Lowering health care costs is a policy priority for public health insurance programs. Policies that divert Emergency Department (ED) care to less costly, more effective health services are a promising avenue for cost savings. Using comprehensive, anonymized administrative data from the State of Rhode Island, which includes Medicaid and other social insurance programs, we demonstrate how government can identify and conduct efficient outreach to Medicaid recipients at risk of becoming high-cost ED users for potentially divertible care. The top predictors from our models include age, Medicaid enrollment and eligibility factors, and prior medical procedures. Our predictive models capture more future divertible spending than existing methods for identifying ED “super-utilizers” based on multiple prior ED visits. By using comprehensive administrative data that includes all the state's social insurance programs, we can also establish connections between predicted high-cost ED users and existing case management in non-Medicaid programs. Policymakers could use these models to improve their identification of divertible spending and reduce the need for
de novo
outreach and case management programs.