Predicting de-ventilator extubation in post-cardiac surgery patients using machine learning

Predicting de-ventilator extubation in post-cardiac surgery patients using machine learning | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Predicting de-ventilator extubation in post-cardiac surgery patients using machine learning Run Fu, Liang Hong, Lei Zhang, Xiaochun Song, Shangyu Chen, Yinying Xue, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7121733/v1 This work is licensed under a CC BY 4.0 License Status: Under Revision Version 1 posted 10 You are reading this latest preprint version Abstract Background : The aim of this study was to develop machine learning models to use machine learning algorithms to predict the factors associated with cardiac surgery that influence patient extubation from a ventilator after cardiac surgery. Method : Clinical data of cardiac surgery patients in Nanjing First Hospital between June 2019 and December 2022 were retrospectively extracted from electronic medical records. Five traditional machine learning algorithms, including logistic regression (LR), decision tree (DT), random forest (RF), extreme gradient boosting (XGB), and light gradient boosting machine (LGB), were used to construct predictive models for risk prediction of long-distance mechanical ventilation in post-cardiovascular surgery patients. The discriminative power of these models was assessed by the area under the receiver operating characteristic curve (AUC). Shapley Additive explanation(SHAP) was used to interpret the predictive models. Results : Data from 4487 patients were employed to train and validate a model of offline extubation risk in post-cardiac surgery patients. Among the full models, the RF model (AUC: 0.86; Sensitivity: 0.781, Specficity: 0.756) and the XGB model (AUC: 0.850; Sensitivity: 0.818; Specificity: 0.768) showed well predictive power for off-ventilator predicting. Eleven variables were finally selected by Boruta and LASSO features selection procedure, including age, hypertension, optime, preoperation EF, preoperation LVPW, HR, reoperation, body-weight, sex, AMBP, ST-II. Among the eleven variables, age, hypertension, operation time, preoperation EF, preoperation LVPW significantly contributed to the prediction model. Conclusion : In this study, we successfully developed several machine learning models to predict factors affecting off-ventilator extubation after cardiac surgery, which may be useful to help clinicians assess the success of off-ventilator extubation in cardiac surgery patients after surgery. Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction With the development of society, the morbidity and mortality rates of cardiovascular diseases have shown a markedly higher trend, and this has led to an increase in the burden of disease in society. According to domestic and international literature [1–2], the incidence of acute respiratory failure from various causes in cardiac intensive care patients after surgery is 3.5–15.9%. Whether it is coronary surgery, valve surgery or aortic coarctation surgery, acute respiratory insufficiency is one of the main complications in the early postoperative period after cardiac surgery [3]. Once acute respiratory insufficiency occurs in critically ill cardiac patients, it can significantly prolong the ICU stay and increase the prognostic risk. Current studies have confirmed that the implementation of mechanical ventilation is the main method of treating acute respiratory insufficiency after cardiac surgery. Its mechanism of action includes: increasing the oxygenation index, improving gas exchange, reducing the work of respiratory muscles, and reducing the preload and afterload of the heart [4]. Analyzed from the perspective of respiratory mechanics, the implementation of positive-pressure mechanical ventilation can increase airway pressure and increase lung volume, open alveoli, and improve lung compliance. However, for patients with low cardiac function, premature or late withdrawal of the ventilator will adversely affect the oxygen metabolic balance and hemodynamics of the patient, which may lead to complications such as pulmonary atelectasis, decreased cardiac output, and ventilator-associated pneumonia [5]. In addition, due to the different severity of postoperative respiratory failure caused by various different reasons, the required duration of mechanical ventilation varies. Therefore, appropriate ventilator deconditioning strategies should be formulated for cardiac intensive care patients to accurately grasp the deconditioning time window, maximize the success rate of deconditioning, and reduce the complications of mechanical ventilation. Failure to extubate as a result of premature or late ventilator withdrawal will raise the risk of reintubating the patient. Moreover, given the diversity and complexity of diseases accompanying cardiac critical care, the strategy set for ventilator deconditioning in postoperative cardiac surgery patients should show an individualized tendency [6].Unlike respiratory diseases such as ARDS and COPD, there is no corresponding clinical guideline or expert consensus on the mechanical ventilation strategy for cardiac critically ill patients [7–8]. Relevant studies [9–10] have indicated that cardiac surgical patients with previous cardiac surgery, reduced left ventricular ejection fraction, shock, and prolonged extracorporeal circulation may require longer periods of mechanical ventilation postoperatively.There is still a lack of high-quality evidence-based medical evidence on ventilator deconditioning strategies for cardiac critically ill patients. To this end, this study analyzes the clinical database of perioperative period of cardiac intensive care patients, combined with machine learning, artificial neural network, random forest algorithm and other technical means, with a view to provide auxiliary decision-making for the clinic through the artificial intelligence technology of ventilator off-loading strategy. Materials and methods Data sources and study population This retrospective study was conducted on 4487 consecutive patients admitted and received cardiac surgery at Nanjing First Hospital from June 2019 to December 2022. Patients who received cardiac surgery during the study period were recruited as the study objects, including coronary artery bypass, heart valve surgery, large vessel surgery, combined surgery,precordial surgery,etc. Exclusion criteria: (1) Patients under 18 years of age. (2) Patients who died or were discharged during or within 48 h after the operation. (3) Patients with incomplete clinical data, such as pre-operation echocardiographic measurements or intraoperative hemodynamic data. Data were collected from electronic medical records (EMR) database, and approval was gained from the Ethics Committee of Nanjing First Hospital(KY20230710-01-KS-01) Definition of patients with successful extubation The patient awakens from general anesthesia after surgery and can move around with commands, performs an SBT test (set ventilator parameters: pressure support (PS) set at 8–10 cmH₂O, PEEP set at 5 cmH₂O.), assesses hemodynamics: heart rate < 140 beats/min, systolic blood pressure 90–160 mmHg, maintained with no or small doses of antihypertensive medications, Oxygenation: P AO₂/FIO₂ ≥150 mmHg or S AO₂ ≥90% in case of FIO₂ ≤40% and PEEP ≤ 5–8 cmH₂O, along with a respiratory rate of 5 ml/kg, not accompanied by significant respiratory acidosis. Removal of tracheal intubation was given after passage, and the respiratory cycle was relatively stable after extubation and no re-tracheal intubation was performed within 48 h. High-flow or non-invasive assisted ventilation could be allowed to assist respiration. Data collection and preprocessing of data Clinical variables extracted from electronic medical records (EMR) database included demographics:age, sex, height, weight,Body Mass Index(BMI); comorbidities: stroke, hypertension, diabetes,coronary heart disease(CAD),chronic renal failure(CRF), atrial fibrillation(AF), chronic obstructive pulmonary disease (COPD); preoperative echocardiographic parameters: aortic arch(AO),left atrial diameter(LAD), left ventricular posterior wall thickness (LVPWT), left ventricular diastolic diameter (LVDd), left ventricular ejection fraction (LVEF); operation information: operation (OP) time, cardiopulmonary bypass (CPB) time, aortic block (Ab)time,temperature,degree of ST-segment elevation in lead II of the electrocardiogram(ST-II),premature ventricular contraction(PVC);Hemodynamic data:heart rate(HR),pulse,arterial systolic blood pressure(ASBP),arterial diastolic blood pressure(ADBP),mean arterial blood pressure (MABP),non-invasive systolic blood pressure(NSBP),non-invasive diastolic blood pressure(NDBP),central venous pressure (CVP);respiration parameter:respiratory rate(RR),SPO2,end-expiratory CO2(ETCO2),fraction of inspiration O2(FIO2);Post-Operative Information༚Whether ECMO, IABP, CRRT or CPR is given after surgery,New-onset atrial fibrillation. We categorized various heart surgeries with different keyword searches: 1. Coronary artery bypass graft surgery: “CABG” “coronary artery bypass graft” “Coronary artery bypass grafting”; 2. Valve surgery: “mitral valve replacement”“mitral valvuloplasty” “aortic valve replacement”“Aortic valvuloplasty”“Tricuspid valve replacement” “Tricuspid valvuloplasty”“MVR “MVP”“AVR” “AVP” “TVR” “TVP”“DVR” “TIVA”“Wheat” “David”; 3. Major vascular surgery: “full arch replacement” “full arch replacement” “right half arch replacement” “right half-arch replacement” “ascending aortic replacement” “ascending aortic replacement” “ascending aortoplasty “Bentall” “AAR” “AAP”; 4. Combined Surgery: Classification 1 and Classification 2; 5. Congenital Heart Surgery: “Atrial septal defect repair” “Ventricular septal defect repair” “ASD repair” “VSD repair”; 6.other cardiac surgeries: surgeries other than the above 5 categories. The detailed percentage is shown in Table 1 Table 1 Patient characteristics and clinical variables. Training set(N = 3140) Test set (N = 1347) P-value Demographic data Age (years) 61.72 ± 11.58 61.49 ± 11.77 0.531 Male, n (%) 1246(39.68%) 522(38.75%) 0.560 Height (cm) 164.94 ± 9.90 165.32 ± 8.06 0.220 Weight (kg) 65.95 ± 11.86 66.05 ± 11.92 0.805 BMI 20.33 ± 2.09 20.29 ± 2.12 0.571 Comorbidities Stroke 377(12.00%) 162(12.02%) 0.985 Hypertension 1530(48.72%) 652(48.40%) 0.843 Diabetes 709(22.57%) 273(20.26%) 0.08 CAD 1392(44.33%) 577(42.83%) 0.355 CRF 148(4.71%) 63(4.68%) 0.958 COPD 143(4.55%) 60(4.45%) 0.883 AF 562(17.89%) 267(19.82%) 0.128 Preoperation AO(mm) 34.53 ± 5.17 34.66 ± 5.46 0.445 Preoperation LAD(mm) 47.20 ± 10.35 47.48 ± 10.20 0.404 Preoperation LVPW(mm) 9.59 ± 1.51 9.57 ± 1.38 0.749 Preoperation LVDd(mm) 54.21 ± 9.07 54.45 ± 9.60 0.433 Preoperation EF(%) 58.14 ± 9.12 57.80 ± 9.81 0.083 Intraoperative hemodynamic data HR(times/minute) 55.82 ± 9.97 55.97 ± 10.26 0.654 PULSE(times/minute) 96.12 ± 20.53 96.52 ± 20.50 0.548 ASBP(mmHg) 75.47 ± 18.67 74.52 ± 19.43 0.124 ADBP(mmHg) 46.21 ± 12.09 45.97 ± 12.26 0.143 AMBP(mmHg) 65.85 ± 6.15 65.48 ± 6.59 0.068 continuation sheet NSBP(mmHg) 24.64 ± 14.13 25.15 ± 14.42 0.268 NDBP(mmHg) 15.64 ± 9.04 15.91 ± 9.33 0.365 CVP(mmHg) 6.83 ± 5.91 6.57 ± 5.22 0.174 Intraoperative respiratory parameters RR 13.91 ± 4.00 13.82 ± 4.05 0.517 SPO2 93.44 ± 8.81 93.46 ± 8.84 0.941 ETCO2 2.57 ± 3.95 2.59 ± 4.10 0.861 FIO2 2.32 ± 0.55 2.32 ± 0.53 0.642 Operative variables OP time 260.78 ± 66.21 261.66 ± 67.79 0.684 CPB time 115.70 ± 40.40 115.52 ± 40.23 0.894 Ab time 80.16 ± 33.12 80.43 ± 33.44 0.801 Temp 33.12 ± 2.97 33.16 ± 2.78 0.658 PVC 4.47 ± 0.99 4.47 ± 0.95 0.874 Table 1 (continued) Training set(N = 3140) Test set (N = 1347) P-value ST-II 3.16 ± 1.51 3.09 ± 1.03 0.161 Postoperative complication IABP 60(1.91%) 26(1.93%) 0.965 defibrill 36(1.15%) 9(0.66%) 0.141 CRRT 65(2.07%) 21(1.56%) 0.252 CPR 34(1.08%) 11(0.82%) 0.412 ECMO 7(0.22%) 2(0.15%) 0.609 Reoperation 66(2.10%) 33(2.44%) 0.467 Operation type CABG only,n(%) 949(30.22%) 391(29.02%) 0.232 Valve surgery only, n(%) 1463(46.59%) 627(46.54%) 0.978 Large Vessel Surgery,n(%) 134(4.26%) 60(4.45%) 0.778 Combined operation, Includes CABG and valve surgery,n(%) 317(10.09%) 135(10.02%) 0.940 Congenital surgery,n(%) 32(1.01%) 15(1.11%) 0.776 Other surgery, n (%) 245(7.80%) 119(8.83%) 0.246 BMI: body mass index;CAD:coronary artery disease;CRF:chronic renal failure;COPD:chronic obstructive pulmonary disease;AF:atrial fibrillation;AO:aortic arch,LAD:left atrial diameter; LVPW:left ventricular posterior wall;LVDd:left ventricular diastolic diameter;LVEF:left ventricular ejection fraction; Optime: operation time; EF: left ventricular ejection fraction; LVPWT: left ventricular posterior wall thickness; HR: heart rate;ASBP: arterial systolic blood pressure;ADBP:arterial diastolic blood pressure;AMBP:arterial mean blood pressure; NSBP:non-invasive systolic blood pressure;NDBP:non-invasive diastolic blood pressure;CVP:central venous pressure;RR:respiration parameter:respiratory rate,SPO2:Saturation of Peripheral Oxygen;ETCO2:end-expiratory CO2;FIO2:fraction of inspiration O2;OP time: Operation time, CPB time:cardio pulmonary bypass time,Ab time:aortic block time,Temp:temperature;PVC:premature ventricular contraction;ST-II:degree of ST-segment elevation in lead II of the electrocardiogram;IABP:Intra-Aortic Balloon Pulsation;CRRT:continuous renal replacement therapy;CPR:cardiopulmonary resuscitation;ECMO:extracorporeal membrane oxygenation. Model construction and evaluating The entire dataset was divided into a training set and a test set (7:3), which means that the proportion of de-ventilated and non-de-ventilated patients remained the same in both subsets. The training set was used to train the model with 10-fold cross-validation and the test set was used to evaluate the performance of the model. We used five traditional machine learning algorithms to construct the predictive model for de-ventilators, including logistic regression (LR), decision tree (DT), random forest (RF), extreme gradient boosting (XGB) and light gradient boosting machine(LGB). Boruta's algorithm and Least Absolute Shrinkage and Selection Operator (LASSO) are used to select the best subset of variables. All variables identified as significant by the Boruta algorithm were entered into the LASSO regression. Finally, the variables included the data identified by the LASSO regression in order to construct simplified models using the same five machine learing algorithms. Statistical analyses The baseline characteristics of patients in the training and test sets were compared. Measurements conforming to a normal distribution were expressed as mean ± standard deviation and analyzed using Student’s t-test. Non-normally distributed measurement data were presented as median [interquartile range (IQR)] and compared using the Wilcoxon rank-sum test. Categorical variables were summarized as frequency (percentage) and evaluated using Pearson’s χ² test, with Fisher’s exact test applied when expected cell frequencies were < 5. Statistical significance was set at P < 0.05. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC-ROC), accuracy, sensitivity, and specificity, while calibration was evaluated via calibration curves and the Brier score. For interpretability, Shapley Additive Explanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) were employed to generate consistent and locally accurate variable attributions within the prediction models. All analyses were performed using R (version 3.6.3) and Python (version 3.7). Results Between June 2019 and December 2022, a total of 4,487 patients were included in the analysis and evaluation in the Cardiovascular ICU of Nanjing First Hospital, Nanjing Medical University, Nanjing, China, all of whom were admitted to the ICU after surgery, and all of whom were assessed for extubation after surgery。 We randomized 70% of these 4487 patients into the training set and the remaining 30% into the test set.The clinical variables of patients in training and test set are listed in Table 1 .There was no difference between the patients in the training and test sets on these clinical variables. Figure 1 lists the AUC on the test set for the different medium models. The full models were conducted with all variables, using the six algorithms including LR, DT, RF, XGB, and LGB for offline extubation predicting, and the AUC, accuracy, sensitivity, and specificity of each full model on test set were presented in Fig. 1 . Among the full models, the RF model (AUC: 0.86; Sensitivity: 0.781, Specficity: 0.756) and the XGB model (AUC: 0.850; Sensitivity: 0.818; Specificity: 0.768) showed well predictive power for off-ventilator predicting. The main parameters of the full RF model were set as follows: bootstrap = True, criterion = “gini,” n_estimators = 500, max_depth = None, min_samples_leaf = 1, min_sample_split = 2. The main parameters of the full XBG model were set as follows: n_estimators = 200, learning_rate = 0.1, max_depth = 9, gamma = 0(Fig. 2 ). Feature selection was performed by the following two steps. First, Boruta algorithm was employed and 38 features were confirmed important to the prediction of off-ventilator patients. Then, Lasso regression was applied to select the best subset features fromthe 38 confirmed important features. Eleven variables were finally selected by Boruta and LASSO features selection procedure, including age, hypertension, optime, preoperation EF, preoperation LVPW, HR, reoperation, body-weight, sex, AMBP, ST-II(Fig. 3 ) . The SHAP summary plot (Fig. 3 ) and dependence plot (Fig. 4 ) represented the contributions of these eleven variables to the prediction of the RF model,with SHAP values above zero indicating an increased risk of offline extubation failure and SHAP values below zero indicating a decreased risk of offline extubation failure. For example, SHAP values for high age (red) were usually more than zero, indicating a icreased risk of offline extubation in patients with higher age. Figure 3 displays the ranking of the features based on the average absolute SHAP value. Among the eleven variables, age, hypertension, optime, preoperation EF, preoperation LVPW were the five variables with the greatest infuence on prediction power. Older age, a history of hypertension, longer duration of surgery, worse preoperative EF and thicker preoperative LV posterior wall thickness suggest an increased likelihood of postoperative cardiac off-ventilator difficulties. Discussion The use of AI technology to analyze clinical big data is a trend in the evolution between traditional medicine and precision medicine, and will contribute to the global application of precision medicine and the emergence of new health management models. In particular, its application in the cardiovascular field provides an extremely valuable research tool for the classification of disease phenotypes, risk prediction, and automatic interpretation of medical images [11–12]. Therefore, with the full cooperation of clinicians and information technology professionals, this study is the first to apply artificial intelligence techniques to construct a risk prediction model for long-term mechanical ventilation in postoperative cardiovascular patients. We selected five conventional machine learning algorithms to construct the prediction model respectively, and the final results showed that the Random Forest and XGBoost algorithms had the best prediction effect in predicting the duration of mechanical ventilation. Since the Random Forest algorithm is an integrated learning method based on decision trees, the model built is robust and can handle nonlinear problems. And XGBoost algorithm is a distributed machine learning algorithm, which runs fast and has high fault tolerance, and has shown good predictive value in risk modeling of several common ICU diseases [13–15]. And this study further confirms that artificial intelligence technology has a broad application prospect in disease risk prediction of cardiovascular critical care patients. Unlike the mathematical and statistical methods used in previous studies, this study applied machine learning algorithms to incorporate a more comprehensive set of clinical parameters, avoiding the bias associated with artificially selected study parameters. However, the complete data generated throughout the patient's disease cycle is too complex, which may lead to overfitting of the prediction model, thus reducing the model validity [16]. Therefore, in this study, data selection focused on past history, preoperative examination and surgery-related clinical parameters from the clinical characteristics of cardiovascular surgical patients, with the aim of utilizing relatively simple and easily accessible clinical indicators for risk prediction. Similar to the results of previous studies [17–19], the model of the present study showed that clinical features such as age, gender, weight, history of hypertension, duration of surgery, preoperative EF, intraoperative voluntary heart rate, and intraoperative mean arterial pressure were important predictive parameters of long-range mechanical ventilation in postoperative cardiac surgery patients. And in this study, the global importance of each predictive feature was calculated by SHAP algorithm. The results showed that age, history of hypertension, duration of surgery, preoperative EF, and preoperative left ventricular posterior wall thickness had a high predictive contribution to the predicted outcome. This shows that left ventricular systolic dysfunction may increase the risk of prolonged mechanical ventilation in postoperative cardiac surgery patients, which is different from the results of previous studies [20–22]. In other machine learning related studies, the interpretability of risk prediction models is often poor [23–24]. In the actual clinical work, if the unexplained prediction model is used directly, it will lead to the physician can only observe the risk but not know where the risk is during the treatment process, and can not take the corresponding medical measures. For this reason, this study used the dependency graph method to analyze the risk thresholds of important quantitative predictive characteristics such as age, weight, and surgery time from a macroscopic point of view, and to determine whether their risk thresholds were consistent with clinical experience. On the other hand, this study also used the LIME method to interpret the prediction results in each case and to demonstrate the characteristics that affect the prediction results from a micro perspective, with the aim of helping users to understand the specific risk factors for positive prediction results. The results of this study are the first of its kind. This study has some limitations. Firstly, we only used the information extracted from the clinical database of cardiac critical care in the Department of Critical Care Medicine of the First Hospital of Nanjing, which is a single-center study and has a relatively small amount of data. The predictive performance of the model may be different if different datasets are used. Second, the existing database cannot encompass all the clinical data of the patients throughout the course of treatment, and perhaps there are potential risk relationships that exist in the unrecorded data that need to be further explored. Finally, our study did not use risk scoring criteria established by other studies. Because each different risk-scoring model involves different characteristic variables, we preferred to explore risk-predictive parameters associated with long-range mechanical ventilation from real databases. Despite the limitations of the data used, the study methodology is generalized and can be used to continue to refine our predictive models based on gradually expanding the amount of data. Conclusion In this study, we successfully developed several machine learning models to predict factors affecting off-ventilator extubation after cardiac surgery, which may be useful to help clinicians assess the success of off-ventilator extubation in cardiac surgery patients after surgery. Declarations Data Availability The data that support the findings of the present study are available from the corresponding author upon reasonable request Ethics statement The study protocol was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Nanjing First Hospital, Nanjing Medical University (KY20230710-01-KS-01). Informed consent was not obtained due to the observational and anonymous nature of data collection. Author contributions RFu, HXu, LHong, LZhang, and CZhang conceived the conception of the study. HXu, RFu, LHong, LZhang, XSong, SChen and YXue acquired the data. LZhang, XSong and CZhang participated in data analyses. RFu, LHong and HXu constructed the predictive model. RFu, HXu and LZhang prepared the first draft of the manuscript. HXu and CZhang led the project and supervised the study. All authors were involved in writing or editing the manuscript, read, and approved the final version of the manuscript. Funding This study was funded by the 2022 Open Project (JSHD2022060); Jiangsu Province Capability Improvement Project through Science, Technology and Education (ZDXYS202210). The study sponsors had no involvement in the study design, the collection, analysis, and interpretation of data, the writing. Conflict of interest The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. References Thanavaro J, Taylor J, Vitt L, Guignon MS, Thanavaro S. Predictors and outcomes of postoperative respiratory failure after cardiac surgery. J Eval Clin Pract. 2020 Oct;26(5):1490-1497. Eremenko AA, Zyulyaeva TP. Postoperative acute respiratory failure in cardiac surgery. Khirurgiia (Mosk). 2019;(8):5-11. Stevens M, Shenoy AV, Munson SH, Yapici HO, Gricar BLA, Zhang X, Shaw AD. Healthcare utilization and costs of cardiopulmonary complications following cardiac surgery in the United States. PLoS One. 2019 Dec 19;14(12):e0226750. Hessels L, Coulson TG, Seevanayagam S, Young P, Pilcher D, Marhoon N, Bellomo R. Development and Validation of a Score to Identify Cardiac Surgery Patients at High Risk of Prolonged Mechanical Ventilation. J Cardiothorac Vasc Anesth. 2019 Oct;33(10):2709-2716. Fan E, Del Sorbo L, Goligher EC, et al. Clinical Practice Guideline: Mechanical Ventilation in Adult Patients with Acute Respiratory Distress Syndrome. Am J Respir Crit Care Med. 2017 May 1;195(9):1253-1263. Tavazzi G. Mechanical ventilation in cardiogenic shock. Curr Opin Crit Care. 2021 Aug 1;27(4):447-453. Pelosi P, Ball L, Barbas CSV, Bellomo R, Burns KEA, Einav S, Gattinoni L, Laffey JG, Marini JJ, Myatra SN, Schultz MJ, Teboul JL, Rocco PRM. Personalized mechanical ventilation in acute respiratory distress syndrome. Crit Care. 2021 Jul 16;25(1):250. Demoule A, Brochard L, Dres M, Heunks L, Jubran A, Laghi F, Mekontso-Dessap A, Nava S, Ouanes-Besbes L, Peñuelas O, Piquilloud L, Vassilakopoulos T, Mancebo J. How to ventilate obstructive and asthmatic patients. Intensive Care Med. 2020 Dec;46(12):2436-2449. Hessels L, Coulson TG, Seevanayagam S, Young P, Pilcher D, Marhoon N, Bellomo R. Development and Validation of a Score to Identify Cardiac Surgery Patients at High Risk of Prolonged Mechanical Ventilation. J Cardiothorac Vasc Anesth. 2019 Oct;33(10):2709-2716. Sim I. Two ways of knowing: big data and evidence-based medicine[J]． Ann Intern Med， 2016，164(8)： 562-563 Itchhaporia D. Artificial intelligence in cardiology. Trends Cardiovasc Med. 2022 Jan;32(1):34-41. Nashef SAM, Ali J. Artificial intelligence and cardiac surgery risk assessment. Eur J Cardiothorac Surg. 2023 Jun 1;63(6):ezad226. Yang L, Wu H, Jin X, et al. Study of cardiovascular disease prediction model based on random forest in eastern China. Sci Rep. 2020 Mar 23;10(1):5245. doi: 10.1038/s41598-020-62133-5. Hou N, Li M, He L, et al. Predicting 30-days mortality for MIMIC-III patients with sepsis-3: a machine learning approach using XGboost. J Transl Med. 2020 Dec 7;18(1):462. doi: 10.1186/s12967-020-02620-5. Qian Q, Wu J, Wang J, et al. Prediction Models for AKI in ICU: A Comparative Study. Int J Gen Med. 2021 Feb 25;14:623-632. doi: 10.2147/IJGM.S289671. Mutasa S, Sun S, Ha R. Understanding artificial intelligence based radiology studies: What is overfitting? Clin Imaging. 2020 Sep;65:96-99. Aksoy R, Karakoc AZ, Cevirme D, Elibol A, Yigit F, Yilmaz Ü, Rabus MB. Predictive Factors of Prolonged Ventilation Following Cardiac Surgery with Cardiopulmonary Bypass. Braz J Cardiovasc Surg. 2021 Dec 3;36(6):780-787. Michaud L, Dureau P, Kerleroux B, Charfeddine A, Regan M, Constantin JM, Leprince P, Bouglé A. Development and Validation of a Predictive Score for Prolonged Mechanical Ventilation After Cardiac Surgery. J Cardiothorac Vasc Anesth. 2022 Mar;36(3):825-832. O'Brien Z, Bellomo R, Williams-Spence J, Reid CM, Coulson T. Development and Validation of Scores to Predict Prolonged Mechanical Ventilation after Cardiac Surgery. J Cardiothorac Vasc Anesth. 2024 Feb;38(2):430-436. Santangelo E, Mongodi S, Bouhemad B, Mojoli F. The weaning from mechanical ventilation: a comprehensive ultrasound approach. Curr Opin Crit Care. 2022 Jun 1;28(3):322-330. Maisat W, Yuki K. Predictive Factors for Postoperative Intensive Care Unit Admission and Mechanical Ventilation After Cardiac Catheterization for Pediatric Pulmonary Vein Stenosis. J Cardiothorac Vasc Anesth. 2022 Aug;36(8 Pt A):2500-2508. Sanfilippo F, Di Falco D, Noto A, Santonocito C, Morelli A, Bignami E, Scolletta S, Vieillard-Baron A, Astuto M. Association of weaning failure from mechanical ventilation with transthoracic echocardiography parameters: a systematic review and meta-analysis. Br J Anaesth. 2021 Jan;126(1):319-330. Pinsky MR, Bedoya A, Bihorac A, Celi L, Churpek M, Economou-Zavlanos NJ, Elbers P, Saria S, Liu V, Lyons PG, Shickel B, Toral P, Tscholl D, Clermont G. Use of artificial intelligence in critical care: opportunities and obstacles. Crit Care. 2024 Apr 8;28(1):113. Gallifant J, Zhang J, Del Pilar Arias Lopez M, Zhu T, Camporota L, Celi LA, Formenti F. Artificial intelligence for mechanical ventilation: systematic review of design, reporting standards, and bias. Br J Anaesth. 2022 Feb;128(2):343-351. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Revision Version 1 posted Editorial decision: Revision requested 24 Apr, 2026 Reviews received at journal 19 Sep, 2025 Reviews received at journal 14 Sep, 2025 Reviewers agreed at journal 08 Sep, 2025 Reviewers agreed at journal 05 Sep, 2025 Reviewers invited by journal 29 Aug, 2025 Editor invited by journal 05 Aug, 2025 Editor assigned by journal 29 Jul, 2025 Submission checks completed at journal 26 Jul, 2025 First submitted to journal 26 Jul, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7121733","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":509654851,"identity":"8649a1de-6f6c-4bc8-9278-459f2c70916f","order_by":0,"name":"Run Fu","email":"","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Run","middleName":"","lastName":"Fu","suffix":""},{"id":509654852,"identity":"d27aea2b-92cc-449b-a318-5f2a41466e41","order_by":1,"name":"Liang Hong","email":"","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Liang","middleName":"","lastName":"Hong","suffix":""},{"id":509654854,"identity":"83aa3351-25d5-43ae-bd9f-5e96f0a78c68","order_by":2,"name":"Lei Zhang","email":"","orcid":"","institution":"Jiangsu Provincial Medical Key Laboratory of Fertility Protection and Health Technology Assessment","correspondingAuthor":false,"prefix":"","firstName":"Lei","middleName":"","lastName":"Zhang","suffix":""},{"id":509654855,"identity":"d5bfaacf-ae27-423c-9ac7-1558935fb091","order_by":3,"name":"Xiaochun Song","email":"","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Xiaochun","middleName":"","lastName":"Song","suffix":""},{"id":509654856,"identity":"3c01b409-592a-4295-9efa-34269a83fccb","order_by":4,"name":"Shangyu Chen","email":"","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Shangyu","middleName":"","lastName":"Chen","suffix":""},{"id":509654857,"identity":"38d796d7-3f77-49ab-a181-bc38f60ceaf7","order_by":5,"name":"Yinying Xue","email":"","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Yinying","middleName":"","lastName":"Xue","suffix":""},{"id":509654858,"identity":"7a915f08-4a7b-4549-9b79-398dd64fee07","order_by":6,"name":"Huan Xu","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAy0lEQVRIiWNgGAWjYDCCAzAGewMDA2MDcVqg6ngOQLWwEa1FIoFILXzHm58/+LiHIXHDzTeGD37uYMjjlyfgOskzxwwbZzxjMDa4nWNs2HuGoViyjYAtBjdyGJuB3pADajGT4G0DWneMSC08BjfPmP/8C9Syn1gtcgY3eMyYwbYQ8j7ILzNnHGAwljyTViwt2yaROONYAn4twBB78OHDAYbEvuOHN35822aT2N98gIA1EPAfiDkMgIQEUcphgP0BScpHwSgYBaNg5AAAfoNIV7GQuJkAAAAASUVORK5CYII=","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":true,"prefix":"","firstName":"Huan","middleName":"","lastName":"Xu","suffix":""},{"id":509654859,"identity":"37d930ed-ba8e-4ba8-b33a-2bc3c672eeb4","order_by":7,"name":"Cui Zhang","email":"","orcid":"","institution":"Nanjing Medical University","correspondingAuthor":false,"prefix":"","firstName":"Cui","middleName":"","lastName":"Zhang","suffix":""}],"badges":[],"createdAt":"2025-07-14 13:38:17","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7121733/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7121733/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90879848,"identity":"faa27dfc-1f14-4b89-add8-fafbad42ec84","added_by":"auto","created_at":"2025-09-09 09:37:21","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":405351,"visible":true,"origin":"","legend":"\u003cp\u003eComparison of AUCs among different machine learning models\u003c/p\u003e\n\u003cp\u003eLR, logistic regression; DT, decision tree; RF, random forestclassifier; LGB, light gradient boosting machine; XGB, extreme gradient boosting machine.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7121733/v1/63ac0155f05162db261ce347.png"},{"id":90881541,"identity":"400b3951-f086-4e2a-aa7a-eaf404995fa8","added_by":"auto","created_at":"2025-09-09 09:45:22","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":419955,"visible":true,"origin":"","legend":"\u003cp\u003eThe calibration curves and the Brier score of different machine learning models.\u003c/p\u003e\n\u003cp\u003eLR, logistic regression; DT, decision tree; RF, random forest classifier; LGB, light gradient boosting machine;\u003c/p\u003e\n\u003cp\u003eXGB, extreme gradient boosting machine\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7121733/v1/befec2d7c9e6103d959ba141.png"},{"id":90881542,"identity":"b6744f16-e46f-4246-9ea1-35ae72363f64","added_by":"auto","created_at":"2025-09-09 09:45:22","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":154788,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP summary plot of the reduced RF model\u003c/p\u003e\n\u003cp\u003eThe plot showed the importance of each variable\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7121733/v1/a77e3ea4436b7325f9c4f2f9.png"},{"id":90879853,"identity":"7589280e-a82c-4309-9832-c79d6055349c","added_by":"auto","created_at":"2025-09-09 09:37:22","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":146611,"visible":true,"origin":"","legend":"\u003cp\u003eSHAP dependence plot of the reduced RF model.\u003c/p\u003e\n\u003cp\u003eOptime: operation time; EF: left ventricular ejection fraction; LVPWT: left ventricular posterior wall thickness; HR: heart rate; AMBP: arterial mean blood pressure; ST-II: degree of ST-segment elevation in lead II of the electrocardiogram.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7121733/v1/8061922ff7e6f8e2a8b7f28c.png"},{"id":90884796,"identity":"8f47188f-055f-41eb-b7fe-fa0f223d03f2","added_by":"auto","created_at":"2025-09-09 10:01:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1854825,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7121733/v1/bcd100c5-3b06-434e-8df4-50d409cce5c6.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Predicting de-ventilator extubation in post-cardiac surgery patients using machine learning","fulltext":[{"header":"Introduction","content":"\u003cp\u003eWith the development of society, the morbidity and mortality rates of cardiovascular diseases have shown a markedly higher trend, and this has led to an increase in the burden of disease in society. According to domestic and international literature [1\u0026ndash;2], the incidence of acute respiratory failure from various causes in cardiac intensive care patients after surgery is 3.5\u0026ndash;15.9%. Whether it is coronary surgery, valve surgery or aortic coarctation surgery, acute respiratory insufficiency is one of the main complications in the early postoperative period after cardiac surgery [3]. Once acute respiratory insufficiency occurs in critically ill cardiac patients, it can significantly prolong the ICU stay and increase the prognostic risk. Current studies have confirmed that the implementation of mechanical ventilation is the main method of treating acute respiratory insufficiency after cardiac surgery. Its mechanism of action includes: increasing the oxygenation index, improving gas exchange, reducing the work of respiratory muscles, and reducing the preload and afterload of the heart [4].\u003c/p\u003e\u003cp\u003eAnalyzed from the perspective of respiratory mechanics, the implementation of positive-pressure mechanical ventilation can increase airway pressure and increase lung volume, open alveoli, and improve lung compliance. However, for patients with low cardiac function, premature or late withdrawal of the ventilator will adversely affect the oxygen metabolic balance and hemodynamics of the patient, which may lead to complications such as pulmonary atelectasis, decreased cardiac output, and ventilator-associated pneumonia [5]. In addition, due to the different severity of postoperative respiratory failure caused by various different reasons, the required duration of mechanical ventilation varies. Therefore, appropriate ventilator deconditioning strategies should be formulated for cardiac intensive care patients to accurately grasp the deconditioning time window, maximize the success rate of deconditioning, and reduce the complications of mechanical ventilation. Failure to extubate as a result of premature or late ventilator withdrawal will raise the risk of reintubating the patient. Moreover, given the diversity and complexity of diseases accompanying cardiac critical care, the strategy set for ventilator deconditioning in postoperative cardiac surgery patients should show an individualized tendency [6].Unlike respiratory diseases such as ARDS and COPD, there is no corresponding clinical guideline or expert consensus on the mechanical ventilation strategy for cardiac critically ill patients [7\u0026ndash;8]. Relevant studies [9\u0026ndash;10] have indicated that cardiac surgical patients with previous cardiac surgery, reduced left ventricular ejection fraction, shock, and prolonged extracorporeal circulation may require longer periods of mechanical ventilation postoperatively.There is still a lack of high-quality evidence-based medical evidence on ventilator deconditioning strategies for cardiac critically ill patients.\u003c/p\u003e\u003cp\u003eTo this end, this study analyzes the clinical database of perioperative period of cardiac intensive care patients, combined with machine learning, artificial neural network, random forest algorithm and other technical means, with a view to provide auxiliary decision-making for the clinic through the artificial intelligence technology of ventilator off-loading strategy.\u003c/p\u003e"},{"header":"Materials and methods","content":"\u003cp\u003e\u003cb\u003eData sources and study population\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThis retrospective study was conducted on 4487 consecutive patients admitted and received cardiac surgery at Nanjing First Hospital from June 2019 to December 2022. Patients who received cardiac surgery during the study period were recruited as the study objects, including\u003c/p\u003e\u003cp\u003ecoronary artery bypass, heart valve surgery, large vessel surgery, combined surgery,precordial surgery,etc. Exclusion criteria: (1) Patients under 18 years of age. (2) Patients who died or were discharged during or within 48 h after the operation. (3) Patients with incomplete clinical data,\u003c/p\u003e\u003cp\u003esuch as pre-operation echocardiographic measurements or intraoperative hemodynamic data. Data were collected from electronic medical records (EMR) database, and approval was gained from the Ethics Committee of Nanjing First Hospital(KY20230710-01-KS-01)\u003c/p\u003e\u003cp\u003e\u003cb\u003eDefinition of patients with successful extubation\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe patient awakens from general anesthesia after surgery and can move around with commands, performs an SBT test (set ventilator parameters: pressure support (PS) set at 8\u0026ndash;10 cmH₂O, PEEP set at 5 cmH₂O.), assesses hemodynamics: heart rate\u0026thinsp;\u0026lt;\u0026thinsp;140 beats/min, systolic blood pressure 90\u0026ndash;160 mmHg, maintained with no or small doses of antihypertensive medications, Oxygenation: P AO₂/FIO₂ \u0026ge;150 mmHg or S AO₂ \u0026ge;90% in case of FIO₂ \u0026le;40% and PEEP\u0026thinsp;\u0026le;\u0026thinsp;5\u0026ndash;8 cmH₂O, along with a respiratory rate of \u0026lt;\u0026thinsp;35 beats/min and a tidal volume (VT) of \u0026gt;\u0026thinsp;5 ml/kg, not accompanied by significant respiratory acidosis. Removal of tracheal intubation was given after passage, and the respiratory cycle was relatively stable after extubation and no re-tracheal intubation was performed within 48 h. High-flow or non-invasive assisted ventilation could be allowed to assist respiration.\u003c/p\u003e\u003cp\u003e\u003cb\u003eData collection and preprocessing of data\u003c/b\u003e\u003c/p\u003e\u003cp\u003eClinical variables extracted from electronic medical records (EMR) database included demographics:age, sex, height, weight,Body Mass Index(BMI); comorbidities: stroke, hypertension, diabetes,coronary heart disease(CAD),chronic renal failure(CRF), atrial fibrillation(AF), chronic obstructive pulmonary disease (COPD); preoperative echocardiographic parameters: aortic arch(AO),left atrial diameter(LAD), left ventricular posterior wall thickness (LVPWT), left ventricular diastolic diameter (LVDd), left ventricular ejection fraction (LVEF);\u003c/p\u003e\u003cp\u003eoperation information: operation (OP) time, cardiopulmonary bypass (CPB) time, aortic block (Ab)time,temperature,degree of ST-segment elevation in lead II of the electrocardiogram(ST-II),premature ventricular contraction(PVC);Hemodynamic data:heart rate(HR),pulse,arterial systolic blood pressure(ASBP),arterial diastolic blood pressure(ADBP),mean arterial blood pressure (MABP),non-invasive systolic blood pressure(NSBP),non-invasive diastolic blood pressure(NDBP),central venous pressure (CVP);respiration parameter:respiratory rate(RR),SPO2,end-expiratory CO2(ETCO2),fraction of inspiration O2(FIO2);Post-Operative Information༚Whether ECMO, IABP, CRRT or CPR is given after surgery,New-onset atrial fibrillation.\u003c/p\u003e\u003cp\u003eWe categorized various heart surgeries with different keyword searches: 1. Coronary artery bypass graft surgery: \u0026ldquo;CABG\u0026rdquo; \u0026ldquo;coronary artery bypass graft\u0026rdquo; \u0026ldquo;Coronary artery bypass grafting\u0026rdquo;; 2. Valve surgery: \u0026ldquo;mitral valve replacement\u0026rdquo;\u0026ldquo;mitral valvuloplasty\u0026rdquo; \u0026ldquo;aortic valve replacement\u0026rdquo;\u0026ldquo;Aortic valvuloplasty\u0026rdquo;\u0026ldquo;Tricuspid valve replacement\u0026rdquo; \u0026ldquo;Tricuspid valvuloplasty\u0026rdquo;\u0026ldquo;MVR \u0026ldquo;MVP\u0026rdquo;\u0026ldquo;AVR\u0026rdquo; \u0026ldquo;AVP\u0026rdquo; \u0026ldquo;TVR\u0026rdquo; \u0026ldquo;TVP\u0026rdquo;\u0026ldquo;DVR\u0026rdquo; \u0026ldquo;TIVA\u0026rdquo;\u0026ldquo;Wheat\u0026rdquo; \u0026ldquo;David\u0026rdquo;; 3. Major vascular surgery: \u0026ldquo;full arch replacement\u0026rdquo; \u0026ldquo;full arch replacement\u0026rdquo; \u0026ldquo;right half arch replacement\u0026rdquo; \u0026ldquo;right half-arch replacement\u0026rdquo; \u0026ldquo;ascending aortic replacement\u0026rdquo; \u0026ldquo;ascending aortic replacement\u0026rdquo; \u0026ldquo;ascending aortoplasty \u0026ldquo;Bentall\u0026rdquo; \u0026ldquo;AAR\u0026rdquo; \u0026ldquo;AAP\u0026rdquo;; 4. Combined Surgery: Classification 1 and Classification 2; 5. Congenital Heart Surgery: \u0026ldquo;Atrial septal defect repair\u0026rdquo; \u0026ldquo;Ventricular septal defect repair\u0026rdquo; \u0026ldquo;ASD repair\u0026rdquo; \u0026ldquo;VSD repair\u0026rdquo;; 6.other cardiac surgeries: surgeries other than the above 5 categories. The detailed percentage is shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePatient characteristics and clinical variables.\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTraining set(N\u0026thinsp;=\u0026thinsp;3140)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTest set (N\u0026thinsp;=\u0026thinsp;1347)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eP-value\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eDemographic data\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAge (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e61.72\u0026thinsp;\u0026plusmn;\u0026thinsp;11.58\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e61.49\u0026thinsp;\u0026plusmn;\u0026thinsp;11.77\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.531\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eMale, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1246(39.68%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e522(38.75%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.560\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHeight (cm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e164.94\u0026thinsp;\u0026plusmn;\u0026thinsp;9.90\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e165.32\u0026thinsp;\u0026plusmn;\u0026thinsp;8.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.220\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWeight (kg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e65.95\u0026thinsp;\u0026plusmn;\u0026thinsp;11.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e66.05\u0026thinsp;\u0026plusmn;\u0026thinsp;11.92\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.805\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eBMI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e20.33\u0026thinsp;\u0026plusmn;\u0026thinsp;2.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e20.29\u0026thinsp;\u0026plusmn;\u0026thinsp;2.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.571\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eComorbidities\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eStroke\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e377(12.00%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e162(12.02%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.985\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHypertension\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1530(48.72%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e652(48.40%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.843\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDiabetes\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e709(22.57%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e273(20.26%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.08\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCAD\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1392(44.33%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e577(42.83%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.355\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCRF\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e148(4.71%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e63(4.68%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.958\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCOPD\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e143(4.55%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60(4.45%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.883\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAF\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e562(17.89%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e267(19.82%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.128\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePreoperation AO(mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e34.53\u0026thinsp;\u0026plusmn;\u0026thinsp;5.17\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e34.66\u0026thinsp;\u0026plusmn;\u0026thinsp;5.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.445\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePreoperation LAD(mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e47.20\u0026thinsp;\u0026plusmn;\u0026thinsp;10.35\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e47.48\u0026thinsp;\u0026plusmn;\u0026thinsp;10.20\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.404\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePreoperation LVPW(mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e9.59\u0026thinsp;\u0026plusmn;\u0026thinsp;1.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9.57\u0026thinsp;\u0026plusmn;\u0026thinsp;1.38\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.749\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePreoperation LVDd(mm)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e54.21\u0026thinsp;\u0026plusmn;\u0026thinsp;9.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e54.45\u0026thinsp;\u0026plusmn;\u0026thinsp;9.60\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.433\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePreoperation EF(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e58.14\u0026thinsp;\u0026plusmn;\u0026thinsp;9.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e57.80\u0026thinsp;\u0026plusmn;\u0026thinsp;9.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.083\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eIntraoperative hemodynamic data\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eHR(times/minute)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e55.82\u0026thinsp;\u0026plusmn;\u0026thinsp;9.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e55.97\u0026thinsp;\u0026plusmn;\u0026thinsp;10.26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.654\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePULSE(times/minute)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e96.12\u0026thinsp;\u0026plusmn;\u0026thinsp;20.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e96.52\u0026thinsp;\u0026plusmn;\u0026thinsp;20.50\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.548\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eASBP(mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e75.47\u0026thinsp;\u0026plusmn;\u0026thinsp;18.67\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e74.52\u0026thinsp;\u0026plusmn;\u0026thinsp;19.43\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.124\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eADBP(mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e46.21\u0026thinsp;\u0026plusmn;\u0026thinsp;12.09\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e45.97\u0026thinsp;\u0026plusmn;\u0026thinsp;12.26\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.143\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAMBP(mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e65.85\u0026thinsp;\u0026plusmn;\u0026thinsp;6.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e65.48\u0026thinsp;\u0026plusmn;\u0026thinsp;6.59\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.068\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003econtinuation sheet\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNSBP(mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e24.64\u0026thinsp;\u0026plusmn;\u0026thinsp;14.13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e25.15\u0026thinsp;\u0026plusmn;\u0026thinsp;14.42\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.268\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNDBP(mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e15.64\u0026thinsp;\u0026plusmn;\u0026thinsp;9.04\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15.91\u0026thinsp;\u0026plusmn;\u0026thinsp;9.33\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.365\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCVP(mmHg)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e6.83\u0026thinsp;\u0026plusmn;\u0026thinsp;5.91\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e6.57\u0026thinsp;\u0026plusmn;\u0026thinsp;5.22\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.174\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eIntraoperative respiratory parameters\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e13.91\u0026thinsp;\u0026plusmn;\u0026thinsp;4.00\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e13.82\u0026thinsp;\u0026plusmn;\u0026thinsp;4.05\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.517\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSPO2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e93.44\u0026thinsp;\u0026plusmn;\u0026thinsp;8.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e93.46\u0026thinsp;\u0026plusmn;\u0026thinsp;8.84\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.941\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eETCO2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.57\u0026thinsp;\u0026plusmn;\u0026thinsp;3.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.59\u0026thinsp;\u0026plusmn;\u0026thinsp;4.10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.861\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eFIO2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e2.32\u0026thinsp;\u0026plusmn;\u0026thinsp;0.55\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2.32\u0026thinsp;\u0026plusmn;\u0026thinsp;0.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.642\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eOperative variables\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOP time\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e260.78\u0026thinsp;\u0026plusmn;\u0026thinsp;66.21\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e261.66\u0026thinsp;\u0026plusmn;\u0026thinsp;67.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.684\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCPB time\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e115.70\u0026thinsp;\u0026plusmn;\u0026thinsp;40.40\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e115.52\u0026thinsp;\u0026plusmn;\u0026thinsp;40.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.894\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAb time\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e80.16\u0026thinsp;\u0026plusmn;\u0026thinsp;33.12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e80.43\u0026thinsp;\u0026plusmn;\u0026thinsp;33.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.801\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTemp\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e33.12\u0026thinsp;\u0026plusmn;\u0026thinsp;2.97\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33.16\u0026thinsp;\u0026plusmn;\u0026thinsp;2.78\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.658\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePVC\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e4.47\u0026thinsp;\u0026plusmn;\u0026thinsp;0.99\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e4.47\u0026thinsp;\u0026plusmn;\u0026thinsp;0.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.874\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e(continued)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTraining set(N\u0026thinsp;=\u0026thinsp;3140)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTest set (N\u0026thinsp;=\u0026thinsp;1347)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eP-value\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eST-II\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e3.16\u0026thinsp;\u0026plusmn;\u0026thinsp;1.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e3.09\u0026thinsp;\u0026plusmn;\u0026thinsp;1.03\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.161\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003ePostoperative complication\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIABP\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e60(1.91%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e26(1.93%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.965\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003edefibrill\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e36(1.15%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e9(0.66%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.141\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCRRT\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e65(2.07%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e21(1.56%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.252\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCPR\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e34(1.08%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e11(0.82%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.412\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eECMO\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e7(0.22%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e2(0.15%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.609\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eReoperation\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e66(2.10%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e33(2.44%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.467\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003cb\u003eOperation type\u003c/b\u003e\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCABG only,n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e949(30.22%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e391(29.02%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.232\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eValve surgery only, n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e1463(46.59%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e627(46.54%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.978\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLarge Vessel Surgery,n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e134(4.26%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e60(4.45%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.778\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCombined operation, Includes CABG and valve surgery,n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e317(10.09%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e135(10.02%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.940\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCongenital surgery,n(%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e32(1.01%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e15(1.11%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.776\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOther surgery, n (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003e245(7.80%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003e119(8.83%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003e0.246\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003ctfoot\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eBMI: body mass index;CAD:coronary artery disease;CRF:chronic renal failure;COPD:chronic obstructive pulmonary disease;AF:atrial fibrillation;AO:aortic arch,LAD:left atrial diameter; LVPW:left ventricular posterior wall;LVDd:left ventricular diastolic diameter;LVEF:left ventricular ejection fraction; Optime: operation time; EF: left ventricular ejection fraction; LVPWT: left ventricular posterior wall thickness; HR: heart rate;ASBP: arterial systolic blood pressure;ADBP:arterial diastolic blood pressure;AMBP:arterial mean blood pressure; NSBP:non-invasive systolic blood pressure;NDBP:non-invasive diastolic blood pressure;CVP:central venous pressure;RR:respiration parameter:respiratory rate,SPO2:Saturation of Peripheral Oxygen;ETCO2:end-expiratory CO2;FIO2:fraction of inspiration O2;OP time:\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd colspan=\"4\"\u003eOperation time, CPB time:cardio pulmonary bypass time,Ab time:aortic block time,Temp:temperature;PVC:premature ventricular contraction;ST-II:degree of ST-segment elevation in lead II of the electrocardiogram;IABP:Intra-Aortic Balloon Pulsation;CRRT:continuous renal replacement therapy;CPR:cardiopulmonary resuscitation;ECMO:extracorporeal membrane oxygenation.\u003c/td\u003e\u003c/tr\u003e\u003c/tfoot\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003e\u003cb\u003eModel construction and evaluating\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe entire dataset was divided into a training set and a test set (7:3), which means that the proportion of de-ventilated and non-de-ventilated patients remained the same in both subsets. The training set was used to train the model with 10-fold cross-validation and the test set was used to evaluate the performance of the model.\u003c/p\u003e\u003cp\u003eWe used five traditional machine learning algorithms to construct the predictive model for de-ventilators, including logistic regression (LR), decision tree (DT), random forest (RF), extreme gradient boosting (XGB) and light gradient boosting machine(LGB).\u003c/p\u003e\u003cp\u003eBoruta's algorithm and Least Absolute Shrinkage and Selection Operator (LASSO) are used to select the best subset of variables. All variables identified as significant by the Boruta algorithm were entered into the LASSO regression. Finally, the variables included the data identified by the LASSO regression in order to construct simplified models using the same five machine learing algorithms.\u003c/p\u003e\u003cp\u003e\u003cb\u003eStatistical analyses\u003c/b\u003e\u003c/p\u003e\u003cp\u003eThe baseline characteristics of patients in the training and test sets were compared. Measurements conforming to a normal distribution were expressed as mean\u0026thinsp;\u0026plusmn;\u0026thinsp;standard deviation and analyzed using Student\u0026rsquo;s t-test. Non-normally distributed measurement data were presented as median [interquartile range (IQR)] and compared using the Wilcoxon rank-sum test. Categorical variables were summarized as frequency (percentage) and evaluated using Pearson\u0026rsquo;s χ\u0026sup2; test, with Fisher\u0026rsquo;s exact test applied when expected cell frequencies were \u0026lt;\u0026thinsp;5. Statistical significance was set at P\u0026thinsp;\u0026lt;\u0026thinsp;0.05. Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC-ROC), accuracy, sensitivity, and specificity, while calibration was evaluated via calibration curves and the Brier score. For interpretability, Shapley Additive Explanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) were employed to generate consistent and locally accurate variable attributions within the prediction models. All analyses were performed using R (version 3.6.3) and Python (version 3.7).\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003eBetween June 2019 and December 2022, a total of 4,487 patients were included in the analysis and evaluation in the Cardiovascular ICU of Nanjing First Hospital, Nanjing Medical University, Nanjing, China, all of whom were admitted to the ICU after surgery, and all of whom were assessed for extubation after surgery。\u003c/p\u003e\n\u003cp\u003eWe randomized 70% of these 4487 patients into the training set and the remaining 30% into the test set.The clinical variables of patients in training and test set are listed in Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e.There was no difference between the patients in the training and test sets on these clinical variables. Figure \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e lists the AUC on the test set for the different medium models.\u003c/p\u003e\n\u003cp\u003eThe full models were conducted with all variables, using the six algorithms including LR, DT, RF, XGB, and LGB for offline extubation predicting, and the AUC, accuracy, sensitivity, and specificity of each full model on test set were presented in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e. Among the full models, the RF model (AUC: 0.86; Sensitivity: 0.781, Specficity: 0.756) and the XGB model (AUC: 0.850; Sensitivity: 0.818; Specificity: 0.768) showed well predictive power for off-ventilator predicting. The main parameters of the full RF model were set as follows: bootstrap\u0026thinsp;=\u0026thinsp;True, criterion = \u0026ldquo;gini,\u0026rdquo; n_estimators\u0026thinsp;=\u0026thinsp;500, max_depth\u0026thinsp;=\u0026thinsp;None, min_samples_leaf\u0026thinsp;=\u0026thinsp;1, min_sample_split\u0026thinsp;=\u0026thinsp;2. The main parameters of the full XBG model were set as follows: n_estimators\u0026thinsp;=\u0026thinsp;200, learning_rate\u0026thinsp;=\u0026thinsp;0.1, max_depth\u0026thinsp;=\u0026thinsp;9, gamma\u0026thinsp;=\u0026thinsp;0(Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\n\u003cp\u003eFeature selection was performed by the following two steps. First, Boruta algorithm was employed and 38 features were confirmed important to the prediction of off-ventilator patients. Then, Lasso regression was applied to select the best subset features fromthe 38 confirmed important features. Eleven variables were finally selected by Boruta and LASSO features selection procedure, including age, hypertension, optime, preoperation EF, preoperation LVPW, HR, reoperation, body-weight, sex, AMBP, ST-II(Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e) .\u003c/p\u003e\n\u003cp\u003eThe SHAP summary plot (Fig. \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e) and dependence plot (Fig. \u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e) represented the contributions of these eleven variables to the prediction of the RF model,with SHAP values above zero indicating an increased risk of offline extubation failure and SHAP values below zero indicating a decreased risk of offline extubation failure. For example, SHAP values for high age (red) were usually more than zero, indicating a icreased risk of offline extubation in patients with higher age. Figure \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e displays the ranking of the features based on the average absolute SHAP value. Among the eleven variables, age, hypertension, optime, preoperation EF, preoperation LVPW were the five variables with the greatest infuence on prediction power. Older age, a history of hypertension, longer duration of surgery, worse preoperative EF and thicker preoperative LV posterior wall thickness suggest an increased likelihood of postoperative cardiac off-ventilator difficulties.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe use of AI technology to analyze clinical big data is a trend in the evolution between traditional medicine and precision medicine, and will contribute to the global application of precision medicine and the emergence of new health management models. In particular, its application in the cardiovascular field provides an extremely valuable research tool for the classification of disease phenotypes, risk prediction, and automatic interpretation of medical images [11\u0026ndash;12]. Therefore, with the full cooperation of clinicians and information technology professionals, this study is the first to apply artificial intelligence techniques to construct a risk prediction model for long-term mechanical ventilation in postoperative cardiovascular patients. We selected five conventional machine learning algorithms to construct the prediction model respectively, and the final results showed that the Random Forest and XGBoost algorithms had the best prediction effect in predicting the duration of mechanical ventilation. Since the Random Forest algorithm is an integrated learning method based on decision trees, the model built is robust and can handle nonlinear problems. And XGBoost algorithm is a distributed machine learning algorithm, which runs fast and has high fault tolerance, and has shown good predictive value in risk modeling of several common ICU diseases [13\u0026ndash;15]. And this study further confirms that artificial intelligence technology has a broad application prospect in disease risk prediction of cardiovascular critical care patients.\u003c/p\u003e\u003cp\u003eUnlike the mathematical and statistical methods used in previous studies, this study applied machine learning algorithms to incorporate a more comprehensive set of clinical parameters, avoiding the bias associated with artificially selected study parameters. However, the complete data generated throughout the patient's disease cycle is too complex, which may lead to overfitting of the prediction model, thus reducing the model validity [16]. Therefore, in this study, data selection focused on past history, preoperative examination and surgery-related clinical parameters from the clinical characteristics of cardiovascular surgical patients, with the aim of utilizing relatively simple and easily accessible clinical indicators for risk prediction.\u003c/p\u003e\u003cp\u003eSimilar to the results of previous studies [17\u0026ndash;19], the model of the present study showed that clinical features such as age, gender, weight, history of hypertension, duration of surgery, preoperative EF, intraoperative voluntary heart rate, and intraoperative mean arterial pressure were important predictive parameters of long-range mechanical ventilation in postoperative cardiac surgery patients. And in this study, the global importance of each predictive feature was calculated by SHAP algorithm. The results showed that age, history of hypertension, duration of surgery, preoperative EF, and preoperative left ventricular posterior wall thickness had a high predictive contribution to the predicted outcome. This shows that left ventricular systolic dysfunction may increase the risk of prolonged mechanical ventilation in postoperative cardiac surgery patients, which is different from the results of previous studies [20\u0026ndash;22].\u003c/p\u003e\u003cp\u003eIn other machine learning related studies, the interpretability of risk prediction models is often poor [23\u0026ndash;24]. In the actual clinical work, if the unexplained prediction model is used directly, it will lead to the physician can only observe the risk but not know where the risk is during the treatment process, and can not take the corresponding medical measures. For this reason, this study used the dependency graph method to analyze the risk thresholds of important quantitative predictive characteristics such as age, weight, and surgery time from a macroscopic point of view, and to determine whether their risk thresholds were consistent with clinical experience. On the other hand, this study also used the LIME method to interpret the prediction results in each case and to demonstrate the characteristics that affect the prediction results from a micro perspective, with the aim of helping users to understand the specific risk factors for positive prediction results. The results of this study are the first of its kind.\u003c/p\u003e\u003cp\u003eThis study has some limitations. Firstly, we only used the information extracted from the clinical database of cardiac critical care in the Department of Critical Care Medicine of the First Hospital of Nanjing, which is a single-center study and has a relatively small amount of data. The predictive performance of the model may be different if different datasets are used. Second, the existing database cannot encompass all the clinical data of the patients throughout the course of treatment, and perhaps there are potential risk relationships that exist in the unrecorded data that need to be further explored. Finally, our study did not use risk scoring criteria established by other studies. Because each different risk-scoring model involves different characteristic variables, we preferred to explore risk-predictive parameters associated with long-range mechanical ventilation from real databases. Despite the limitations of the data used, the study methodology is generalized and can be used to continue to refine our predictive models based on gradually expanding the amount of data.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn this study, we successfully developed several machine learning models to predict factors affecting off-ventilator extubation after cardiac surgery, which may be useful to help clinicians assess the success of off-ventilator extubation in cardiac surgery patients after surgery.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData Availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe data that support the findings of the present study are available from the corresponding author upon reasonable request\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics statement\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe study protocol was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of Nanjing First Hospital, Nanjing Medical University (KY20230710-01-KS-01). Informed consent was not obtained due to the observational and anonymous nature of data collection.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRFu, HXu, LHong, LZhang, and CZhang conceived the conception of the study. HXu, RFu, LHong, LZhang, XSong, SChen and YXue acquired the data. LZhang, XSong and CZhang participated in data analyses. RFu, LHong and HXu constructed the predictive model. RFu, HXu and LZhang prepared the first draft of the manuscript. HXu and CZhang led the project and supervised the study. All authors were involved in writing or editing the manuscript, read, and approved the final version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was funded by the 2022 Open Project (JSHD2022060); Jiangsu Province Capability Improvement Project through Science, Technology and Education (ZDXYS202210). The study sponsors had no involvement in the study design, the collection, analysis, and interpretation of data, the writing.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConflict of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eThanavaro J, Taylor J, Vitt L, Guignon MS, Thanavaro S. Predictors and outcomes of postoperative respiratory failure after cardiac surgery. J Eval Clin Pract. 2020 Oct;26(5):1490-1497.\u003c/li\u003e\n\u003cli\u003eEremenko AA, Zyulyaeva TP. Postoperative acute respiratory failure in cardiac surgery. Khirurgiia (Mosk). 2019;(8):5-11.\u003c/li\u003e\n\u003cli\u003eStevens M, Shenoy AV, Munson SH, Yapici HO, Gricar BLA, Zhang X, Shaw AD. Healthcare utilization and costs of cardiopulmonary complications following cardiac surgery in the United States. PLoS One. 2019 Dec 19;14(12):e0226750.\u003c/li\u003e\n\u003cli\u003eHessels L, Coulson TG, Seevanayagam S, Young P, Pilcher D, Marhoon N, Bellomo R. Development and Validation of a Score to Identify Cardiac Surgery Patients at High Risk of Prolonged Mechanical Ventilation. J Cardiothorac Vasc Anesth. 2019 Oct;33(10):2709-2716.\u003c/li\u003e\n\u003cli\u003eFan E, Del Sorbo L, Goligher EC, et al. Clinical Practice Guideline: Mechanical Ventilation in Adult Patients with Acute Respiratory Distress Syndrome. Am J Respir Crit Care Med. 2017 May 1;195(9):1253-1263.\u003c/li\u003e\n\u003cli\u003eTavazzi G. Mechanical ventilation in cardiogenic shock. Curr Opin Crit Care. 2021 Aug 1;27(4):447-453.\u003c/li\u003e\n\u003cli\u003ePelosi P, Ball L, Barbas CSV, Bellomo R, Burns KEA, Einav S, Gattinoni L, Laffey JG, Marini JJ, Myatra SN, Schultz MJ, Teboul JL, Rocco PRM. Personalized mechanical ventilation in acute respiratory distress syndrome. Crit Care. 2021 Jul 16;25(1):250.\u003c/li\u003e\n\u003cli\u003eDemoule A, Brochard L, Dres M, Heunks L, Jubran A, Laghi F, Mekontso-Dessap A, Nava S, Ouanes-Besbes L, Pe\u0026ntilde;uelas O, Piquilloud L, Vassilakopoulos T, Mancebo J. How to ventilate obstructive and asthmatic patients. Intensive Care Med. 2020 Dec;46(12):2436-2449.\u003c/li\u003e\n\u003cli\u003eHessels L, Coulson TG, Seevanayagam S, Young P, Pilcher D, Marhoon N, Bellomo R. Development and Validation of a Score to Identify Cardiac Surgery Patients at High Risk of Prolonged Mechanical Ventilation. J Cardiothorac Vasc Anesth. 2019 Oct;33(10):2709-2716.\u003c/li\u003e\n\u003cli\u003eSim I. Two ways of knowing: big data and evidence-based medicine[J]． Ann Intern Med， 2016，164(8)： 562-563\u003c/li\u003e\n\u003cli\u003eItchhaporia D. Artificial intelligence in cardiology. Trends Cardiovasc Med. 2022 Jan;32(1):34-41.\u003c/li\u003e\n\u003cli\u003eNashef SAM, Ali J. Artificial intelligence and cardiac surgery risk assessment. Eur J Cardiothorac Surg. 2023 Jun 1;63(6):ezad226.\u003c/li\u003e\n\u003cli\u003eYang L, Wu H, Jin X, et al. Study of cardiovascular disease prediction model based on random forest in eastern China. Sci Rep. 2020 Mar 23;10(1):5245. doi: 10.1038/s41598-020-62133-5.\u003c/li\u003e\n\u003cli\u003eHou N, Li M, He L, et al. Predicting 30-days mortality for MIMIC-III patients with sepsis-3: a machine learning approach using XGboost. J Transl Med. 2020 Dec 7;18(1):462. doi: 10.1186/s12967-020-02620-5.\u003c/li\u003e\n\u003cli\u003eQian Q, Wu J, Wang J, et al. Prediction Models for AKI in ICU: A Comparative Study. Int J Gen Med. 2021 Feb 25;14:623-632. doi: 10.2147/IJGM.S289671.\u003c/li\u003e\n\u003cli\u003eMutasa S, Sun S, Ha R. Understanding artificial intelligence based radiology studies: What is overfitting? Clin Imaging. 2020 Sep;65:96-99.\u003c/li\u003e\n\u003cli\u003eAksoy R, Karakoc AZ, Cevirme D, Elibol A, Yigit F, Yilmaz \u0026Uuml;, Rabus MB. Predictive Factors of Prolonged Ventilation Following Cardiac Surgery with Cardiopulmonary Bypass. Braz J Cardiovasc Surg. 2021 Dec 3;36(6):780-787. \u003c/li\u003e\n\u003cli\u003eMichaud L, Dureau P, Kerleroux B, Charfeddine A, Regan M, Constantin JM, Leprince P, Bougl\u0026eacute; A. Development and Validation of a Predictive Score for Prolonged Mechanical Ventilation After Cardiac Surgery. J Cardiothorac Vasc Anesth. 2022 Mar;36(3):825-832.\u003c/li\u003e\n\u003cli\u003eO\u0026apos;Brien Z, Bellomo R, Williams-Spence J, Reid CM, Coulson T. Development and Validation of Scores to Predict Prolonged Mechanical Ventilation after Cardiac Surgery. J Cardiothorac Vasc Anesth. 2024 Feb;38(2):430-436.\u003c/li\u003e\n\u003cli\u003eSantangelo E, Mongodi S, Bouhemad B, Mojoli F. The weaning from mechanical ventilation: a comprehensive ultrasound approach. Curr Opin Crit Care. 2022 Jun 1;28(3):322-330.\u003c/li\u003e\n\u003cli\u003eMaisat W, Yuki K. Predictive Factors for Postoperative Intensive Care Unit Admission and Mechanical Ventilation After Cardiac Catheterization for Pediatric Pulmonary Vein Stenosis. J Cardiothorac Vasc Anesth. 2022 Aug;36(8 Pt A):2500-2508.\u003c/li\u003e\n\u003cli\u003eSanfilippo F, Di Falco D, Noto A, Santonocito C, Morelli A, Bignami E, Scolletta S, Vieillard-Baron A, Astuto M. Association of weaning failure from mechanical ventilation with transthoracic echocardiography parameters: a systematic review and meta-analysis. Br J Anaesth. 2021 Jan;126(1):319-330.\u003c/li\u003e\n\u003cli\u003ePinsky MR, Bedoya A, Bihorac A, Celi L, Churpek M, Economou-Zavlanos NJ, Elbers P, Saria S, Liu V, Lyons PG, Shickel B, Toral P, Tscholl D, Clermont G. Use of artificial intelligence in critical care: opportunities and obstacles. Crit Care. 2024 Apr 8;28(1):113.\u003c/li\u003e\n\u003cli\u003eGallifant J, Zhang J, Del Pilar Arias Lopez M, Zhu T, Camporota L, Celi LA, Formenti F. Artificial intelligence for mechanical ventilation: systematic review of design, reporting standards, and bias. Br J Anaesth. 2022 Feb;128(2):343-351. \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7121733/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7121733/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground\u003c/strong\u003e: The aim of this study was to develop machine learning models to use machine learning algorithms to predict the factors associated with cardiac surgery that influence patient extubation from a ventilator after cardiac surgery. \u003cstrong\u003eMethod\u003c/strong\u003e: Clinical data of cardiac surgery patients in Nanjing First Hospital between June 2019 and December 2022 were retrospectively extracted from electronic medical records. Five traditional machine learning algorithms, including logistic regression (LR), decision tree (DT), random forest (RF), extreme gradient boosting (XGB), and light gradient boosting machine (LGB), were used to construct predictive models for risk prediction of long-distance mechanical ventilation in post-cardiovascular surgery patients. The discriminative power of these models was assessed by the area under the receiver operating characteristic curve (AUC). Shapley Additive explanation(SHAP) was used to interpret the predictive models.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults\u003c/strong\u003e: Data from 4487 patients were employed to train and validate a model of offline extubation risk in post-cardiac surgery patients. Among the full models, the RF model (AUC: 0.86; Sensitivity: 0.781, Specficity: 0.756) and the XGB model (AUC: 0.850; Sensitivity: 0.818; Specificity: 0.768) showed well predictive power for off-ventilator predicting. Eleven variables were finally selected by Boruta and LASSO features selection procedure, including age, hypertension, optime, preoperation EF, preoperation LVPW, HR, reoperation, body-weight, sex, AMBP, ST-II. Among the eleven variables, age, hypertension, operation time, preoperation EF, preoperation LVPW significantly contributed to the prediction model. \u003cstrong\u003eConclusion\u003c/strong\u003e: In this study, we successfully developed several machine learning models to predict factors affecting off-ventilator extubation after cardiac surgery, which may be useful to help clinicians assess the success of off-ventilator extubation in cardiac surgery patients after surgery.\u003c/p\u003e","manuscriptTitle":"Predicting de-ventilator extubation in post-cardiac surgery patients using machine learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-09 09:37:17","doi":"10.21203/rs.3.rs-7121733/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"decision","content":"Revision requested","date":"2026-04-24T13:46:29+00:00","index":"","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-19T16:31:07+00:00","index":"hide","fulltext":""},{"type":"editorInvitedReview","content":"","date":"2025-09-14T20:00:21+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"315749040789435033510822278550797662914","date":"2025-09-09T03:09:24+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"238185557996888859731790119352741900022","date":"2025-09-05T16:12:22+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-08-29T10:14:59+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-08-05T08:59:20+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-07-29T08:11:39+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-07-26T16:05:46+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Medical Informatics and Decision Making","date":"2025-07-26T13:39:59+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-medical-informatics-and-decision-making","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"midm","sideBox":"Learn more about [BMC Medical Informatics and Decision Making](http://bmcmedinformdecismak.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/midm/default.aspx","title":"BMC Medical Informatics and Decision Making","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a0f4fc58-5e09-4ea8-b721-8c7a0cfe8274","owner":[],"postedDate":"September 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"in-revision","subjectAreas":[],"tags":[],"updatedAt":"2026-04-24T13:56:15+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-09 09:37:17","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7121733","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7121733","identity":"rs-7121733","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}