Establishment Of The Chinese Heart Valve Surgery Database And Risk Prediction Model For In-hospital Mortality After Heart Valve Surgery

[Objective]In this study, a database and network information platform of patients undergoing heart valve surgrey are about to be set up by multi-subject cooperation.The clinical data of patients undergoing heart valve surgery at our centre from Janaury 1998 to December 2008 are about to be collected. We aim to evaluate the performance of the EuroSCORE model for predicting in-hospital mortality,and establish a risk prediction model and risk score in patients undergoing heart valve surgery at our center.[Methods]1.Establishment of the Chinese Heart Valve Surgery Database:(1)A set of data registry table for the clinical data and follow-up information in patients undergoing heart valve surgery was set up by division discussion and expert counseling,including selection, definitions, and codings of candidate predictor variables.(2)The database and network information platform of patients undergoing heart valve surgrey were set up by computer network centre of the Second Military Medical University, including data registry software,function of data anaylisis and statistics,and exclusive network platform for patients' follow-up. The clinical data of patients undergoing heart valve surgery at our centre from Janaury 1998 to December 2008 was collected and registried. After development of the function of data anaylisis and statistics,the data that had been collected were corrected seriously.2.Validation of European System for Cardiac Operative Risk Evaluation (EuroSCORE) in Chinese Heart Valve Surgery:(1)4155 patients whose major etiology were heart valve disease in the database were considered for inclusion. The endpoint of all parts in this study was in-hospital mortality.(2)Data for the EuroSCORE risk factors were obtained retrospectively from hospital medical records according to the definitions in EuroSCORE model.(3)A11 patients were scored according to the additive and logistic EuroSCORE models,and were then divided into three subgroups according to the additive EuroSCORE algorithm:low risk group; medium risk group; and high risk group. (4)Predicted mortality was compared to observed mortality for the entire cohort population and each risk subgroup. Model calibration was analysed by Hosmer—Lemeshow goodness-of-fit statistic,and model discriminationwas tested by calculating the area under the receiver operating characteristic (ROC) curve.3.Establishment of risk prediction model and risk score for in-hospital mortality after heart valve surgery:(1) The patients undergoing four types of heart valve surgery were considered for inclusion,including aortic valve replacement,mitral valve repair,mitral valve replacement, and aortic and mitral combination procedure.All patients were divided into three subgroups according to the surgery site of left atrioventricular valve:mitral valve surgery group; aortic valve surgery group; and mitral and aortic valve surgery group.Risk factors that were possibly associated with mortality were collected from the clinical variables in the database, and then the coding of candidate predictor variables were determined.(2) The data was splited into development (60%) and validation (40%) data sets, and then the risk model was developed by using a logistic regression model according to the data in development data set.(3) Predicted mortality was compared to observed mortality for the development data set, validation data set, and each subgroup. Model calibration was analysed by Hosmer—Lemeshow goodness-of-fit statistic,and model discriminationwas tested by calculating the area under the receiver operating characteristic (ROC) curve.(4) Risk score was finally set up according to the coefficientβand rank of variables in logistic regression model.[Results]1.Establishment of the Chinese Heart Valve Surgery Database:(1) Data registry table for heart valve surgery:The table is made up by clinical data and follow-up data.The clinical data table includes 13 parts which are made up by demographics information of patients, preoperative risk factors, preoperative cardiac status, preoperative inspection and laboratory test,preoperative medications, operative and extracorporeal circulation information, heart valve surgery, concomitant other cardiac procedures, post operative, complications, and prognosis. The follow-up data table includes 4 parts which are made up by time, methods, and information of follow-up.(2) Data registry software for heart valve surgery:The software is developed depending on data registry table,and the function of data anaylisis and statistics is exploited at meantime. After development of the data registry software,the clinical data of 5065 consecutive patients undergoing heart valve surgery at our centre from Janaury 1998 to December 2008 was collected and registried.(3) Network platform for follow-up:The network platform provides all kinds of connections with patients who are discharged, such as letter, telephone, message and so on.2.Validation of European System for Cardiac Operative Risk Evaluation (EuroSCORE) in Chinese Heart Valve Surgery:(1)There were significant differences in the prevalence of risk factors between the study sample and the European cardiac surgical populations.(2)Observed mortality was 4.86% overall in all the 4155 patients. The additive EuroSCORE predicted a mortality rate of 3.78% and the logistic EuroSCORE predicted a mortality rate of 3.30%. The results mean that both the additive EuroSCORE and the logistic EuroSCORE underpredicted observed mortality (Hosmer-Lemeshow:P=0.025 and P<0.001).(3)There were 981,2492 and 682 patients classified as low-, medium-and high-risk group according to the additive EuroSCORE score,and the observed mortality was 1.53%,4.86%,and 10.11%, respectively. The additive EuroSCORE model showed good calibration in predicting in-hospital mortality at high-risk group subgroup (Hosmer-Lemeshow:P=0.307),but showed poor calibration at low- and medium-risk group subgroup (Hosmer-Lemeshow:P<0.001). The logistic EuroSCORE model showed good calibration in predicting in-hospital mortality at low- and high-risk group subgroup (Hosmer-Lemeshow:P=0.879 and P=0.111), but underpredicted observed mortality at medium risk group subgroup (Hosmer-Lemeshow:P<0.001).(4)The discriminative power of both models for the entire cohort and three subgroup was poor with whose the area under the ROC curve was lower than 0.70. 3.Establishment of risk prediction model and risk score for in-hospital mortality after heart valve surgery:(1)Observed mortality was 4.74% overall in all the 4032 patients.There were 1910, 724 and 1398 patients classified as mitral valve surgery, aortic valve surgery, and mitral and aortic valve surgery subgroup according to the type of valve surgery,in which the observed mortality was 4.45%,4.42%,and 5.29%, respectively. (2)In our risk prediction model, there were 8 variables independently influenced operative mortality:chronic lung disease(OR:2.11),serum creatinine (OR:4.16), NYHAⅢ-Ⅳ(OR:2.75),critical preoperative state (OR:2.69), left ventricular ejection fraction (OR:1.55), tricuspid valve regurgitation (OR:1.33),arotic valve stenosis (OR:1.34), and concomitant CABG (OR:3.02).(3)Our risk model showed good calibration calibration and discriminative power for the development data set,validation data set, and three subgroup in which Hosmer-Lemeshow test's P value were greater than 0.05 and the area under the ROC curve were greater than 0.70.(4)Results in our risk score:chronic lung disease:3 score, serum creatinine> 110umol/l:5 score, NYHAⅢ-Ⅳ:4 score, critical preoperative state:3 score, left ventricular ejection fraction(40%-50%:2 score,30%-40%:4 score,<30%:6 score), tricuspid valve regurgitation(mild:1 score, moderate:2 score, severe:3 score), arotic valve stenosis(mild:1 score, moderate:2 score, severe:3 score) and concomitant CABG:4 score.[Conclusions]1.We have established a new database for patients undergoing heart valve surgery in China by cooperation with multicentres, which can be a solid safeguard to clinical researches in Chinese valve surgrey domain in the furture.2.The EuroSCORE model give an imprecise prediction for individual operative risk in Chinese patients undergoing heart valve surgery at our center, due to their poor calibration and discriminative power. Therefore, the use of the EuroSCORE models for risk evaluation maybe unsuitable and more caution should be exercised when using it in the future. Furthermore, creation of a new model, which accurately predicts outcomes in patients undergoing heart valve surgery for our center, is maybe required.3.We have created a new risk prediction model and risk score, which can accurately predicts outcomes in patients undergoing heart valve surgery for our center. Furthermore, our risk model can also enable benchmarking and comparisons between multicenters in a meaningful way in the furture.