Predicting psychiatric emergency room recidivism

To address the need to reduce psychiatric emergency-room (ER) recidivism and to promote continuity of care, this study developed a model using administratively obtainable variables to predict psychiatric ER visits in the six months following an index ER visit for a psychiatric condition. Data on member characteristics, preindex psychiatric ER use, index ER information, and postindex utilization for 1,029 adult HMO members visiting the ER for a psychiatric condition were collected and randomly divided in half. A regression model predicting psychiatric ER visits in the six months following the initial psychiatric ER visit was developed in the first data set and tested in the second. In both models, Medicaid insurance coverage, and preindex inpatient admissions for depression or substance abuse were identified as significant predictors of future psychiatric ER utilization. Administratively identifiable variables can be used to identify members at elevated risk of ER recidivism for psychiatric conditions. Through improved identification of risk, case management interventions can be strategically directed to members with the greatest need of services.