Multiplevariate Logistic Analysis Of Postoperative Complications And Establishment Of Risk Model In Gastric Carcinoma

Objective To investigate the main risk factors of postoperative complications in gastric carcinoma, to establish the Logistic regression model and to evaluate the accuracy, sensitivity and specificity for predicting postoperative complications. Methods Case-control study was used to retrospectively investigate the clinical data of 650 gastric carcinoma cases . Firstly, single factor analysis, including Independent-sample T Test for scale variables, Chi-square Test for binomial and nominal variables, Nonparametric Test for ordinal variables, was used in 60 clinical variables. Then, multiplevariate Logistic regression analysis was used in the selected variables. At last, the Logistic regression equation was set up:P=Exp ∑ BiXi / 1+Exp ∑ BiXi, the odds ratio was calculated :OR=Exp(B), and the accuracy, sensitivity and specificity of the model was evaluated individually. Results 18 clinical variables, including age,diameter of tumor, blood loss during operation operative time, Child-Pugh score, count of leukocyte, serum level of albumin, preoperative comorbidity, portal hypertension, total gastrectomy, lymphadenectomy for No.16a and No.16b, combined pancreatosplenectomy , cardiac disease , COPD, Case count of surgeon , degree of lymphadenectomy, were selected through single factor analysis. By entering these factors into stepwise Logistic regression analysis, it was revealed that there were eight significant factors for postoperative complications in diminishing sequence as preoperative comorbidity( OR=2.933), combined pancreatosplenectomy(OR=2.723), lymphadenectomy for No.l6a( OR=2.674) and No.l6b( OR= 2.457), total gastrectomy ( OR=1.948), Child-Pugh score (OR=1.573), blood loss during operation (OR=1.003), Case count of surgeon(OR=0.254). Logistic regression equation: P=1/1+Exp∑ (4.327 - 1.076X1- 1.002X2-0.984X3-0.899X4 - 0.667 X5 - 0.453 X6 - 0.003 X7...