Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

This paper proposes to take advantage of the unique way that Medicaid managed care was implemented in Kentucky to identify the impact of managed care on the health outcomes of children. As Kaestner, Dubay, and Kenney (2002) explain in their NBER working paper, 'Surprisingly, despite Medicaid's growing reliance on managed care, there is a relative paucity of empirical research on the subject. In particular, the effects of Medicaid managed care on health and how it affects children are two areas where empirical research is particularly scant.' KDK summarize much of the previous research on Medicaid managed care across different populations, concluding that the findings across many of the studies are inconsistent, and offer several critiques. First, much of the previous work is based on poorly thought-out non-experimental research designs. Issues about selection and omitted variables bias potentially contaminate the estimates of Medicaid managed care. Second, they criticize a number of empirical studies constructed from demonstration projects based on relatively small sample sizes and where the applicability of the results may be limited. The approaches takenin our paper address the problems of omitted variables bias and small sample sizes.

Kentucky introduced a Medicaid managed care program, Passport, into the county containing Louisville and the 16 counties immediately surrounding it in late 1997. We have obtained Medicaid claims and enrollment micro-data for all children in these 17 counties as well as all Medicaid data for all children in the ring of 14 counties surrounding the Passport counties that continued traditional fee-for-service (FFS) Medicaid. Our sample of full year enrolled children between 1996 and 1999 contains almost 42,000 children and over a million child months of data. This data provides detailed information on any contact with the health care system for which Medicaid was billed. Some of the relevant variables included would be the procedures that were done, their cost, physician and provider identifiers, diagnosis codes, and recipient demographics.

The Medicaid claims data allows users to find the exact address for a Medicaid enrollee. Instead of comparing recipients in Marion County (a Passport county) to Taylor County (a FFS county), for example, it is possible to compare recipients that live within one mile of the Marion/Taylor border. It is likely that as one makes the geographic area finer, the omitted county characteristics are less important for example, the job market and physical proximity to health care is essentially identical for those this close to the county border. On the other hand, the key difference that does not disappear with these finer groupings is the fact that the county of residence determines whether the Medicaid enrollee is in Passport or FFS. Therefore, by comparing the health care utilization and outcomes of 'border' counties to adjacent FFS counties, the analysis will be better able to control for selection and omitted variables bias. Initial results suggest that the introduction of managed care led to a reduction in the monthly number of children with positive Medicaid expenditures in the Passport counties relative to the FFS counties.