A Simple Predictive Model Of Prolonged Intensive Care Unit Stay After Surgery For Acquired Heart Valve Disease

Part 1Validation of EuroSCORE in Chinese Patients Operated for Acquired HeartValve DiseaseObjective To evaluate the validation of EuroSCORE in Chinesepatients operated for acquired heart valve disease. Methods Between January 2004 toJanuary 2006, 2218 consecutive patients with acquired heart valve disease wereoperated. EuroSCORE were performed to predict mortality, prolonged ICU stay andmajor postoperative complications. The patients with congenital heart valve diseaseand concomitant thoracic aortic surgery were excluded. A C statistic (receiveroperating characteristic curve) was used to test the discrimination of the models.Calibration was assessed by Hosmer-Lemeshow goodness-of-fit statistic. ResultsDiscriminating abilities of logistic EuroSCORE algorithm were, 0.71 for mortality,0.67 for prolonged ICU stay, 0.72 for respiratory failure, 0.70 for renal failure, 0.55for re-exploration for bleeding. As for additive algorithm, areas under ROC curvewere, 0.69 for mortality, 0.66 for prolonged ICU stay, 0.71 for respiratory failure, 0.74for renal failure, 0.55 for re-exploration for bleeding. Calibration of logistic andadditive algorithm in predicting mortality, prolonged ICU stay and majorpostoperative complications were not satisfactory. However, logistic algorithm couldbe used to predict postoperative respiratory (C statistic: 0.72; Hosmer-Lemeshow:p=0.20). Conclusions: EuroSCORE is not an accurate predictor in predictingmortality, prolonged ICU stay and major postoperative complications expect forrespiratory failure in Chinese patients operated for acquired heart valve disease. Part 2Validation of Parsonnet Score in Chinese Patients Operated for Acquired HeartValve DiseaseObjective To evaluate the validation of Parsonnet score in Chinesepatients operated for acquired heart valve disease. Methods Between January 2004 toJanuary 2006, 2218 consecutive patients with acquired heart valve disease wereoperated. The patients with congenital heart valve disease and concomitant thoracicaortic surgery were excluded. Parsonnet score were performed to predict mortality,prolonged ICU stay and major postoperative complications. A C statistic (receiveroperating characteristic curve) was used to test the discrimination of the models.Calibration was assessed by Hosmer-Lemeshow goodness-of-fit statistic. ResnltsDiscriminating abilities of Parsonnet score were, 0.69 for mortality, 0.63 forprolonged ICU stay, 0.65 for respiratory failure, 0.65 for renal failure, 0.63 forre-exploration for bleeding. Calibration of Parsonnet score in predicting mortality,prolonged ICU stay and major postoperative complications were poor (p＜0.001).Conclusions: Parsonnet score is not an accurate predictor in predicting mortality,prolonged ICU stay and major postoperative complications in Chinese patientsoperated for acquired heart valve disease.Part 3The Risk Factors for Prolonged Intensive Care Unit Stay After the Surgery forAcquired Heart Valve DiseaseObjective To analyze risk factors for prolonged stay in intensive careunit(ICU) after valvular surgery. Methods Between January 2004 and January 2006,2218 consecutive patients undergoing valvular surgery were studied. All the patientswere suffered from acquired heart valve disease. The patients with concomitantthoracic aortic surgery were excluded. Prolonged stay in ICU was defined as five days stay or more. The patients were divided into two groups. GroupⅠ: patients requiredprolonged ICU stay. GroupⅡ: patients did not required prolonged ICU stay.Univariate and multivariate logistic analysis were used to identify the risk factors.Results Three hundred and forty-five patients required prolonged ICU stay.Univariate risk factors showed that the proportion of previous heart surgery, strokehistory, ischemic mitral regurgitation, smoking history, pulmonary hypertension,combined with coronary artery bypass grafting, double valve surgery and re-CPBsupport, the CPB time and aortic cross-clamping time of groupⅠwere higher thanthat of groupⅡ. The NYHA function class, left ventricular ejection fraction,pulmonary function, cardio-thoracic ratio of groupⅠwere worse than that of groupⅡ. Logistic regression identified that preoperative age≥65 years (OR=3.03),pulmonary dysfunction (OR=1.88), stroke history (OR=1.75), previous heart surgery(OR=2.09), ischemic mitral regurgitation (OR=2.71), 40%＜left ventricular ejectionfraction≤0.50 (OR=2.01), left ventricular ejection fraction≤40%(OR=5.61),cardio-thoracic ratio≥0.72 (OR=2.84), 0.68≤cardio-thoracic ratio＜0.72 (OR=2.73),pulmonary hypertension (OR=1.55), combined with coronary artery bypass grafting(OR=1.77), double valve surgery(OR=1.40), repeat cardiopulmonary support(OR=4.21)and cardiopulmonary time≥185 minutes (OR=3.10) were risk factors forprolonged ICU stay after the surgery for acquired heart valve disease. ConclusionsPreoperative age≥65 years, pulmonary dysfunction, stroke history, previous heartsurgery, ischemic mitral regurgitation, left ventricular ejection fraction,cardio-thoracic rati, pulmonary hypertension, combined with coronary artery bypassgrafting, double valve surgery, repeat cardiopulmonary support and cardiopulmonarytime were risk factors for prolonged ICU stay after the surgery for acquired heartvalve disease. The patients with risk factors described above need more prei-andpostoperative care to avoid mortality, morbidity and prolonged ICU stay after valvularsurgery. Part 4A Simple Predictive Model of Prolonged Intensive Care Unit Stay after Surgeryfor Acquired Heart Valve DiseaseObjective The study aim was to construct a simple model (the Fuwairisk score) to predict prolonged intensive care unit (ICU) stay after surgery to treatacquired heart valve disease. Methods Data on 2218 consecutive patients operated foracquired heart valve disease were retrospectively collected. Prolonged stay in the ICUwas defined as 5 days stay or more. The simple logistic score was calculated bylogistic coefficient and the additive score by odds ratio. A C statistic (receiveroperating characteristic curve) was used to test the discrimination of the models.Calibration was assessed by Hosmer-Lemeshow goodness-of-fit statistic. ResultsSimple logistic model of Fuwai risk score had very good discriminating ability (Cstatistic: 0.76) and calibration (Hosmer-Lemeshow: p=0.25) in predicting prolongedICU stay, while additive algorithm had good discriminating ability (C statistic: 0.75)but poor calibration (13＜0.001). Additive algorithm greatly underestimated the risk ofhigh-risk patients. Conclusions Simple logistic algorithm of Fuwai risk score is asimple, objective and accurate score system which could be used to predict aprolonged ICU stay conveniently for the surgery of acquired heart valve diseasePart 5Risk Stratification in Valvualr Surgery: Comparison of Fuwai Risk Score,EuroSCORE and Parsonnet ScoreObjective This study aimed to compare the predictive abilities ofFuwai risk score, EuroSCORE and Parsonnet score in predicting mortality, prolongedintensive care unit stay, (ICU) and major postoperative complications for the surgeryof acquired heart valve disease. Methods Data on 2218 consecutive patients operatedfor acquired heart valve disease were retrospectively collected. Prolonged stay in theICU was defined as 5 days stay or more. Fuwai risk score, EuroSCORE and Parsonnetscore were applied to predict mortality, prolonged intensive care unit stay and major postoperative complications. A C statistic (receiver operating characteristic curve)was used to test the discrimination of the models. Calibration was assessed byHosmer-Lemeshow goodness-of-fit statistic. Results ROC area of Fuwai risk scorewas 0.62 for predicting mortality, 0.76 for prolonged ICU stay, 0.84 for respiratoryfailure, 0.70 for renal failure, 0.68 for re-exploration for bleeding. The calibration ofthe model was only acceptable in predicting prolonged ICU stay (p=0.25).Discriminating abilities of logistic EuroSCORE algorithm were, 0.71 for mortality,0.67 for prolonged ICU stay, 0.72 for respiratory failure, 0.70 for renal failure, 0.55for re-exploration for bleeding. As for additive algorithm, areas under ROC curvewere, 0.69 for mortality, 0.66 for prolonged ICU stay, 0.71 for respiratory failure,0.74 for renal failure, 0.55 for re-exploration for bleeding. Calibration of logistic andadditive EuroSCORE algorithm in predicting mortality, prolonged ICU stay andmajor postoperative complications were not satisfactory (p＜0.001). However, logisticalgorithm could be used to predict postoperative respiratory (C statistic: 0.72;Hosmer-Lemeshow: p=0.20). Discriminating abilities of Parsonnet score were, 0.69for mortality, 0.63 for prolonged ICU stay, 0.65 for respiratory failure, 0.65 for renalfailure, 0.63 for re-exploration for bleeding. Calibration of Parsonnet score inpredicting mortality, prolonged ICU stay and major postoperative complications werenot satisfactory (p＜0.001). The Discriminating abilities of Fuwai risk score wassuperior to EuroSCORE and Parsonnet score in predicting mortality, prolonged ICUstay and major postoperative complications except for renal failure. Conclusions:Althought Fuwai risk score in not a good predictor in predicting mortality and majorpostoperative complications, simple logistic algorithm of Fuwai risk score is a simple,objective and accurate score system in predicting a prolonged ICU stay for thesurgery of acquired heart valve disease. EUroSCORE and Parsonnet score are notsuperior to Fuwai risk score.