Psychiatric Inpatient Length of Stay and Needs: A Cluster Analysis using Machine Learning Algorithms.

Understanding distinct inpatient profiles among individuals with mental health disorders (MHD) can inform service planning and the development of targeted interventions. This study aimed to identify patient profiles based on hospital length of stay (LOS) and associated needs, predisposing, and enabling factors. A total of 600 MHD patients hospitalized in Quebec (Canada) between 2012 and 2018 were interviewed. Using K-means cluster analysis, three distinct profiles were identified and compared based on their clinical and sociodemographic characteristics. Profile 1 (27%) included mostly younger men, with the second-longest LOS. These patients were frequently diagnosed with personality and substance use disorders (SUD), exhibited high self-reported severity of psychotic and SUD symptoms, and had histories of aggressive behavior. They were labeled as unstable psychotic patients with comorbidities living in the community. Profile 2 (37%) had the longest LOS and consisted of older, unemployed individuals with psychotic disorders or autism spectrum disorder. Many lived in supervised housing or with family and were labeled as long-term severely psychotic patients supported in the community. Profile 3 (36%) had the shortest LOS and comprised mostly women with higher education, living with a spouse. They showed high distress at admission—characterized by mood and anxiety disorders, the highest allostatic load score, suicidal ideation, and low social functioning—and were labeled as patients in suicidal crisis from distressed community environments. Findings suggest that tailored interventions—Assertive Community Treatment for Profile 1, intensive home care for Profile 2, and crisis resolution teams for Profile 3—could enhance care and potentially reduce LOS.