Development Of Prenatal Emergency Assessment Model For Maternal And Neonatal Adverse Complications In Pregnant Women With Congenital Heart Disease

Background:In China,due to medical and health conditions,women with congenital heart disease(CHD)usually lack essential health advisories and a basic follow-up system.Due to the high-risk of cardiac complications during pregnancy,clinical management of such cases during pregnancy is difficult for clinicians.At the same time,for the cause of the maternal heart disease,neonates also have a high risk of adverse events during delivery.Prenatal emergency assessment strategy is necessary in such a situation to evaluate the incidence of maternal cardiac and neonatal complications during childbirth,and expected to provide a reference for clinical treatment of pregnant women with CHD.Methods:A retrospective analysis was performed of pregnancies in 270 women with CHD who delivered after 28 gestational weeks.Univariable and multivariable logistic regression analysis contributed to building predictive models for both mothers and neonates using the basic characteristic clinical data of the CHD patients in the perinatal period.The area under the receiver operating characteristic(ROC)curve(AUC)is utilized to detect the discrimination of the model,and the Hosmer-Lemeshow test(H-L)is used to detect the calibration of the model.Finally,the cross-validation of the support vector machine(SVM)was applied as an external validation to access the accuracy of two predictive models.Results:A total of 270 pregnant women with CHD were included in this study.The main types of CHD include:101(37.4%)patients with atrial septal defects(ASD),67(24.8%)patients with ventricular septal defect(VSD),25(9.3%)patients with the Eisenmenger’s syndrome(ES),20(7.4%)patients with tetralogy of Fallot(TOF),21(7.8%)patients with persistent ductus artriosus(PDA),and 36(13.3%)patients with other type of CHD,including Marfan syndrome,mitral regurgitation,aortic stricture,single ventricle,atrioventricular septal defect,congenitally corrected transposition of the great arteries,transposition of the great arteries and so on.In 270 pregnant women with CHD,36(13.3%)women had more than one adverse cardiac complication,of which 25(9.30%)women had heart failure(HF),18(6.7%)women had arrhythmia requiring treatment(Arr),and 11(4.1%)women had cardiac death(CD),and 2(0.7%)patients had peripartum cardiomyopathy(PPC).Seven high-risk factors were discovered in the maternal cardiac model,including:left ventricular ejection fraction(LVEF)<0.55(p=0.023),left ventricle diameter(LV)>50mm(p=0.012),platelet count(PLT)<100×109/L(p=0.002),total serum protein(TSP)<55g/L(p=0.022),pulmonary arterial systolic pressure(PASP)>40mmHg or pulmonary artery mean pressure(PAMP)>36mmHg(p=0.001),sinus tachycardia(p=0.015),ectopic heart rhythm(p=0.049).The p value of the H-L test of the maternal adverse event prediction model was 0.96,AUC=0.871(95%CI 0.808-0.934,p<0.001).External SVM cross-validation results showed that the best prediction accuracy of maternal heart model was 92.6%.In 270 pregnant women with CHD,4 patients had twin pregnancy,a total of 274 newborns.There were 77(28.10%)neonates had more than one adverse cardiac complication,of which 63(23.0%)neonates had preterm labor(PTL),54(19.7%)neonates had low birth weight(LBW),11(4.0%)neonates had small for gestational age(SGW),4(1.5%)neonates had neonatal death(ND),and 2(0.7%)fetuses had intrauterine fetal death(IFD).Five high-risk factors were discovered in the neonatal model,including:left ventricle diameter(LV)>50mm(p<0.001),percutaneous oxygen saturation(SpO2)≤90%(p=0.003),hemoglobin(HB)>140 g/L(p=0.001),PASP>40mmHg or PAMP>36mmHg(p<0.001),and gestational hypertension or preeclampsia(p<0.001).The p value of H-L test of the maternal adverse event prediction model was 0.756,AUC=0.809(95%CI 0.747-0.871,p<0.001).External SVM cross-validation results showed that the best prediction accuracy of maternal heart model was 81.5%.Conclusions:Pregnant women with CHD have a high-risk of adverse maternal cardiac complications and neonatal complications in the third trimester.Two prenatal emergency assessment models were built successfully,and the two models have excellent discrimination,calibration,and prediction accuracy.Utilizing the prenatal emergency evaluation models,obstetrics and cardiology clinicians could evaluate CHD patients after 28 weeks of gestation in a more objective way,and expected to provide a powerful tool for the clinical treatment of pregnant women with CHD.Due to the limitations of sample size and statistical methods,the prenatal emergency assessment models established by this study still require a large amount of clinical data to support validation.