Multivariate Logistic Regression Analysis of Postoperative Complications and Risk Model Establishment of Gastrectomy for Gastric Cancer: A Single-Center Cohort Report.
Jinzhe Zhou,Yanbing Zhou,Shougen Cao,Shikuan Li,Hao Wang,Zhaojian Niu,Dong Chen,Dongsheng Wang,Liang Lv,Jian Zhang,Yu Li,Xuelong Jiao,Xiaojie Tan,Jianli Zhang,Haibo Wang,Bingyuan Zhang,Yun Lu,Zhenqing Sun
DOI: https://doi.org/10.3109/00365521.2015.1063153
2007-01-01
Abstract:Background. Reporting of surgical complications is common, but few provide information about the severity and estimate risk factors of complications. If have, but lack of specificity. Methods. We retrospectively analyzed data on 2795 gastric cancer patients underwent surgical procedure at the Affiliated Hospital of Qingdao University between June 2007 and June 2012, established multivariate logistic regression model to predictive risk factors related to the postoperative complications according to the Clavien-Dindo classification system. Results. Twenty-four out of 86 variables were identified statistically significant in univariate logistic regression analysis, 11 significant variables entered multivariate analysis were employed to produce the risk model. Liver cirrhosis, diabetes mellitus, Child classification, invasion of neighboring organs, combined resection, introperative transfusion, Billroth II anastomosis of reconstruction, malnutrition, surgical volume of surgeons, operating time and age were independent risk factors for postoperative complications after gastrectomy. Based on logistic regression equation, p=Exp n-ary sumation BiXi / (1+Exp n-ary sumation BiXi), multivariate logistic regression predictive model that calculated the risk of postoperative morbidity was developed, p = 1/(1 + e((4.810-1.287X1-0.504X2-0.500X3-0.474X4-0.405X5-0.318X6-0.316X7-0.305X8-0.278X9-0.255X10-0.138X11))). The accuracy, sensitivity and specificity of the model to predict the postoperative complications were 86.7%, 76.2% and 88.6%, respectively. Conclusions. This risk model based on Clavien-Dindo grading severity of complications system and logistic regression analysis can predict severe morbidity specific to an individual patient's risk factors, estimate patients' risks and benefits of gastric surgery as an accurate decision-making tool and may serve as a template for the development of risk models for other surgical groups.