Predicting Patients' Readmission

Abstract

This study explores whether clinicians or a statistical model can better identify patients at risk of early readmission and investigates variables potentially associated with clinicians' risk judgment. We focus on a total of 142 patients discharged from acute psychiatric wards in the Verona Mental Health Department (Italy). Psychiatrists assessed patients' risk of readmission at 30 and 90 days postdischarge, predicted their postdischarge compliance, and assessed their Global Assessment of Functioning (GAF) score at admission and discharge. Clinicians' judgment outperformed the statistical model, with the difference reaching statistical significance for 30-day readmission. Clinicians' readmission risk judgment, both for 30 and 90 days, was found to be statistically associated with predicted compliance with community treatment and GAF score at discharge. Clinicians' superior performance might be explained by their risk judgment depending on nonmeasurable factors, such as experience and intuition. Patients with a poorer GAF score at discharge and poor assumed compliance were predicted to have a higher risk of readmission.