Multivariable Predictors of Postoperative Venous Thromboembolic Events after General and Vascular Surgery: Results from the Patient Safety in Surgery Study

Abstract

Venous thromboembolism (VTE) is a potentially preventable postoperative complication. Accurate risk prediction is an essential first step toward limiting serious, and sometimes fatal, postoperative VTE. We sought to develop and test a model to predict patients at high risk for postoperative VTE. Data from the Patient Safety in Surgery (PSS) Study were used to develop and test a predictive model of VTE using multiple logistic regression analyses. VTE occurred in 1,162 of 183,069 (0.63%) patients undergoing vascular and general surgical procedures. The 30-day mortality in patients who suffered a VTE was 11.19%. Fifteen variables independently associated with increased risk of VTE included patient factors (female gender, higher American Society of Anesthesiologists class, ventilator dependence, preoperative dyspnea, disseminated cancer, chemotherapy within 30 days, and > 4 U packed red blood cell transfusion in the 72 hours before operation), preoperative laboratory values (albumin < 3.5 mg/dL, bilirubin > 1.0 mg/dL, sodium > 145 mmol/L, and hematocrit < 38%), and operative characteristics (type of surgical procedure, emergency operation, work relative value units, and infected/contaminated wounds). These variables were used to develop a predictive model for postoperative VTE (c-index = 0.7647) and a risk score that can be used in the preoperative assessment of patients undergoing major operations. Venous thromboembolic events after noncardiac operations are relatively infrequent but highly lethal. Important multivariable risk factors for VTE in this setting were identified in the large PSS database. The risk-prediction scoring system, developed by using the logistic regression odds ratios, helps to identify patients at risk for postoperative VTE and to institute appropriate perioperative prophylactic measures.