The Value of Risk Scores to Predict Clinical Outcomes in Patients with Variceal and Non-Variceal Upper Gastrointestinal Bleeding

摘要

I read with interest a prospective validation study by Chang et al. [1] on the use of the AIMS65 score, Glasgow-Blatchford score (GBS), and Rockall score (RS) for the prediction of clinical outcomes in patients with acute upper gastrointestinal bleeding (AUGIB). The authors studied 337 patients who demonstrated overt signs of AUGIB. They used the area under the receiving operator characteristic (AUROC), which is a performance metric that enables clinicians to discriminate between events and nonevents in patients, to determine the utility of each score type. In general, an AUROC of <0.7 is considered a sub-optimal performance. In a single hospital, the AIMS65 score, with an AUROC of 0.747, was found to be better at predicting hospital mortality than the GBS and RS, which had AUROC values of 0.671 and 0.660, respectively. The performances of the AIMS65, GBS, and RS for predictions of the need for endoscopic intervention were only moderate, with AUROCs of 0.619, 0.645, and 0.600, respectively. Similarly, these scores were not useful for predicting recurrent bleeding, with AUROCs around 0.6. In this cohort, none of the scores performed sufficiently to predict the clinical events. The AIMS65 was useful only for predicting mortality.