|
TRENDSThe Impact Of Program Structure On Children’s Disenrollment From Medicaid And SCHIP
Program fragmentation might exacerbate disenrollment of children from Medicaid and the State Children’s Health Insurance Program (SCHIP). Using data from 2001–2004, I estimated the number of children who switched programs and the number who "dropped out" of public insurance—becoming uninsured despite continuing eligibility. Roughly two million children a year switched programs; 12.5 percent a year dropped out of Medicaid, and 15.6 percent a year dropped out of SCHIP. Children in states with separate SCHIP and Medicaid programs were 45 percent more likely to drop out. This effect persisted after controlling for demographic and policy variables. Administering multiple distinct public insurance programs appears to be counterproductive in terms of retention.
Incrementalism has become the method of choice in expanding health coverage to the roughly forty-five million uninsured Americans. Recent policies have been narrowly targeted and rarely coordinated with one another: the expansion of the Medicaid program to near-poor children and pregnant women in the late 1980s and early 1990s, the creation of the State Children’s Health Insurance Program (SCHIP) in 1997, and various state-specific waiver programs. This approach has had some success, especially among children, whose rate of uninsurance remains much lower than that of adults, largely because of Medicaid and SCHIP expansions.1 However, this fragmented approach—having multiple, separately run public programs for different age and income groups—could lead to problems with retention. If so, a more coordinated approach to insurance expansion might be needed to replace the current trend of patchwork solutions. Prevalence and impact of children’s coverage. Nearly 8.5 million U.S. children did not have any health coverage in 2003; however, more than 60 percent of them were eligible for Medicaid or SCHIP.2 Although some eligible children never enroll, others do enroll but then drop out. One recent study indicates that as many as three million children leave Medicaid/SCHIP each year and become uninsured, despite continuing eligibility.3 A large body of research indicates that insurance is critical to children’s health. Uninsured children get less frequent checkups and vaccinations and are more likely to go without needed care when sick, compared with their insured peers.4 Children with discontinuous coverage are 50 percent more likely than others to lack a regular source of care, which results in a tenfold increased risk of hospitalization for preventable conditions.5 Clearly, retention of eligible and needy children in Medicaid/SCHIP should be a policy priority. Program structure. Medicaid and SCHIP are administered by individual states using federal matching funds. In 2003, Medicaid covered twenty-four million children, while SCHIP covered four million.6 When Congress created SCHIP in 1997, each state had the option of creating a separately run program or using the additional funds to expand Medicaid eligibility. By 2001, thirty-three states had created SCHIP plans separate from Medicaid;, two more states set up separate programs by 2002. The remaining fifteen states and the District of Columbia administered a single program.7 Disenrollment. In both programs, families must renew coverage at least once a year by recertifying their eligibility. Failure to renew leads to loss of coverage. Previous research has categorized disenrollment into (1) loss of eligibility; (2) acquisition of other health insurance; and (3) "dropout," when a child remains eligible for Medicaid/SCHIP but leaves public insurance and becomes uninsured.8 The creation of two distinct programs in a single state adds a fourth possibility: program switching, which occurs when the renewal process indicates that a child in Medicaid has become eligible for SCHIP (or vice versa) because of changes in age or income. Program switching is administratively inefficient, with increased costs for monitoring and placing children in the correct program. But the more central outcome in terms of health care access is dropout. Previous studies have found that older children, healthier children, those with families paying premiums, and those living in regions with less physician access are all more likely to leave public insurance.9 Passive renewal, which assumes that enrollees remain eligible unless they report otherwise, has been found to greatly reduce disenrollment.10 An analysis of dropout in particular found that having a sibling in public insurance was the strongest protective factor against dropout and that higher Medicaid physician payment rates reduced dropout.11 But, overall, previous research suggests that the most important cause of dropout is procedural complexity. A review of SCHIP administrative files in eight states indicated that nearly one-fourth of renewal applications were denied because of incorrect paperwork, and 40 percent of enrollees never even reapplied.12 A survey of parents whose children left SCHIP found that confusion about the renewal process was a leading factor in their children’s disenrollment: 38 percent of parents did not know that they had to renew at all.13 Although some renewal procedures have been studied in previous research, the impact of Medicaid/SCHIP program structure on retention—one of the primary differences in programs across states—has not been explored. One previous study found that public insurance take-up among eligible children is roughly two to three percentage points higher in states with combined programs than in those with a separate SCHIP, but the study did not examine retention, where program coordination might have an even larger effect.14 Given the impact of parental confusion on retention, it seems likely that states with separate rather than combined programs would experience more dropout. This study explores the effects of program fragmentation on public insurance disenrollment. The two primary objectives are (1) to estimate how many children switch each year between Medicaid and SCHIP; and (2) to determine whether administering two separate programs—rather than a combined program—exacerbates dropout.
Data. The sample comes from the Current Population Survey (CPS) March supplement, a nationally representative survey conducted by the U.S. Census Bureau. The CPS uses a cycle of four months on, eight months off, and four months on for its respondents. Thus, in any given March, half of the households are contacted the following March. Households that move or fail to respond are dropped from the survey. The potential bias of this attrition, estimated in a similar previous CPS-based study to be 16 percent, is addressed in the Discussion section. Each child in this study’s sample is present in two consecutive years of the survey, and the sample (N = 8,473) comprises all children (ages 0–18) reporting Medicaid or SCHIP as their only form of insurance during their first year in the survey. The study period is 2001–2004, the first years for which the CPS differentiated between SCHIP and Medicaid. The CPS provides data on insurance coverage, demographics, income, and household sampling weights (used in all analyses). State-and year-specific eligibility for children is based on the threshold percentage of the federal poverty level for public coverage. To evaluate the effect on retention of combined versus separate Medicaid/SCHIP programs, I used multivariate regression to control for other policies affecting retention. Broadly speaking, there are two categories of policies to consider: those affecting the complexity of renewal, and those affecting the value of Medicaid/SCHIP coverage. In the former category, procedural variables include the frequency of renewal, a face-to-face renewal interview requirement, presumptive eligibility (allowing parents to enroll their children immediately in case of illness, verifying eligibility afterward), renewal outreach, and twelvemonth continuous eligibility.15 The monthly premium for each enrollee, required by some states for certain families, is imputed based on age, program, and family income.16 In the second category are state policies affecting the marginal benefit of public insurance, compared to a family’s next-best option, such as private insurance or charity care. These parameters include the Medicaid physician reimbursement rates and pediatrician participation rates for Medicaid and SCHIP, respectively.17 The analysis also includes Medicaid/SCHIP managed care penetration rates, program dollars spent per enrollee, and charity care availability.18 Financial variables are adjusted using inflation and state cost-of-living indices.19 Although this list of policies is meant to be comprehensive, it is impossible to capture every program difference across fifty states and the District of Columbia. To control for unobserved state policies and to proxy the level of stigma toward public insurance, the analysis includes a state ideology index. The index is based on the gap between the percentage of a state’s population that describes itself as "liberal" and "conservative," respectively, using survey data from 1995–1999.20 This index is a valid, reliable measure of public opinion and a very strong predictor of attitudes toward welfare spending.21 Of course, stigma is not strictly a state-level parameter and varies across individuals, but given that the CPS does not ask attitudinal questions, state-level ideology is the best available proxy for stigma. Methods. The key outcomes of this analysis are disenrollment and dropout. Disenrollment is the opposite of retention; it is the fraction of the population enrolled at one point in time no longer enrolled at a second point in time. For this analysis, retention can be defined in two ways: (1) Program-specific retention, which is the percentage of Medicaid enrollees from Year 1 who are still enrolled in Medicaid at Year 2, and the percentage of SCHIP enrollees from Year 1 who are still enrolled in SCHIP at Year 2. This approach classifies a child who switches from one program to the other as "disenrolling." (2) Overall public insurance retention, which is the percentage of Medicaid enrollees from Year 1 who are still enrolled in either SCHIP or Medicaid at Year 2, and the percentage of SCHIP enrollees from Year 1 who are still enrolled in either program at Year 2. The difference between program-specific retention and overall public insurance retention indicates the fraction of enrollees switching between programs. For the multivariate analysis, I first identified which children dropped out—becoming uninsured despite continued eligibility. Then I used logistic regression to estimate the impact of combined versus separate Medicaid/SCHIP programs on dropout. I estimated three models: For each regression, "dropout" is the outcome variable, and the key independent variable, "combined program," is whether a child lives in a state that runs separate or combined Medicaid/SCHIP programs. The regression sample excludes children that left public insurance because of new insurance or lost eligibility. Standard errors are clustered at the household level. The analysis uses three models to test the robustness of the findings. Each model has its advantages. Regression 1 is the most straightforward: "Combined program" is the only independent variable, so the regression estimates an unadjusted bivariate relationship. Since time trends or differences in populations and ideologies of states may confound bivariate results, Regression 2 controls for the year, demographics, and the ideology index, which is intended to capture stigma and unobservable variation across states. The demographic variables are age, highest level of parental education, family income (percentage of poverty), race/ethnicity, sex, sibling and parental coverage in Medicaid/SCHIP, urban versus rural residence, and self-reported health status. Regression 3 builds on Regression 2 by controlling for all of the state policies discussed in the previous section. As the most inclusive model, Regression 3 best addresses concerns of confounding. On the other hand, the sheer number of policy variables raises the possibility of collinearity; however, the more parsimonious Regression 2 should not suffer from this problem. Together, the three models provide a robust test of the impact of combined versus separate Medicaid/SCHIP programs. Nonetheless, Regressions 1–3 are cross-sectional analyses, which may be biased by unobserved state characteristics. One final test is to look at the effect of changes in program organization within a given state. Two states—Maryland and South Dakota—created new separate SCHIP programs during the study period. Using a multivariate regression in which the policies of Regression 3 are replaced by dummy variables for each state, I estimated the within-state effect of this program switch. To translate logistic odds ratios (ORs) into relative risks (RRs), I used the method of recycled predictions, in which the samplewide average probability of dropout is predicted using the regression coefficients, first when "combined program" is set equal to 0 for the whole sample, and again when it is set equal to 1.22 All data analysis was conducted using Stata 7.0.
Disenrollment, program switching, and dropout. SCHIP had much higher program-specific disenrollment than Medicaid (Exhibit 1  ). Overall public insurance disenrollment rates for the two programs were closer but still significantly different. Additionally, 27 percent of SCHIP enrollees switched to Medicaid each year, and 6 percent of Medicaid enrollees switched to SCHIP. In terms of numbers of children, the picture is one of nearly symmetric switching between programs: 27 percent of SCHIP enrollment is roughly one million children, and 6 percent of Medicaid enrollment is 1.4 million children. Also, of children who left public insurance, similar proportions of Medicaid and SCHIP disenrollees acquired new insurance or dropped out. But roughly twice as many SCHIP disenrollees lost eligibility.
Separate versus combined programs: impact on dropout. States with combined programs experienced an annual dropout rate of 9.6 percent, compared with 13.9 percent in states with separate programs—a statistically significant difference (chi square = 8.43, p = .004).
Exhibit 2
In Regression 3, which added to Regression 2 the Medicaid/SCHIP policies discussed above, the OR on "combined program" was 0.64 and statistically significant (p = .026). Using predicted probabilities from Regression 3, this result indicates that running separate Medicaid and SCHIP programs increases the risk of dropout by 45 percent (RR = 1.45). Only two other state-level variables were significant predictors of dropout: more generous per enrollee spending and more liberal state ideology both were associated with less dropout. Focusing on the states that switched program structures during the study period—Maryland and South Dakota—I found that after switching from a combined program to separate programs, these states’ aggregate dropout rate jumped from 4.6 percent to 36.6 percent (chi square = 8.19, p = .0056). In a longitudinal regression controlling for demographics and state of residence, the adjusted OR for "combined program" was 0.06 (95 percent confidence interval = 0.01, 0.51), a highly statistically significant result.
Administering two separate public insurance programs within the same state creates cracks through which children can fall out of public coverage: Adjusting for demographic and policy confounders, children living in states with separate Medicaid and SCHIP programs were 45 percent more likely to drop out than children living in states with combined programs. Furthermore, more than two million children a year switched back and forth between SCHIP and Medicaid, placing a heavy administrative burden on states with no discernible benefit in terms of health coverage. Protective effects of a combined program. The protective effect of a combined program persisted across multiple cross-sectional analyses controlling for a comprehensive list of demographic and policy variables. The same pattern was evident in a longitudinal analysis of states that changed program structures during the study period. However, the longitudinal results must be interpreted with a great deal of caution, since they depended on changes in the policies of only two states, which could have been altering other unmeasured policies simultaneously. Nonetheless, the longitudinal analysis supports the central finding that separate Medicaid/SCHIP programs experience more dropout than combined programs. Running separate programs creates fragmented family coverage, under which some children may be in SCHIP and others in Medicaid. Even if renewal procedures are streamlined, the existence of multiple programs likely confuses families about their coverage options and might end up discouraging parents from spending the time and energy to renew. Finally, adding into this discussion the fate of families who move to a new state (families not in this study, because the CPS does not retain movers), the complexity of having two programs per state—often going by different names state-to-state—is likely to be even more counterproductive, as was evident in the confusion regarding Medicaid coverage for refugees following Hurricane Katrina. Possible advantages of separate programs. Administering separate programs has no immediately obvious policy advantage. One might suspect that SCHIP programs would have higher take-up or lower dropout rates, because of less stigma toward the newer program than toward Medicaid, or because of more flexibility for states when designing SCHIP enrollment and outreach procedures. Yet in this study, dropout was actually significantly higher in SCHIP than in Medicaid (15.6 percent versus 12.5 percent), and previous research suggests that Medicaid/SCHIP take-up rates are lower in states where SCHIP is a separate program.23 A more cynical policy justification mentioned in previous research is that exacerbating dropout might in fact be the goal in some states seeking to reduce program spending.24 But given the research evidence on the adverse health outcomes of uninsurance, this clearly makes for flawed—and even punitive—public policy. Although administering separate programs might not have a convincing policy rationale, it likely reflects political considerations that cannot be ignored. In recent years, SCHIP has been less targeted for state and federal budget cutting than Medicaid, perhaps reflecting a greater willingness to maintain funding for a children-only program than for one that also covers adults. But the results of this analysis indicate that running a separate SCHIP—while potentially politically attractive—is both administratively costly and counterproductive in terms of promoting health care access among needy children. Study limitations. This analysis has several limitations, most notably pertaining to the CPS data set. Attrition bias in the linked samples and underreporting of Medicaid coverage in the CPS likely lead to, if anything, an underestimate of the true dropout rate: Families that mistakenly say they are not in Medicaid/SCHIP or who fall out of the CPS presumably are less likely to stay enrolled in public insurance than those families that acknowledge they are in Medicaid/SCHIP and whose circumstances are stable enough for them to remain in the survey for twelve months. Another concern is that the CPS provides annual income data rather than monthly data, which means that eligibility is estimated imperfectly; however, income mobility among Medicaid enrollees tends to be relatively limited, which means that any bias from this imprecision in income measurement is likely to be small.25
In terms of the reliability of the CPS’s new measure of SCHIP coverage, a comparison of CPS estimates with other SCHIP enrollment data indicates that there is an undercount, but it is smaller than the well-studied Medicaid undercount and should generate minimal bias, if any, in these results.26 Overall, the validity of the approach used here is supported by the fact that the disenrollment figures in Exhibit 1 The primary limitation of the multivariate analysis is that the relationships are correlational and not necessarily causal. Nonetheless, the important protective effect of a combined program was a persistent finding across multiple regressions with comprehensive demographic and policy controls. The robustness of the finding, combined with a plausible mechanism for the effect of program structure on retention, makes alternative explanations for the results less likely. This analysis adds to the growing evidence that retention is a critical challenge in improving children’s health in the United States. Both Medicaid and SCHIP have far too many eligible children—12–16 percent of enrollees—becoming uninsured each year. The incremental approach to health insurance expansion in the United States could be one of the key factors in this dropout problem, because it has created cracks in the system through which needy children can fall. To seal these cracks, states need to bring public insurance programs together under coordinated administration. Otherwise, any attempt to expand health coverage to the 8.5 million uninsured U.S. children—and to the forty-five million uninsured Americans overall—will be plagued by the steady dropout of those already enrolled in Medicaid and SCHIP.
Benjamin Sommers (bsommers{at}post.harvard.edu) holds a doctorate in health policy from Harvard Medical School and is in the school’s MD/PhD Program in the Social Sciences. This paper was written with the support of a graduate research fellowship from the National Science Foundation. The author thanks David Cutler, Kathy Swartz, and Joe Newhouse for reading through drafts and offering advice. He also thanks Bob Blendon, Mollyann Brodie, Janet Currie, Tom McGuire, Melissa Wachterman, and the participants in the Health Policy Research Seminar at Harvard for their helpful input. Any remaining errors are solely his responsibility.
This article has been cited by other articles:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Top
Study Data And Methods
Results
