Advances in predictive modeling of postoperative delirium risk in cardiac surgery based on machine learning algorithms
10.3969/j.issn.1671-8348.2025.05.032
- VernacularTitle:基于机器学习算法的心脏外科术后谵妄风险预测模型的研究进展
- Author:
Jingli CHEN
1
;
Li GENG
Author Information
1. 武汉科技大学医学院,武汉 430065
- Keywords:
machine learning;
cardiac surgery;
delirium;
prediction model
- From:
Chongqing Medicine
2025;54(5):1225-1229,1234
- CountryChina
- Language:Chinese
-
Abstract:
Postoperative delirium(POD),as a highly prevalent acute neurological complication after cardiac surgery,has become a key challenge in perioperative management as it is significantly associated with adverse outcomes such as prolonged hospital stay,delayed functional recovery and increased mortality.With the expansion of medical data scale and the advancement of artificial intelligence technology,machine learning-based POD risk prediction models,with their advantages of efficiently integrating multi-dimensional data and analysing complex non-linear relationships,provide a new avenue for early detection and intervention of POD.Existing studies have shown that supervised learning and integration algorithms can significantly improve pre-diction performance through feature screening and weight optimisation,but model construction still faces chal-lenges such as high data heterogeneity and sample bias.Although external validation is a key link to promote clinical translation of models,barriers to inter-centre data sharing and privacy issues limit their practical appli-cation.In addition,current research generally suffers from shortcomings such as insufficient model interpret-ability,real-time prediction delay,and limited clinical adaptability.This article systematically reviewed the ap-plication progress,potential,and limitations of machine learning in predicting POD risks in cardiac surgery,providing multidimensional theoretical support for building a precise and personalized perioperative risk man-agement system.The aim is to promote the upgrading of clinical decision-making systems,improve patient prognosis,and achieve intelligent allocation of medical resources.
