Machine Learning For The Prediction Of Extubation In Mechanically Ventilated Patients In The Intensive Care Unit:A Meta-analysis

Purpose: We aimed to perform a meta-analysis to assess machine learning’s performance in the prediction of extubation for patients in the intensive care unit(ICU)and and to provide a theoretical basis for the use of machine learning models in the intensive care unit to predict the timing of extubation.Methods: We systematically searched for studies on machine learning to predict the timing of extubation in mechanically ventilated patients in the ICU in the seven databases until August 2021.The quality of the qualified literature after screening was evaluated and data was extracted.The primary endpoint was the areas under the receiver operating characteristic curve.Meta Di Sc 1.4,Stata 16.0 and In Stat software were used for meta analysis to pool sensitivity,specificity,positive likelihood ratio,negative likelihood ratio and diagnostic odds ratio,and the summary receiver operating characteristic curve was drawn.Results: We identified 250 citations that were published,of which 7 papers met the eligibility criteria and were used for the quantitative analysis.Compared to the logistic regression model with a pooled AUC of 0.81(interquartile range=0.77–0.84),the best performing machine learning models(n=6)pooled a higher AUC of 0.90(interquartile range=0.85–0.94).In the subgroup analysis,the DNN pooled a highest AUC of 0.9059.Conclusions: This meta-analysis shows that DNN in machine learning models achieved a better discrimination ability in predicting extubation in mechanically ventilated patients in the intensive care unit,which can provide relatively reliable guidance for clinical decision-making.