Abstract / Summary
Abstract Survival of treated patients with cardiac arrest (CA) in the intensive care unit (ICU) remains low, and the Oxford Severity of Illness Score, Simplified Acute Physiology Score II (SAPS II), Systemic Inflammatory Response Syndrome criteria critically ill patients, but the cardiac arrest patients have not been studied. However, the results of studies in which different severity scores predict survival status in patients with cardiac arrest are unclear.The purpose of this study is to evaluate the effectiveness of the OASIA, SAPS II, SIRS, and SOFA scores in predicting 28-day mortality among patients with cardic ogenic shock (CS) in the intensive care unit (ICU), and to identify the scoring system with the highest diagnostic value. Methods : This study screened patients defined as CA from the MIMIC Ⅳ database. The OASIS, SAPS II, SIRS, and SOFA scores was divided into low, medium and high groups based on the Jorden Index. Statistical methods used included univariate and multivariate Cox regression analysis, ROC curves, propensity score matching, and generalized linear model analysis with inverse probability weighting, and patients identified with CA by ICD-9 codes were included in the sensitivity analysis. Results : This study included a total of 2,535 ICU patients following cardiac arrest, of whom 1,068 (42.1%) died within 28 days. Univariate analysis revealed that age, sex, multiple laboratory parameters, and four severity scores (OASIS, SAPS II, SIRS, SOFA) were all significantly associated with 28-day mortality (all P < 0.05). Multivariate Cox regression analysis further confirmed that blood urea nitrogen (HR = 1.003, P = 0.034), lactate (HR = 1.076, P < 0.001), and bilirubin (HR = 1.021, P = 0.044) were independent risk factors, while albumin (HR = 0.817, P < 0.001) and PO₂ (HR = 0.999, P < 0.001) were independent protective factors. Among the four scoring systems, OASIS (HR = 1.017, P < 0.001) and SAPS II (HR = 1.018, P < 0.001) retained independent predictive ability in the multivariate model, whereas SIRS and SOFA lost statistical significance. Stratified analysis showed that the moderate and high groups of OASIS, SAPS II, and SIRS were all significantly associated with 28-day mortality (all P < 0.001), whereas none of the SOFA groups showed a significant association. Sensitivity analysis (patients with cardiac arrest) and inverse probability weighting analysis further confirmed the robustness of these results. ROC curve analysis indicated that the predictive performance of SAPS II was slightly better than that of OASIS ( P = 0.023), but the AUC for both was below 0.7. Dicussion: Conclusion : The OASIS, SAPS II, and SIRS scores are robust independent predictors of 28-day mortality in ICU patients following cardiac arrest, whereas the predictive performance of the SOFA score has been inconsistent across different analytical methods. The predictive ability of SAPS II is slightly superior to that of OASIS, but both are only moderate. The clinical predictive value of a single scoring system is limited; therefore, it is recommended to combine multiple scoring systems with key laboratory parameters (particularly lactate and albumin) for a comprehensive prognostic assessment.