Predictors of In-Hospital Cardiac Arrest Outcomes: A Single-Center Observational Study

Abstract

Background/Objectives: In-hospital cardiac arrest (IHCA) carries high mortality and substantial risk of neurological and functional impairment. Given that contemporary, clinically relevant risk models remain limited, especially within Southern European systems, the aim of this study was to develop a process-aware model for bedside risk stratification. Methods: We retrospectively analyzed a single-center cohort from a prospectively maintained resuscitation registry (AHEPA University General Hospital, Thessaloniki). Adults (≥18 years) with index IHCA in 2017–2019 were included. Utstein-defined variables underwent univariable screening, LASSO selection, and collinearity checks before multivariable logistic regression for in-hospital mortality. We assessed discrimination (AUC) and calibration (Hosmer–Lemeshow). Results: Among 826 IHCAs, 137 survived to discharge and 689 died. Higher mortality was independently associated with longer CPR (aOR = 1.115, 95% CI: 1.080–1.158), older age (aOR = 1.034, 95% CI: 1.014–1.055), and CCU location (aOR = 7.303, 95% CI: 2.557–25.798), while operating room (aOR = 0.029, 95% CI: 0.003–0.252), ICU/HDU (aOR = 0.203, 95% CI: 0.065–0.630), and an initial shockable rhythm (aOR = 0.297, 95% CI: 0.144–0.611) were protective. Longer time to CPR initiation also predicted mortality (aOR = 1.746, 95% CI: 1.001–3.162). Model performance was strong (AUC = 0.897, 95% CI: 0.865–0.928) with good calibration (Hosmer–Lemeshow p = 0.879). Conclusions: A process-aware model integrating patient factors, intra-arrest metrics, and location showed excellent internal performance for predicting IHCA mortality. Findings reaffirm the prognostic importance of age, rhythm, and resuscitation timeliness/intensity and support future work extending prediction to neurological/functional outcomes and testing targeted care bundles in high-risk strata.