Evaluation of scoring systems without endoscopic findings for predicting outcomes in patients with upper gastrointestinal bleeding

摘要

Abstract Background Risk scoring systems are used to evaluate patients with upper gastrointestinal bleeding (UGIB). We compared Glasgow-Blatchford score (GBS), modified GBS (mGBS), and Pre-endoscopy Rockall score (Pre-E RS) for immediate application without endoscopic findings in predicting the need of interventions and the 30-day mortality in patients with UGIB. Methods Patients who visited the emergency room with UGIB from January 2007 to June 2016 were included. GBS, mGBS, and Pre-E RS were obtained for all patients. The area under the receiver-operating characteristic curves (AUC) was used to assess the accuracy of the scoring systems to determine the need for interventions and 30-day mortality. Also, we investigated the potential cutoff scores for predicting 30-day mortality and the need for interventions. Results In predicting the need for interventions, GBS (AUC = 0.727) and mGBS (AUC = 0.733) outperformed Pre-E RS (AUC = 0.564, P < 0.0001). In predicting 30-day mortality, Pre-E RS (AUC = 0.929) outperformed GBS (AUC = 0.664, P < 0.0001) and mGBS (AUC = 0.652, P < 0.0001). Based on AUC analyses of sensitivities and specificities, the optimal cutoff mGBS and GBS for the need for interventions was 9 (70.71% sensitivity, 89.35% specificity) and 9 (73.57% sensitivity, 82.90% specificity) respectively, and optimal cutoff Pre-E RS for 30-day mortality was 4 (88.0% sensitivity, 97.52% specificity). Conclusions GBS and mGBS are considered to be moderately accurate in making an early decision about the need of interventions in patients with UGIB. Pre-E RS is considered to be highly accurate in early detection of patients at high risk for 30-day mortality without endoscopic findings. In addition, we suggested potential cutoff scores to predict the need of interventions for GBS and mGBS, and 30-day mortality for Pre-E RS. Further studies are needed to confirm the clinical applicability of results.