contrast, the number of Medicaid-eligible but uninsured children declined by about a million from 1996 to 2002. This decline occurred mainly after 1998 and was concentrated among children eligible because of the Medicaid expansions.
Take-up rates. The take-up rate equals the number of children with public coverage divided by the number who were enrolled in public coverage or uninsured. Exhibit 4 reveals three key findings. First, although take-up among welfare-related, Medicaid-eligible children dipped slightly (but not significantly) in the wake of welfare reform, by 2002 take-up had returned to its 1996 level. If welfare reforms had the hypothesized negative effect on take-up, by 2002 that effect was offset by other factors such as improved outreach, simplified enrollment, and increased retention.
Second, and more remarkably, take-up rates among expansion-eligible children rose from 61 percent in 1996 to 68 percent in 1998 (the difference is only significant at the 10 percent level) and 77 percent in 2002. In 1996 Medicaid take-up rates among expansion-related eligibles were well below those among welfare-related eligibles. In a major reversal of this pattern, by 2002 take-up rates for these two groups had largely converged. Third, SCHIP take-up rates rose throughout SCHIP’s implementation. As expected, take-up rates were initially quite low. As of 2000, predictions that SCHIP take-up would be under 50 percent absent major improvements in outreach were largely correct.24 However, SCHIP take-up continued to rise, and by 2002 it had reached 60.4 percent.
As a final step, we used multivariate analysis to examine whether the trends in Exhibit 4
Limitations. Our study has three notable limitations. First, although we are careful to account for rules governing family composition, income net of disregards, assets, and immigration status, all of these variables may be measured with error in our data. Also, we do not simulate transitional or continuation eligibility, disability-related eligibility, or presumptive eligibility. For this reason, we may underestimate the number of eligible children, the number of eligible but uninsured children, and take-up rates. Second, we deem children to be uninsured only if they continuously lacked coverage during the round (typically four to five months). By ignoring shorter spells of uninsurance, our results may understate the true extent to which eligible children lack coverage. Third, an issue with all household surveys is the accuracy of coverage data. Comparisons with administrative data suggest that MEPS may modestly undercount enrollment in public coverage. Nevertheless, MEPS is widely regarded as providing more accurate and consistent public coverage estimates than the Current Population Survey (CPS), perhaps because MEPS asks numerous detailed questions regarding the presence, source, and duration of coverage.26 Also noteworthy is that MEPS and the CPS find higher levels of uninsurance than either the National Survey of America’s Families (NSAF) or the NHIS. Indeed, this difference has grown since the late 1990s, with NSAF and the NHIS showing faster declines in uninsurance and faster increases in public coverage than either MEPS or the CPS.27 Comparison of results. Given the potential for differences across surveys and eligibility simulations, it is useful to compare our results with other published estimates. Our estimates of uninsured eligible children in 2000 are close to those from the 1999 NSAF. We estimate that 6.4 million eligible children were uninsured throughout the round (4.0 million were eligible for Medicaid and 2.4 million for SCHIP). The 1999 NSAF estimate is that 6.8 million eligible children were uninsured as of the interview date (4.6 million were eligible for Medicaid and 2.2 million for SCHIP).28 Similarly, our public coverage take-up rate for 2000 is 68.8 percent, which is close to the 68 percent take-up rate from the 1999 NSAF.29 Our 2002 MEPS estimates are also close to a CPS-based study that found that 6.5 million children were eligible for Medicaid or SCHIP but were uninsured in 2002.30 In contrast, adjusting CPS data in an effort to correct for undercounted public coverage yields an estimate that fewer than five million children in 2000 were eligible for Medicaid or SCHIP but were uninsured.31 Clearly, no survey or eligibility simulation is free from potential errors, and estimates from any one survey or eligibility simulation should be interpreted with caution.
Between 1996 and 2002 the proportion of all children who were eligible for free or heavily subsidized public health insurance coverage rose from 28.6 percent to 47.1 percent. Not surprisingly, the number of eligible uninsured children rose as well, from 4.6 million in 1996 to 6.2 million in 2002. As of 2002, more than 60 percent of all uninsured children were eligible for public coverage. What was true in 1996 was even more true in 2002: Participation remains a key problem among uninsured children. Children’s enrollment in public coverage remains far from complete, yet dramatic progress has been made. First, we observe no net decline in take-up among welfare-related eligible children. Early studies found evidence of declines in Medicaid participation in the wake of immigration and welfare reform.32 Our results for 1996 to 1998 mirror those findings. However, if immigration and welfare reforms had the hypothesized negative effect on take-up, by 2002 those effects were offset by state efforts to improve outreach, simplify enrollment, and increase retention. Second, we find large increases in take-up among Medicaid expansion–eligible children, from the low levels observed in 1996 to levels in 2002 that were close to those for welfare-related eligibles.33 Third, we observe remarkable increases in take-up among SCHIP-eligible children. These findings are confirmed by our multivariate analysis, in which we controlled for a wide range of socioeconomic characteristics of children and their families. We interpret our results as evidence of the effects of improved outreach, reduced stigma, enrollment simplification, continuous coverage, and the myriad other improvements in Medicaid and SCHIP implemented since the mid-1990s.34 In many states, reform efforts focused on SCHIP, but we observe equally large improvements in the take-up rates of both Medicaid and SCHIP. This suggests that these programs may have important spillover effects on each other. It may be overly optimistic to anticipate continued improvements in take-up rates beyond the period of our study. Recently, states have faced intense fiscal pressures. These pressures, combined with rapidly rising enrollment in public coverage, have led many states to begin rolling back SCHIP eligibility or the outreach programs and simplifications that are likely to have increased take-up rates.35 Whether these gains can be preserved in the present fiscal climate is an important open question that we hope will be answered by future waves of MEPS.
Thomas Selden (tselden{at}ahrq.gov) and Julie Hudson are economists, and Jessica Banthin is director, Division of Modeling and Simulation, Center for Financing, Access, and Cost Trends, at the Agency for Healthcare Research and Quality in Rockville, Maryland. The authors are grateful for the helpful suggestions of Cindy Brach, Randolph Capps, Lisa Dubay, Trena Ezzati-Rice, Sarah Grantham, Carol Irvin, Genevieve Kenney, Julie Martinez, John Moeller, and John Sommers, along with the expert programming assistance of Devi Katikineni and Jackie Malone of Social and Scientific Systems. They also thank Leighton Ku, Randolph Capps, Jeffrey Passel, Kenneth Sucher, and Carol Irvin for generously sharing data. Any remaining errors are the authors’ own. The views expressed in this paper are those of the authors, and no official endorsement by the Department of Health and Human Services or the Agency for Healthcare Research and Quality is intended or should be inferred.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Data And Methods
Study Results
