Recovery trajectories across intensive psychiatric settings: a multivariate analysis characterized by baseline profiles

Background: Understanding recovery variability in intensive psychiatric care is crucial, as variable-centered analyses often obscure individual differences. Aims: This study aimed to identify distinct recovery trajectories among adults in intensive psychiatric care settings and assess whether treatment modality or patient characteristics could predict trajectory membership. Methods: We used a KML3D machine-learning approach to analyze longitudinal data from 169 adults across three acute care settings: inpatient hospitalization, Soteria homes, and tech-assisted home care. Recovery was modeled across five outcome measures (symptomatology, functioning, quality of life), and a random forest model assessed predictors of trajectory membership. Results: Two distinct trajectories were identified: “Persistent Challenges” (54.4%) and “Accelerated Recovery” (45.6%). The “Persistent Challenges” group had more severe baseline presentations and improved slowly, while the “Accelerated Recovery” group started with minor-to-moderate severity and improved faster. Treatment modality and other predictors showed no meaningful relationship with recovery trajectory membership (kappa = 0.17, AUC = 0.67). Conclusions: Higher multidimensional baseline severity predicted more challenging recovery courses. The lack of predictive power for treatment modality supports the viability of hospitalization alternatives and highlights the value of multidimensional assessment in intensive psychiatric care settings.