Preliminary exploration of mathematical model in predicting the prognosis of Chinese people undergoing aortic surgery
10.3760/cma.j.issn.1001-4497.2015.08.008
- VernacularTitle:国人预测主动脉手术风险评估系统数学模型的初步探索
- Author:
Yipeng GE
;
Chengnan LI
;
Lei CHEN
;
Wei LIU
;
Junming ZHU
;
Lizhong SUN
- Publication Type:Journal Article
- Keywords:
Aortic dissection;
Risk assessment;
Intensive care;
Logistic models
- From:
Chinese Journal of Thoracic and Cardiovascular Surgery
2015;31(8):481-485
- CountryChina
- Language:Chinese
-
Abstract:
Objective To establish the risk scoring system in predicting prolonged intensive care unit(ICU) stay after Sun' s procedure(total aortic arch replacement with stented elephant trunk implantation) for Stanford type A aortic dissection.Accumulate experience in establishing the mathematical model in predicting the prognosis of Chinese people undergoing aortic surgery.Methods Between February 2009 and February 2012,data from 384 consecutive patients in Bejing Anzhen Hospotal,who underwent aortic surgery using Sun' s procedure,were collected retrospectively.Lengths of ICU stay longer than 7 days was defined as prolonged ICU stay.All the factors related to prolonged ICU stay were entered into univariate analysis.Then the variables with statistical difference were entered into multiple logistic analysis.The mathematical model was established based on the logistic analysis.The C-statistic was used to test discrimination of the model.Calibration was assessed with the Hosmer-Lemeshow goodness-of-fit statistic.Results The in-hospital mortality was 8.07％.The mean length of ICU stay was 3.06 days.42 patients stayed in ICU for 7 days or more.Logistic regression identified that preoperative stroke history(P =0.001,0R =9.40,regression coefficient =2.24),creatinine clearance ≤ 50 ml/min (P =0.03,OR =2.91,regression coefficient =1.07) surgical period from symptom onset shorter than 1 week (P =0.003,OR =2.89,regression coefficient =1.06),combining with coronary artery bypass grafting(P =0.03,OR =3.14,regression coefficient =1.15) were final independent risk factor for prolonged ICU stay.Simple logistic score were defined as:stroke history 22 points,creatinine clearance ≤50 ml/min 11 points,surgical period from symptom onset shorter than 1 week 11 points,combining with coronary artery bypass grafting 12 points.Additive score was defined as:stroke history 9 points,surgical period from symptom onset shorter than 1 week 3 points,creatinine clearance≤50 ml/min 11 points 3 points,combining with coronary artery bypass grafting 3 points.C statistic (receiver operating characteristic curve) for logistic algorithm was 0.72 and for additive model was 0.71.However,Hosmer Lemeshow goodness-of-fit was good (P ＞ 0.05) for logistic algorithm,was poor for additive model(P ＜ 0.05).Conclusion Using logistic regression to establish a scoring system predicting the prognosis of aortic surgery is feasible.The predicting model utilizing regression coefficient is accurate and is convenient for clinical using.Additive algorithm is not accurate and should be abandoned.
