The development and validation of a novel model for predicting surgical complications in colorectal cancer of elderly patients: Results from 1008 cases

Abstract

ObjectiveTo establish predicting models of surgical complications in elderly colorectal cancer patients. BackgroundSurgical complications are usually critical and lethal in the elderly patients. However, none of the current models are specifically designed to predict surgical complications in elderly colorectal cancer patients. MethodsDetails of 1008 cases of elderly colorectal cancer patients (age ≥ 65) were collected retrospectively from January 1998 to December 2013. Seventy-six clinicopathological variables which might affect postoperative complications in elderly patients were recorded. Multivariate stepwise logistic regression analysis was used to develop the risk model equations. The performance of the developed model was evaluated by measures of calibration (Hosmer-Lemeshow test) and discrimination (the area under the receiver-operator characteristic curve, AUC). ResultsThe AUC of our established Surgical Complication Score for Elderly Colorectal Cancer patients (SCSECC) model was 0.743 (sensitivity, 82.1%; specificity, 78.3%). There was no significant discrepancy between observed and predicted incidence rates of surgical complications (AUC, 0.820; P = .812). The Surgical Site Infection Score for Elderly Colorectal Cancer patients (SSISECC) model showed significantly better prediction power compared to the National Nosocomial Infections Surveillance index (NNIS) (AUC, 0.732; P ˂ 0.001) and Efficacy of Nosocomial Infection Control index (SENIC) (AUC; 0.686; P˂0.001) models. ConclusionsThe SCSECC and SSISECC models show good prediction power for postoperative surgical complication morbidity and surgical site infection in elderly colorectal cancer patients.