The Empoli REMS after five years: a single-centre retrospective observational cohort study of patient characteristics and time to exit

Background Evidence on time to exit from Italian Residences for the Execution of Security Measures (REMS) and its clinical and forensic correlates remains limited. This study described patients admitted during the first five years of activity of the Empoli REMS and estimated time to permanent exit. Methods This single-centre retrospective observational cohort included all 34 patients admitted between 22 July 2020 and 31 July 2025. Descriptive analyses used the full cohort. Two patients with a recorded stay of 0 months were excluded from time-to-event analyses. Permanent exit from the Empoli REMS was the event; patients still admitted at study end were right-censored. Kaplan-Meier estimates, log-rank tests, Cox models, and Benjamini-Hochberg false-discovery-rate correction were used. Results The sample was predominantly male (n = 30, 88.2%), with a mean age of 37.4 years (SD = 10.9); schizophrenia-spectrum disorders were the most frequent diagnoses (n = 19, 55.8%). The time-to-event cohort included 32 patients, with 13 exits and 19 censored observations. Observed follow-up was 20.1 months (SD = 13.3), and the Kaplan-Meier median time to exit was 24 months (95% CI: 19 months to not reached). A homicide-related index offence (HR = 0.19, 95% CI: 0.06-0.67, p = .009) and each additional 12 months of court-ordered security measure duration (HR = 0.38, 95% CI: 0.18-0.77, p = .008) were associated with a lower exit hazard; both associations remained significant after false-discovery-rate correction. No statistically significant associations were detected for migrant status, LAI treatment, or history of illicit drug use; however, the corresponding estimates were imprecise and the confidence intervals did not exclude potentially meaningful associations. Conclusions Within this small exploratory cohort, time to exit was associated primarily with offence-related and judicial variables. The findings require confirmation in larger multicentre cohorts.