Leveraging the Random Assignment of Medicaid Managed Care Plans to Study Plan Choices, Treatment Effects, and Cream Skimming

In South Carolina, the Medicaid program is administered through Managed Car Organizations (MCOs), which offer different health care plans to Medicaid beneficiaries. These plans differ in their generosity, network coverage, and other attributes, and they are ranked by the state using a system of “star ratings.” The system of MCOs offers choices to health care consumers and allows plans to compete for consumers. In South Carolina, when consumers do not make an active plan choice, the state uses an algorithm to assign plans to consumers automatically. Starting in 2017, this auto-assignment will be made using an explicitly random process. We propose to use this assignment feature to study the effect of plan assignment on patient outcomes such as health care utilization and expenditures. This prospective analysis will be complemented with a retrospective analysis that takes advantage of the state’s existing quasi-random round-robin assignment procedure to allocate households to plans. Additionally, we propose to combine the analysis of the randomly or quasi-randomly assigned population with the population that made active choices to try to distinguish between treatment and selection in accounting for which plans perform better.