Development Of Prediction Models To Estimate Extubation Time And Midterm Recovery Time Of Ophthalmic Patients Undergoing General Anesthesia:a Cross-sectional Study

ObjectivesTo analyze risk factors of extubation time and midterm recovery time in ophthalmic patients underwent general anesthesia,and to develop prediction models to predict post-anesthesia extubation time and midterm recovery time for assisting medical staffs recognize high risk emergence delayed patients.MethodsThe perioperative data of patients underwent general anesthesia in Joint Shantou International Eye Center from January 2018 to October 2020 were retrospectively collected,including general patient data,laboratory examination results,anesthesia and surgical information.SPSS22.0 software was used for statistical analysis,and multiple linear regression was used to analyze risk factors of extubation time and midterm recovery time.The data set was randomly divided into training dataset,validation dataset and check dataset according to the ratio of 7:1.5:1.5.Matlab R2018a software was used to develop different prediction models of extubation time and midterm recovery time,including fuzzy neural network,stepwise linear regression model,regression tree model,ensembles of trees regression model and artificial neural network with different algorithms(Levenberg Marquardt Algorithm、Bayesian Regularization Algorithm、Scaled Conjugate Gradient Algorithm).Mean square error(MSE),root mean square error(RMSE),goodness of fit(R-squared/R~2)and model training time was calculated for training performance evaluation.Using the same and separated check dataset(N=274)to perform generalization evaluation for each model,MSE,RMSE,mean absolute error(MAE),mean absolute percentage error(MAPE)and R~2 were calculated respectively.ResultsA total of 1824 cases were collected for risk factors analysis and model development.There are 19 risk factors for extubation time(p<0.05)which are anesthesia time,Creatine kinase MB isoenzyme(CK-MB),mean corpuscular volume(MCV),ondansetron usage,mean corpuscular hemoglobin concentration,muscle relaxant types,preoperative atropine,nalbuphine,end-tidal CO2(ETCO2),cystatin,urinalysis,tidal volume,creatinine,platelet ratio,transfusion volume,operation history,anesthesiologists,surgery types,and surgeons.There are 21 risk factors for midterm recovery time(p<0.05)which are extubation time,postoperative body temperature,dexamethasone,operation time,preoperative atropine,nalbuphine,preoperative body temperature,transfusion volume,red blood cell distribution width,postoperative complications,total carbon dioxide,underlying diseases,dexmedetomidine usage,ondansetron usage,ETCO2,serum total cholesterol,serum calcium,muscle relaxant types,anesthesiologists,surgery types,and surgeons.The extubation time prediction models’goodness of fit from high to low are as followed:fuzzy neural network,stepwise linear regression model,artificial neural network(LM Algorithm),artificial neural network(BR Algorithm),artificial neural network(SCG Algorithm),ensembles of trees regression model and regression tree model with its R-squared value of 0.956,0.954,0.949,0.947,0.937,0.899,0.890.And the models’accuracy from high to low are fuzzy neural network,stepwise linear regression model,artificial neural network(LM Algorithm),artificial neural network(BR Algorithm),artificial neural network(SCG Algorithm),ensembles of trees regression model and regression tree model with the RMSE value of 6.637,6.778,7.144,7.304,7.935,10.092,10.504.The midterm recovery time prediction models’goodness of fit from high to low are as followed:fuzzy neural network,stepwise linear regression model and ensembles of trees regression model(same value),artificial neural network(BR Algorithm),regression tree model,artificial neural network(SCG Algorithm)and artificial neural network(LM Algorithm)with the R-squared value of 0.885,0.873,0.873,0.866,0.863,0.849,0.810.And the models’accuracy from high to low are fuzzy neural network,ensembles of trees regression model,stepwise linear regression model,artificial neural network(BR Algorithm),regression tree model,artificial neural network(SCG Algorithm)and artificial neural network(LM Algorithm)with the RMSE value of9.285,9.768,9.782,10.032,10.157,10.646,11.949.ConclusionThe prediction models constructed in this study had good predictive performance in predicting both extubation time and midterm recovery time of ophthalmic patients undergoing general anesthesia(extubation models R~2:0.890～0.956;midterm recovery models R~2:0.810～0.885).The fuzzy neural network developed in this study shows the best prediction performance and fast computing ability in both the extubation time and midterm recovery time prediction.