Is psychiatric residential facility discharge possible and predictable? A multivariate analytical approach applied to a prospective study in Italy

Background A growing number of severely ill patients require long-term care in non-hospital residential facilities (RFs). Despite the magnitude of this development, longitudinal studies surveying fairly large resident samples and yielding important information on this population have been very few. Aims The aims of the study were (1) to describe the socio-demographic, clinical, and treatment-related characteristics of RF patients during an index period in 2010; (2) to identify predictors and characteristics associated with discharge at the 1-year follow-up; (3) to evaluate clinicians' predictions about each patient's likelihood of home discharge (HD). Methods A prospective observational cohort study was conducted involving all patients staying in 23 medium-long-term RFs of the St John of God Order with a primary psychiatric diagnosis. A comprehensive set of socio-demographic, clinical, and treatment-related information was gathered and standardized assessments (BPRS, HONOS, PSP, PHI, SLOF, RBANS) were administered to each participant. Logistic regression analyses were run to identify independent discharge predictors. Results The study involved 403 patients (66.7 % male), with a mean age of 49 years (SD = 10). The participants' average illness duration was 23 years; median value for length of stay in the RF was 2.2 years. The most frequent diagnosis was schizophrenia (67.5 %). 104 (25.8 %) were discharged: 13.6 % to home, 8.2 % to other RFs, 2.2 % to supported housing, and 1.5 % to prison. Clinicians' predictions about HD were generally erroneous. Conclusions Very few patients were discharged to independent accommodations after 1 year. The main variables associated with a higher HD likelihood were: illness duration of <15 years and effective social support during the previous year. Lower severity of psychopathology and higher working skill levels were also associated with a significantly greater HD likelihood.