{"title":"机器学习算法在预测胸部钝挫伤患者机械通气时间延长方面的性能。","authors":"Yifei Chen, Xiaoning Lu, Yuefei Zhang, Yang Bao, Yong Li, Bing Zhang","doi":"10.2147/TCRM.S482662","DOIUrl":null,"url":null,"abstract":"<p><strong>Purpose: </strong>Mechanical ventilation (MV) is one of the most common treatments for patients with blunt chest trauma (BCT) admitted to the intensive care unit (ICU). Our study aimed to investigate the performance of machine learning algorithms in predicting the prolonged duration of mechanical ventilation (PDMV) in patients with BCT.</p><p><strong>Methods: </strong>In this single-center observational study, patients with BCT who were treated with MV through nasal or oral intubation were selected. PDMV was defined as the duration of mechanical ventilation ≥7 days after endotracheal intubation (normal vs prolonged MV; dichotomous outcomes). K-means was used to cluster data from the original cohort by an unsupervised learning method. Multiple machine learning algorithms were used to predict DMV categories. The most significant predictors were identified by feature importance analysis. Finally, a decision tree based on the chi-square automatic interaction detection (CHAID) algorithm was developed to study the cutoff points of predictors in clinical decision-making.</p><p><strong>Results: </strong>A total of 426 patients and 35 characteristics were included. K-means clustering divided the cohort into two clusters (high risk and low risk). The area under the curve (AUC) of the DMV classification algorithms ranged from 0.753 to 0.923. The importance analysis showed that the volume of pulmonary contusion (VPC) was the most important feature to predict DMV. The prediction accuracy of the decision tree based on CHAID reached 86.4%.</p><p><strong>Conclusion: </strong>Machine learning algorithms can predict PDMV in patients with BCT. Therefore, limited medical resources can be more appropriately allocated to BCT patients at risk for PDMV.</p>","PeriodicalId":22977,"journal":{"name":"Therapeutics and Clinical Risk Management","volume":null,"pages":null},"PeriodicalIF":2.8000,"publicationDate":"2024-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11421453/pdf/","citationCount":"0","resultStr":"{\"title\":\"Performance of Machine Learning Algorithms in Predicting Prolonged Mechanical Ventilation in Patients with Blunt Chest Trauma.\",\"authors\":\"Yifei Chen, Xiaoning Lu, Yuefei Zhang, Yang Bao, Yong Li, Bing Zhang\",\"doi\":\"10.2147/TCRM.S482662\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<p><strong>Purpose: </strong>Mechanical ventilation (MV) is one of the most common treatments for patients with blunt chest trauma (BCT) admitted to the intensive care unit (ICU). Our study aimed to investigate the performance of machine learning algorithms in predicting the prolonged duration of mechanical ventilation (PDMV) in patients with BCT.</p><p><strong>Methods: </strong>In this single-center observational study, patients with BCT who were treated with MV through nasal or oral intubation were selected. PDMV was defined as the duration of mechanical ventilation ≥7 days after endotracheal intubation (normal vs prolonged MV; dichotomous outcomes). K-means was used to cluster data from the original cohort by an unsupervised learning method. Multiple machine learning algorithms were used to predict DMV categories. The most significant predictors were identified by feature importance analysis. Finally, a decision tree based on the chi-square automatic interaction detection (CHAID) algorithm was developed to study the cutoff points of predictors in clinical decision-making.</p><p><strong>Results: </strong>A total of 426 patients and 35 characteristics were included. K-means clustering divided the cohort into two clusters (high risk and low risk). The area under the curve (AUC) of the DMV classification algorithms ranged from 0.753 to 0.923. The importance analysis showed that the volume of pulmonary contusion (VPC) was the most important feature to predict DMV. The prediction accuracy of the decision tree based on CHAID reached 86.4%.</p><p><strong>Conclusion: </strong>Machine learning algorithms can predict PDMV in patients with BCT. Therefore, limited medical resources can be more appropriately allocated to BCT patients at risk for PDMV.</p>\",\"PeriodicalId\":22977,\"journal\":{\"name\":\"Therapeutics and Clinical Risk Management\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":2.8000,\"publicationDate\":\"2024-09-20\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11421453/pdf/\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Therapeutics and Clinical Risk Management\",\"FirstCategoryId\":\"3\",\"ListUrlMain\":\"https://doi.org/10.2147/TCRM.S482662\",\"RegionNum\":3,\"RegionCategory\":\"医学\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"2024/1/1 0:00:00\",\"PubModel\":\"eCollection\",\"JCR\":\"Q1\",\"JCRName\":\"Pharmacology, Toxicology and Pharmaceutics\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Therapeutics and Clinical Risk Management","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.2147/TCRM.S482662","RegionNum":3,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2024/1/1 0:00:00","PubModel":"eCollection","JCR":"Q1","JCRName":"Pharmacology, Toxicology and Pharmaceutics","Score":null,"Total":0}
Performance of Machine Learning Algorithms in Predicting Prolonged Mechanical Ventilation in Patients with Blunt Chest Trauma.
Purpose: Mechanical ventilation (MV) is one of the most common treatments for patients with blunt chest trauma (BCT) admitted to the intensive care unit (ICU). Our study aimed to investigate the performance of machine learning algorithms in predicting the prolonged duration of mechanical ventilation (PDMV) in patients with BCT.
Methods: In this single-center observational study, patients with BCT who were treated with MV through nasal or oral intubation were selected. PDMV was defined as the duration of mechanical ventilation ≥7 days after endotracheal intubation (normal vs prolonged MV; dichotomous outcomes). K-means was used to cluster data from the original cohort by an unsupervised learning method. Multiple machine learning algorithms were used to predict DMV categories. The most significant predictors were identified by feature importance analysis. Finally, a decision tree based on the chi-square automatic interaction detection (CHAID) algorithm was developed to study the cutoff points of predictors in clinical decision-making.
Results: A total of 426 patients and 35 characteristics were included. K-means clustering divided the cohort into two clusters (high risk and low risk). The area under the curve (AUC) of the DMV classification algorithms ranged from 0.753 to 0.923. The importance analysis showed that the volume of pulmonary contusion (VPC) was the most important feature to predict DMV. The prediction accuracy of the decision tree based on CHAID reached 86.4%.
Conclusion: Machine learning algorithms can predict PDMV in patients with BCT. Therefore, limited medical resources can be more appropriately allocated to BCT patients at risk for PDMV.
期刊介绍:
Therapeutics and Clinical Risk Management is an international, peer-reviewed journal of clinical therapeutics and risk management, focusing on concise rapid reporting of clinical studies in all therapeutic areas, outcomes, safety, and programs for the effective, safe, and sustained use of medicines, therapeutic and surgical interventions in all clinical areas.
The journal welcomes submissions covering original research, clinical and epidemiological studies, reviews, guidelines, expert opinion and commentary. The journal will consider case reports but only if they make a valuable and original contribution to the literature.
As of 18th March 2019, Therapeutics and Clinical Risk Management will no longer consider meta-analyses for publication.
The journal does not accept study protocols, animal-based or cell line-based studies.