|
SCHIP’s Impact In Three States: How Do The Most Vulnerable Children Fare?
This study provides consistent evidence, from three very diverse states with heterogeneous populations and distinct programs (Florida, Kansas, and New York), that the State Children’s Health Insurance Program (SCHIP) increased access to and satisfaction with health care among enrolled low-income children and that vulnerable children—minorities, children and adolescents with special health care needs, and children who were uninsured for long periods of time—shared in these improvements. We highlight some areas to target for future improvement, such as reducing the high levels of unmet needs among special-needs children and increasing preventive care, especially for Hispanic children.
Congress enacted the State Children’s Health Insurance Program (SCHIP) in 1997 to insure children from low-income families that earned too much to be eligible for Medicaid and did not have access to affordable dependent coverage from their employers. Providing insurance to low-income children has led to measurable improvements in access, use, and some quality measures in state-funded insurance programs that predate SCHIP.1 Evidence has since accumulated that SCHIP improves access to health care, but the question of whether all children benefit equally from SCHIP has not yet been answered.2 Information about SCHIP’s impact on the most vulnerable enrollees—minority children, children and adolescents with special health care needs, and children who were uninsured for long periods of time—is critical for states to ensure equitable, efficient, and effective use of program resources. This study looks at how some particularly vulnerable children fared under SCHIP in three states—Florida, Kansas, and New York. The three subgroups of children were selected for study because all three groups experience poor access to health care and are prevalent in SCHIP.3 Minority children have experienced poorer access to health care independent of their insurance status.4 Special-needs children and adolescents also have problems getting services, and some have expressed concern that such children might have difficulty obtaining needed services in separate freestanding SCHIP programs that rely on commercial insurers.5 The long-term uninsured are also at risk if past difficulties getting health care reflect nonfinancial barriers to care that are not resolved by the provision of insurance.6 Although providing SCHIP eliminates financial barriers and is likely to improve access, it is unclear whether long-term uninsured children will be able to take as much advantage of the coverage as children who have more experience with insurance. Concern about these vulnerable populations drives this study’s research questions: Does SCHIP increase access to and satisfaction with health care for the enrolled population? Does it increase access to and satisfaction with health care for the three vulnerable subgroups studied here? And is SCHIP’s impact on access to and ratings of care similar for vulnerable subgroups and other enrollees? This study builds upon recent studies that investigated these questions in New York.7 It expands that work to Florida and Kansas, to determine if similar patterns in access to and ratings of care occurred across a variety of states and SCHIP programs.
Setting. This study uses longitudinal data from evaluations of separate free-standing SCHIP programs in Kansas, New York, and Florida that are part of the Child Health Insurance Research Initiative (CHIRI).8 New York and Florida have large, mature programs that had prototype programs in place before SCHIP was enacted. Together these two states accounted for 25 percent of SCHIP enrollees in 2001.9 Kansas has a smaller program that was established as a result of the SCHIP legislation. Each state offers a distinct set of features and environments under which the programs operate (Exhibit 1  ).10
Study design. Details about the data and methods of the three state studies are described elsewhere.11 All used a pre-post (T1–T2) longitudinal design in which samples were drawn from newly enrolled children between July 2000 and March 2001. Interviews were conducted shortly after enrollment about the twelve months prior to enrollment (T1), and again thirteen months after enrollment about the twelve months following enrollment (T2). The Kansas and New York studies drew stratified random samples of children ages 1–17. Florida drew a random sample of adolescents only (ages 12–17). Survey response. Kansas had 434 completed T1 and T2 surveys, for an overall response rate of 35 percent; Florida had 944 completed T1 and T2 surveys, for a combined response rate of 30 percent; and New York had 2,290 completed T1 and T2 surveys, for a combined response rate of 55 percent. We performed multivariate analyses to investigate whether T1 nonresponse was related to observable characteristics in the administrative data and whether T2 nonresponse was related to survey responses in T1.12 We adjusted our weights to account for nonresponse.
Measures.
As part of CHIRI, the three studies included a set of common core measures (Exhibit 2
Analyses. Only children for whom interviews were completed in both periods were included in the analyses. To account for differences in T2 interviews across the three studies, analyses in New York and Florida were further limited to children and adolescents who remained enrolled in SCHIP for at least twelve months. We were unable to do the same in Kansas because of the small sample size. The Kansas results, therefore, were estimated using samples that included children who disenrolled before twelve months elapsed. Additional analyses of the New York data were conducted on the sample of adolescents for comparison to Florida’s results. Sample-size limitations prevented adolescent-only analyses of the Kansas data. Analyses were performed to assess SCHIP’s impact on each of the three groups of vulnerable children and adolescents and the differences between vulnerable and other SCHIP enrollees both before and after SCHIP. We used multivariate population-average, random-effect logistic regressions for usual source of care, unmet need, and use of preventive care; and population-average, random-effect regression models to generate "adjusted" rates of care that eliminate the effects of differences in the characteristics of the population subgroups, thereby allowing for "fair" comparisons within states. We used standard methods to generate "adjusted" rates from model estimates.13 We did not pool data across states, however, so adjusted rates are based on each state’s characteristics (for example, the Kansas adjusted rates are influenced more by nonmetropolitan children than either the Florida or New York rates are). We calculated bootstrap standard errors of the differences in the adjusted rates before and after SCHIP to perform statistical tests. Tests of the equivalence of the adjusted outcomes across vulnerable subgroups were performed using Wald tests of the appropriate model restrictions.
The model specifications included all independent variables described above and summarized in Exhibit 2
A sizable proportion of SCHIP enrollees were black or Hispanic, had special health care needs, or were uninsured, or some combination, during the entire year prior to enrollment (Exhibit 2  ).
Access and satisfaction.
Overall, SCHIP improved children’s and adolescents’ access to and rating of health care (Exhibit 3
Improvements among minority children. Minority children and adolescents saw improvements on many measures. For some measures, results were relatively constant across the three states. For example, black children and adolescents were more likely to have had a preventive visit after SCHIP than before, but Hispanic children and adolescents were not. For other measures, SCHIP’s impact varied. The unmet needs of minority children and adolescents fell in Kansas and New York but not in Florida. SCHIP’s impact on minority children was greatest in New York, where black or Hispanic children, or both, improved on all four measures. Special-needs children. In all three states, special-needs children and adolescents experienced large improvements in ratings of care. The proportion with a usual source of care increased, and the proportion with unmet needs decreased in Kansas and New York. Although nearly all special-needs children and adolescents in Kansas and New York had a usual source of care after SCHIP enrollment, almost one-third still had unmet needs. Only in Florida did these children and adolescents experience statistically significant increases in the probability of having a preventive visit. Uninsured children and adolescents. Children and adolescents in Florida and New York who had been uninsured for more than one year had strong and statistically significant improvements in all four measures. In Kansas such children experienced a 62 percent reduction in unmet needs, but none of the increases in usual source of care, preventive care, or rating of health care were statistically significant. Comparisons of SCHIP’s impact. We compared improvements for vulnerable children and adolescents with other SCHIP enrollees and examined whether there were disparities between the two groups after SCHIP enrollment.
Minority versus white children.
Gains in access and satisfaction measures for minority children and adolescents were, for the most part, similar to the gains among white children and adolescents (Exhibit 4
Special-needs enrollees and their counterparts. SCHIP’s impact on special-needs enrollees was variable, depending on the measure and the state (Exhibit 5 ). There were few disparities after SCHIP enrollment between special-needs enrollees and their counterparts without special needs. Also, special-needs children and adolescents had greater unmet needs after SCHIP enrollment than other SCHIP enrollees had (except in Florida) but were better off than other enrollees on other measures (such as usual source of care in Kansas and preventive care in Florida). Disparities between special-needs and other adolescents in rating of health care that had existed before SCHIP enrollment were eliminated after enrollment.
Long-term uninsured versus previously insured. Enrollees who had been among the long-term uninsured experienced large gains after SCHIP enrollment (Exhibit 6 ). Children and adolescents who had at least some insurance during the year before they enrolled in SCHIP maintained the access levels they had prior to enrollment. Although disparities between long-term uninsured and insured children and adolescents were relatively common in the period before SCHIP enrollment, in every instance those disparities were eliminated by greater improvements among the long-term uninsured.
This study provides consistent evidence, from three highly diverse states with heterogeneous populations and distinct programs, that many vulnerable children had improved access to and satisfaction with health care after enrolling in SCHIP, improvements on par with those experienced by other enrollees. Our results demonstrate that all other things being equal, vulnerable children and adolescents shared in the benefits of SCHIP; however, all things are not equal. Vulnerable SCHIP enrollees may have different characteristics that are associated with worse outcomes (for example, minority enrollees were more likely to come from poorer, single-parent families with lower educational achievement and to have been uninsured for a long time).14 Studies that compare vulnerable and other enrollees without controlling for those characteristics are likely to have different results. Explaining the New York difference. In New York (unlike in Kansas or Florida), minority children and adolescents constituted a majority of SCHIP enrollees. This may have made it easier for New York health plans, providers, and policymakers to organize the delivery system in ways that increased minorities’ access to care. The high level of competition among health plans for SCHIP enrollees in the New York City market, in contrast to the Kansas market (one plan) and Florida market (at most two plans) could have caused plans to be more responsive to minority enrollees’ needs.15 Unmet needs after SCHIP enrollment. SCHIP reduced unmet needs. The fact that special-needs enrollees were more likely than other children to have unmet needs after SCHIP enrollment may in part be attributable to the fact that policy-makers in most states did not actively consider the more complex and more numerous health care needs of these enrollees when they decided to use commercial insurance norms in designing their freestanding SCHIP programs.16 Florida is one of the exceptions. It refers seriously impaired, SCHIP-eligible children and adolescents to the Children’s Medical Services Program (CMS), a Title V carve-out program. Florida’s special-needs children and adolescents in our study, however, were those who were not placed in CMS, either because their conditions were not severe enough or because CMS could not accommodate them because of a statewide cap that limits the number of enrollees with mental health conditions to 300. Florida’s special-needs enrollees who remained in the regular (not CMS) SCHIP program did not experience the reduction of unmet needs that other enrollees experienced. While special-needs children and adolescents tend to have higher unmet needs regardless of their insurance, states can pursue strategies to minimize them, including conducting needs assessments, identifying this population and risk-adjusting their capitation rates, and expanding benefits or arranging for wraparound services from other agencies. Leveling the playing field for the long-term uninsured. SCHIP enrollees who had been uninsured for a year prior to enrollment were no more likely than other enrollees to experience nonfinancial barriers to SCHIP; after SCHIP enrollment, the two groups of enrollees were virtually indistinguishable. It should be noted, however, that the long-term uninsured enrollees had a substantial connection with the health care system before enrolling in SCHIP that is not typical of uninsured children.17 This indicates that the long-term uninsured children who enrolled in SCHIP may be those who can best take advantage of their new coverage. Lack of gains in most measures among previously insured children and adolescents does not mean that they did not benefit from SCHIP. Many previously insured children lost coverage because of a change in their parents’ employment, the high costs of prior insurance, or a life change such as loss of Medicaid.18 In our samples, most of the previously insured children had been Medicaid enrollees. Many of them may have joined the ranks of the uninsured in the absence of SCHIP. Thus, SCHIP may have played a central role in maintaining the levels of access that these children enjoyed before enrolling in SCHIP. Reaching adolescents. Adolescents account for a sizable proportion of new SCHIP enrollees—almost one-third in 2001.19 Nevertheless, little attention was placed on adolescents’ unique needs when states initially designed their SCHIP benefit packages.20 It is therefore noteworthy that adolescent enrollees saw such marked improvement in health care and satisfaction. Targeting areas for future improvement. Our study shows that the usual-source-of-care rate can be increased to very high levels within one year of enrollment, and near-universal levels can be achieved for some subgroups. States may want to turn their attention to the quality of care provided by enrollees’ usual source of care and to developing mechanisms for smoothing transitions when enrollees change their source of care. Preventive visits showed only small gains, especially among Hispanics. States can explore a variety of options to improve this. Potential strategies include educating parents about the importance of preventive care, providing health plans or physicians with incentives for increasing the proportion of enrollees who meet guidelines for preventive care use, assessing the adequacy of health plans’ pediatric and family physicians and their location and hours of operation, forming coalitions with community-based organizations that work with underserved populations, and collaborating with educators to promote preventive care use (for example, permitting SCHIP reimbursement for school-based health clinics and having schools make preventive visits with a physician a prerequisite for participation in after-school sports). Study limitations. Low response rates. Several limitations of our study must be noted. First, the response rates of the surveys were low. Poor contact information (particularly phone numbers) contributed greatly to low T1 response rates. To the extent that the contact information was randomly missing (or wrong), this would not generate a selection bias. Multivariate models using administrative data found modest associations between T1 response rates and age, urban residence, and Hispanic ethnicity. Loss to T2 follow-up (13–50 percent) was also a strong contributor to the low overall response rates. Multivariate models predicting loss to follow-up as a function of T1 measures found that few measures were predictive (usual source of care, unmet needs, and age in Kansas; urban/rural status in New York; and Hispanic ethnicity in Florida). Nevertheless, if loss to follow-up was related to the pre-post effect size, our results could be biased.21 Different enrollment criteria for Kansas. Second, unlike Florida and New York, Kansas results included children who disenrolled from SCHIP within twelve months of enrollment, and measures of access and satisfaction included the period of time during which they were not enrolled. We performed bivariate and multivariate analyses for the full sample of children in New York to check for any possible bias this may have introduced into our results. We found no differences in dependent variables between children who disenrolled early and those who remained enrolled for at least twelve months; nor did we find any substantive differences when we estimated the multivariate models. The lack of sensitivity of the New York results to the inclusion of the early disenrollees, as well as the consistent patterns in the results, suggests that these differences may not be important. Regression to the mean. Third, we could not control for the possibility of regression to the mean. Our measures, however, showed higher levels of access before SCHIP than have been reported among uninsured children nationally, mitigating this concern.22 Concluding comments. SCHIP was passed to provide health insurance to children of low-income, working families. The current state fiscal environment has put that goal in jeopardy. A stagnant economy has led to widespread declines in state revenues, and states have responded with budget cuts to public programs. As of November 2003, six of the thirty-five states with separate or combination SCHIP programs had frozen enrollment, and other states have reduced eligibility, imposed more restrictive enrollment procedures, increased cost sharing, dropped benefits, and frozen or decreased provider reimbursement.23 When making decisions about SCHIP’s future, policymakers should consider our findings that SCHIP not only improved access to and satisfaction with health care for all enrollees, but it also conferred these benefits on the most vulnerable of low-income children.
Andrew Dick (Andrew_Dick{at}URMC.rochester.edu) is an assistant professor, Department of Community and Preventive Medicine, University of Rochester School of Medicine and Dentistry. Cindy Brach is a senior health policy researcher, Agency for Healthcare Research and Quality (AHRQ) Center for Delivery, Organizations, and Markets, in Rockville, Maryland. Andrew Allison is a researcher and director of health care finance and organization at the Kansas Health Institute in Topeka. Elizabeth Shenkman is an associate professor, Institute for Child Health Policy, University of Florida in Gainesville. Laura Shone is an assistant professor, Peter Szilagyi is a professor, and Jonathan Klein is an associate professor, Department of Pediatrics and Strong Children’s Research Center, University of Rochester School of Medicine and Dentistry, in Rochester, New York. Eugene Lewit is senior program manager at the David and Lucile Packard Foundation in Los Altos, California. The authors thank Nat Boyett and Karen VanLandeghem for their critical contributions, and Robert St. Peter for his leadership of the Kansas project. This work was supported by three cooperative agreements issued by the Agency for Healthcare Research and Quality (AHRQ) that included cofunding from the David and Lucile Packard Foundation and the Health Resources and Services Administration (AHRQ Grant nos. HS10465, HS10536, and HS10450). Additional support was provided for the Kansas project by the Kansas Health Foundation, the United Methodist Health Ministry Fund, and the Prime Health Foundation.
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Data And Methods
Study Results
