Multiple Switching in Medicaid Managed Care: A Proportional Hazards Model

Abstract

This case study (n=41,969) aims to discover managerially useful predictors of multiple switching among HMOs in a Medicaid managed care population observed over 33 months. Cox's proportional hazards model was used to analyze eligibility data for the entire population, claims data for Medicaid services received during six months prior to HMO enrollment (sample n=2,474) and telephone interviews (sample n=656). Each analytic stage involved four comparisons: (1) enrollees with one switch compared with enrollees with no switches; (2) enrollees with multiple switches compared with those having no switches; (3) in relation to making the first switch, enrollees with multiple switches compared with those having one switch; and (4) in relation to making the second switch, multiple compared with those having one switch. Interruptions (which were independent of switches) predicted switching versus non-switching. Medical claims and, weakly, greater age were associated with multiple switching. Managers can use these three predictors to reduce switching and multiple switching. The finding that switching was associated with greater utilization before HMO enrollment contradicts findings for employer-sponsored insurance, but a possible explanation is offered.