The Predictive Value Of Preoperative Monitoring With Electrical Cardiometry For Postoperative Recovery In Infants With Congenital Heart Disease

Objective: Using electrical cardiometry(EC)in evaluating the preoperative cardiac function impairment in infants with ventricular septal defect(VSD).Methods: A retrospective study was conducted on infants diagnosed with VSD and admitted to the cardiac surgery department of the Capital Institute of Pediatrics from January 2017 to January 2023.Infants were grouped based on the classification of cardiac function severity,with those classified as mild(including functional class II and below)categorized as the mild impairment group and those classified as severe(functional class II and above)categorized as the severe impairment group.Hemodynamic monitoring using the ICON device(Osypka Medical Gmb H,Germany)was performed on the enrolled infants from the day of admission to the day before surgery.Cardiac index(CI),index of contractility(ICON),systolic time ratio(STR),and thoracic fluid content(TFC)were recorded.Key factors influencing the grouping of cardiac function were analyzed.Results: A total of 51 children were included in the study.Significant differences were observed between the severe and mild impairment groups in terms of weight,NT-pro BNP,baseline CI,maximum increase rate of CI,maximum increase rate of ICON,and maximum TFC value.There were no significant differences in the duration of mechanical ventilation(DMV),length of ICU stay,incidence of complications,and maximum vasoactive-inotropic score(VIS)within 24 hours.Correlation analysis showed the strongest correlation between cardiac function classification and the maximum increase rate of ICON(r=0.336,p<0.05),and asignificant positive correlation between maximum TFC value and NT-pro BNP(r=0.301,p=0.032<0.05).Logistic regression analysis indicated that NT-pro BNP≥1500 pg/ml(regression coefficient 1.517,p=0.018<0.05,odds ratio 4.560)and ICONR ≥25% were independent factors for assessing the severity of cardiac function impairment(regression coefficient for NT-pro BNP ≥1500 was 1.470,p=0.030<0.05,odds ratio 4.350).Conclusion: NT-pro BNP ≥1500 pg/ml and ICONR ≥25% are independent factors for assessing the severity of preoperative cardiac function impairment.Preoperative hemodynamic monitoring based on EC is of significant value in evaluating cardiac function.The degree of improvement in cardiac function after proactive adjustment in children with congenital heart disease can be indicated by the changing trend of hemodynamic parameters and has a certain impact on improving surgical outcomes.Objective: This study aimed to explore the association between hemodynamic parameters measured by Electrical Cardiometry(EC)and postoperative prognostic indicators in infants with ventricular septal defect(VSD)with based on principal component analysis.Methods: Infants diagnosed with VSD and admitted to the Department of Cardiac Surgery at the Capital Pediatric Institute from January 2017 to January 2023 were included in the study.EC hemodynamic monitoring was performed using the ICON device(Osypka Medical Gmb H,Germany)from the day of admission until the day before surgery.Principal component analysis was applied to produce new principal components,and a comprehensive score was calculated.Correlation analysis and multiple regression analysis were conducted to figure out the crucial factors to predicting postoperative prognostic.The diagnostic value was evaluated by ROC curves.Results: A total of 51 cases of VSD were collected.Three principal components were extracted through principal component analysis,including basic characteristics,ICON features,and TFC features,with 72.18% cumulative variance.It revealed a significant correlation between ICON features and duration of mechanical ventilation(DMV)(r=-0.470,p=0.001<0.05),as well as ICU stay duration(r=-0.417,p=0.002<0.05).All the patients were divided into the normal DMV group(DMV≤12h)and the prolonged DMV group(DMV>12 h).Meanwhile,they were also divided to to normal ICU stays group and was classified into the normal group and prolonged ICU stays group according to whether their ICU stays were over 1 daypostoperatively.Multivariate analysis revealed that the comprehensive score,basic characteristics,and ICON features were independent factors influencing prolonged DMV and ICU stay(p<0.05),while the duration of cardiopulmonary bypass(CPB)only had a significant effect on the latter.The regression coefficient for predicting prolonged DMV was-1.257,with an odds ratio(OR)of 0.284 for ICON features,and the regression coefficient for predicting prolonged ICU stay was-2.358,with an OR of 0.319 for ICON features.ROC curve analysis showed that the comprehensive score had an AUC= 0.856,with SE=0.857 and SP= 0.733 at a cutoff value of 0.002 for predicting prolonged DMV.ICON features had an AUC =0.737,with SE=0.762 and SP=0.700 at a cutoff value of 0.462 for predicting prolonged DMV.The comprehensive score for predicting prolonged ICU stays had an AUC=0.835,with SE=0.909 and SP=0.655 at a cutoff value of-0123.The ICON features for predicting prolonged ICU stays had an AUC=0.727,with SE=0.818 and SP=0.655 at a cutoff value of 0.655.Conclusion: The combined use of preoperative of hemodynamic parameters recorded by EC monitoring has significant value in predicting prolonged DMV and ICU stays.The comprehensive score,basic features,and ICON features produced by principal component analysis were identified as independent factors on the prolonged DMV and ICU stays.For each unit increase in ICON features,the risk of DMV prolongation is reduced by 0.284 times,and the risk of prolonged ICU stay is reduced by 0.319 times.Objective: Hemodynamic parameters measured by preoperative monitoring devices has a significant impact on postoperative outcomes,but traditional evaluation methods have limitations.Studies suggest that machine learning models can effectively address the complex data and machine learning-based risk predictive models have very high predictive value [1].This study aims to explore the predictive value of machine learning models analyzing hemodynamic parameters provided by Electrical Cardiometry(EC)in predicting postoperative outcomes in infants with congenital heart disease(CHD).Methods: Infants diagnosed with CHD and requiring surgery admitted to the Cardiac Surgery Department of the Capital Pediatric Research Institute from January2017 to April 2023 were included.EC hemodynamic monitoring was performed using the ICON device(Osypka Medical Gmb H,Germany)from the day of admission until the day before surgery.Using mechanical ventilation time(DMV)of 12 hours as the cutoff point,DMV within 12 hours(including 12 hours)was defined as normal,and exceeding 12 hours was defined as prolonged DMV.K-Nearest Neighbor(KNN),Neural Networks(NNs),Decision Trees,and Random Forest machine learning models were used to analyze preoperative hemodynamic monitoring data to predict whether DMV would be prolonged in CHD patients.Results: A total of 65 cases of CHD children were collected,with 27 cases in the normal mechanical ventilation duration(DMV)group and 38 cases in theprolonged DMV group.There were significant differences(p<0.05)in age,height,weight,NT-pro BNP,and respiratory rate between the two groups.All the postoperative outcome showed significant differences(p<0.05)between the normal DMV group and prolonged DMV group.The KNN algorithm demonstrated the best predictive performance,with an accuracy of 0.750,recall rate of 0.667,F1 score of0.750,and AUC of 0.788.In the random forest algorithm model,important variables contributing to the prediction were identified through feature variable importance ranking,including height,age,weight,presence of respiratory system disease,BNP≥1500pg/ml,heart function classification,maximum increase rate of ICON≥25%,composite score,and identity feature.ICON feature variables also contributed to the prediction to some extent.With important feature variables selected,the decision tree model showed the best predictive performance with an AUC of 0.822.Clinical decision guidelines were provided: Age was identified as the most important predictive factor.For children between 3 and 6 months:(1)If TFCmax<43,DMV is normal when the maximum increase rate of ICON≥25% or CI≥4.1L/min/m2;otherwise,DMV is prolonged.(2)If TFCmax≥43,DMV is normal when the maximum increase rate of CI is above 39.3%;otherwise,DMV is prolonged.Conclusion: Machine learning can effectively analyze preoperative hemodynamic parameters monitored by EC in CHD infants and has significant value in predicting postoperative recovery.Height,age,weight,presence of respiratory system disease,BNP≥1500pg/ml,preoperative heart function classification,maximum increase rate of ICON≥25%,as well as baseline features,ICON features,and composite scores from principal component analysis are important factors influencing prolonged ventilation time after surgery.The decision tree model indicates that age is the most important influencing factor,followed by TFC.Preoperative proactive inotropic therapy to increase cardiac functional reserve,improve cardiac index(CI),and ICON is beneficial for early postoperative recovery in CHD infants.Additionally,promoting diuresis to reduce TFC is also beneficial.