Does selection bias affect children in Medicaid managed care? Enrollment and transitions of children in child welfare

Medicaid managed care (MC) has become the dominant form of Medicaid for children in many states, although it is a heterogeneous group, ranging from behavioral health carve-out to comprehensive physical care. Although some states now deliver Medicaid exclusively through MC or through traditional fee-for-service (FFS), many have a range of beneficiaries enrolled in different types of plans. While some enrollment patterns can be explained by geography, it is critical for public finance, public health, and disparities research to understand patterns of enrollment, disenrollment, and transitions in Medicaid MC. Theory and empirical evidence in other populations, such as Medicare adults, has demonstrated that MC organizations have an incentive to “cream skim” lower cost, healthier beneficiaries. However, limited analysis has been conducted to date on Medicaid children, partly as a result of data limitations.

We studied enrollment and transitions patterns among Medicaid children by using an innovative linkage of Medicaid claims (MAX) and survey data on abused or neglected children in child welfare. The MAX data span 36 states from 2000-2003. Panel and cross-sectional linear, logit, and multinomial logit models were used to analyze enrollment in behavioral health MC, comprehensive MC, and FFS plans. Controls included demographics, out-of-home placement (including foster care), general health status, rural, mental health needs (child behavior checklist (CBCL)), dummies for 4 types of child maltreatment, and state and year controls.

Descriptive (cross-sectional) models show that Blacks and older children were more likely to be enrolled in Medicaid behavioral health, and neglected children and those with higher mental health needs (CBCL) were less likely. White children, those with higher CBCL, neglected or physically abused, and those placed outside the home were more likely to receive FFS; blacks were less likely. Comprehensive but non-behavioral MC plans were more likely to include blacks and Hispanics, and much less likely to include children in fair/poor health, placed outside the home, or in rural areas. The FFS population was more stable (fewer dis-enrolled from these plans) than that in comprehensive or behavioral MC plans: 65% FFS never changed plans in the 4-year period compared to 54% comprehensive MC and 47% behavioral MC. Fewer factors predicted individual transitions, but older children were more likely to join behavioral MC plans and those with fair/poor health were much less likely.

Some evidence of favorable selection into (or retaining in) Medicaid managed care plans and adverse selection into (or retaining in) FFS was observed. Prior research has shown a strong linkage between the predictors in this analysis and increased Medicaid expenditures. Future research will need to assess whether MC enrollment is driven by caseworker behavior or MC plan behavior and the implications for state Medicaid finance.