Psychiatry

One-year KISMET program increased validated 12-month smoking cessation in severe mental illness

In a pragmatic cluster-RCT across 21 outpatient teams, the KISMET intervention increased validated 12-month cessation (OR 4.2, 95% CI 1.0–17.2) versus usual care. Benefits appeared at 3 months (OR 12.1) and persisted at 12 months but not at 6 months. Dropout was substantially higher in the intervention arm (58% vs 32%), limiting confidence in effect size estimates. No signal of physical or psychopathological harms was reported, but secondary outcomes showed no between-group differences. 1

LASSO-derived 30-day psychiatric readmission risk score for UAE inpatients

Using 14,994 admissions, a LASSO model predicted 30-day readmission with modest discrimination (validation AUC 0.649; temporal AUC 0.676). Key predictors included prior admissions, schizoaffective diagnosis, comorbidity burden, and medication markers such as first-generation antipsychotics. A simplified three-tier score stratified observed readmission risk into 6.7%, 10.0%, and 18.0% groups. Model outperformed recalibrated LACE and READMIT-subset indices but remains region-specific and requires external transport validation. 2

ML for comparative antidepressant selection remains immature; external validation limited

This systematic review included 19 studies comparing ML approaches for selecting among antidepressants, with AUCs varying from 0.59 to 0.95. Higher reported performance clustered in small samples using high-dimensional biological features and internal-only validation. Only seven studies performed external validation, which generally reduced performance estimates. Authors highlight confounding between prognostic and predictive features and scarce explainability as barriers to clinical use. 3