Development and validation of the risk of prolonged psychiatric admission score (R-PPAS) among adults in England: a population-based cohort study

Abstract Purpose Reducing longer hospital​ stays of acute mental health admissions is an urgent priority across psychiatric hospitals. However, there are limited risk prediction tools to identify prolonged admissions to support timelier treatment and discharge planning. We developed and validated a pragmatic risk prediction tool—the Risk of Prolonged Psychiatric Admission Score (R-PPAS)—to identify adult psychiatric admissions likely to exceed 59 days. Methods We derived a population-based cohort using electronic health records of 4208 adult acute psychiatric admissions (aged 18–65) across six wards in two English counties (January 2016 -March 2021). We developed a prediction model from 31 candidate predictors in a derivation cohort (four wards), and converted estimates into a simplified 0–7 scoring tool. We validated the final risk-scoring tool in a validation cohort involving two different wards in a different county. Results Overall, 977 (23.2%) of 4208 admissions were prolonged stays (derivation cohort: 579/2295 [25.2%]; validation cohort: 398/1913 [20.8%]). The final risk prediction tool (R-PPAS; score range 0–7) incorporated 14 predictors, including specific diagnoses (e.g., paranoid schizophrenia, severe depression), homelessness, risk of self-neglect or non-compliance with care, aggression and delayed allocation of a community-based care coordinator upon admission. The tool showed acceptable discrimination in the validation cohort (area under the curve=0.72, 95% CI 0.69–0.74), with sensitivity ranging from 77.6% at the recommended cut-off of ≥3 to 1.5% at ≥7, and specificity 55.2% to 99.9%. Conclusions We developed the R-PPAS, a validated, user-friendly risk stratification tool to identify adult patients at risk of prolonged psychiatric hospital stays to support timely triage, resource allocation and discharge planning. Further external validation is needed to determine its generalisability across regions, countries and healthcare systems.