Prediction of Early Mortality in Post-CPR ICU Patients Using Machine Learning and Statistical Methods: A Retrospective Cohort Study
Rattachement africain : tr, kh, mm. Niveau de preuve : code pays fourni par la source.
Le résumé fourni par la source
Background/Objectives: Mortality rates remain high among patients admitted to the intensive care unit (ICU) following successful return of spontaneous circulation (ROSC) after cardiopulmonary resuscitation (CPR). Identifying risk factors specific to this patient group may directly inform clinical decision-making processes. This study aimed to identify the clinical and laboratory parameters associated with mortality in post-CPR ICU patients and to compare machine learning models developed using these parameters with traditional statistical analyses. Methods: This retrospective study included a total of 82 patients treated in a tertiary-level ICU between 2020 and 2023. The post-CPR group (n=41) consisted of patients admitted to the ICU following effective CPR and ROSC, while the control group (n=41) included randomly selected patients with similar clinical characteristics who had not undergone CPR. Demographic data, clinical scores (APACHE II, SOFA, NUTRIC), laboratory values, and survival outcomes were recorded. Mortality prediction models were developed using the Random Forest algorithm applied to class-balanced datasets generated with the ADASYN method. Results: The post-CPR group had significantly higher scores and biomarker levels, including APACHE II, SOFA, and CRP, whereas albumin and GFR levels were notably lower. Both ICU and hospital mortality rates were significantly elevated in this group (75.6% and 80.5%, respectively; p< 0.001). In general ICU mortality models developed using Random Forest, variables such as inotropic support, APACHE II, SOFA, and CRP emerged as prominent predictors, and the model demonstrated high predictive performance (AUC: 0.914). In the subgroup of post-CPR patients, factors such as thrombocyte count, mean platelet volume, and sex were found to be particularly influential in predicting mortality. Conclusions: Both traditional statistical analyses and machine learning models provide clinically meaningful results in predicting early mortality among post-CPR patients. In particular, the need for inotropic support and elevated inflammatory markers appear to be strong predictors of mortality. The high predictive performance of AI-supported models, even with small sample sizes, highlights their potential clinical utility, though prospective observational studies are needed to further validate these models. Registration: The dataset used for model development, along with the executable Python scripts, is available for sharing.
Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.
Le contrôle bibliographique ouvert
DOI retrouvé dans Crossref DOI retrouvé ; titre concordant.
- Titre Crossref
- Prediction of Early Mortality in Post-CPR ICU Patients Using Machine Learning and Statistical Methods: A Retrospective Cohort Study
- Date Crossref
- 07/05/2025
- Éditeur
- MDPI AG
- Type
- posted-content
Ce recoupement confirme des métadonnées liées au DOI. Il ne confirme ni la méthode ni les conclusions de l’étude, et il ne compte pas comme une seconde source scientifique indépendante.
Où se fait cette recherche
-
Gülhane Askerî Tıp Akademisi pays non établi dans la noticeÉtablissement de santé
-
University of Health Science pays non établi dans la noticeUniversité ou école supérieure
-
Ankara Atatürk Eğitim ve Araştırma Hastanesi pays non établi dans la noticeÉtablissement de santé
-
Sağlık Bilimleri Üniversitesi pays non établi dans la noticeUniversité ou école supérieure
-
Alanya University pays non établi dans la noticeUniversité ou école supérieure
-
Department of Public Health pays non établi dans la noticeOrganisme public
-
Ankara University pays non établi dans la noticeUniversité ou école supérieure
-
Memorial Ankara Hospital pays non établi dans la noticeÉtablissement de santé
-
Gulhane Training and Research Hospital Department of Intensive Care Unit pays non établi dans la noticeÉtablissement de santé
-
Faculty of Medicine Department of Anesthesiology pays non établi dans la noticeUniversité ou école supérieure
-
Mamak Public Hospital pays non établi dans la noticeÉtablissement de santé
-
Etlik City Hospital Department of Intensive Care pays non établi dans la noticeÉtablissement de santé
Gülhane Askerî Tıp Akademisi, University of Health Science et Ankara Atatürk Eğitim ve Araştırma Hastanesi, avec 9 autres affiliations.
Une affiliation ne permet pas de déduire la nationalité d’un auteur.