Predicting High Utilization of Emergency Department Services Among Patients with a Diagnosis of Psychosis in a Medicaid Managed Care Organization

Abstract

Studies have demonstrated increased utilization of medical services for patients with behavioral health diagnoses. Medicaid managed care organizations (MMCOs) that operate under behavioral health carve-outs face the challenge of effectively targeting disease management initiatives in the absence of information on behavioral diagnoses. This study sought to develop a predictive model of emergency department(ED) utilization for patients where a diagnosis of psychosis could be identified from a claim associated with a medical service provider visit. A retrospective cohort analysis was performed using medical and pharmacy claims from an MMCO in Philadelphia, Pennsylvania, to identify patients known to have a diagnosis of psychosis and to develop the predictive model. Demographics, comorbidities, medical utilization, and medications were assessed as predictor variables. Within the MMCO, 764 members were identified with at least one medical claim having a psychosis diagnosis. Ordinary least squares multiple regression analysis was performed to measure the correlation between independent variables and ED visits. Variables with significant F ratios in the regression analysis were retained as factors in a risk model to evaluate their additive and cumulative effects. Four variables were significant predictors of high ED utilization: prior number of ED visits, prior number of hospitalizations, history of alcohol abuse, and history of depression. ED utilization increased as the number of risk factors increased: With no risk factors, mean ED use was 0.58 visits (per 6 months), while the cumulative effects of all four factors equated to 8.5 ED visits. The model may be useful to other MMCOs, or similar organizations, seeking to risk-stratify their ED-related disease management activities for patients identified with psychosis.