Development and validation of multivariable risk adjustment models for return of spontaneous circulation and survival to hospital discharge following out-of-hospital cardiac arrest in England

Risk adjustment models can support clinical decision-making and enable comparative reporting to drive quality improvement. To develop and validate risk adjustment models for return of spontaneous circulation (ROSC) at hospital handover and survival to hospital discharge among patients experiencing out-of-hospital cardiac arrest (OHCA), using national registry data.

Patients with OHCA in England from 1 January 2016 and 31 December 2017, where resuscitation was attempted by the emergency medical service, were included. Data were sourced from the Out-of-Hospital Cardiac Arrest Outcomes registry. The 2016 cohort (n = 27 942) was used for model development and the 2017 cohort (n = 28 425) for validation. Outcomes were ROSC at hospital handover and survival to hospital discharge. Candidate predictors were age, sex, witnessed status, aetiology, bystander CPR, initial rhythm, and public access defibrillator use. Multivariable logistic regression models were developed using backward stepwise selection. Model performance was assessed using area under the receiver operating characteristic curve (AUC), Brier score, calibration plots, Hosmer–Lemeshow test, and classification metrics. Across the full study dataset, ROSC occurred in 28.6% of cases and survival in 8.2%. All candidate predictors were retained for ROSC, while sex was excluded from the final survival model. The ROSC model had an AUC of 0.702 (95% CI: 0.694–0.711) in development and 0.712 (0.704–0.719) in validation, with Brier scores of 0.182 in both. The survival model had an AUC of 0.877 (0.868–0.887) in development and 0.871 (0.862–0.879) in validation, with Brier scores of 0.059 and 0.061, respectively.

These validated models demonstrated strong performance, improved on previous models, and may support benchmarking, audit, and quality improvement initiatives.