ACUTE LOWER GASTROINTESTINAL BLEEDING: WHICH IS THE BETTER SCORE FOR EACH OUTCOME?

Aims: Introduction Although several scores have been proposed as risk stratification tools for acute lower gastrointestinal bleeding (ALGIB) there are few comparative studies among them.

Aims To assess the effectiveness of different scoring systems - NOBLADS, Oakland (OS), AIMS65, Glasgow-Blatchford (GBS) - at predicting severe bleeding, relapse, need for transfusion, need for haemostatic intervention and mortality.

Methods Retrospective review of all ALGIB admissions that required hospitalization, from January 2017 to July 2019. Clinical, demographic and laboratory data was collected. Different scores and area under the receiver operating characteristic curve (AUROC) were calculated and then compared using the Delong method.

Results 119 patients included, 54.6 % females, average age 77.0 ± 10.9 years. Most frequent aetiologies for ALGIB were ischemic colitis (39%) and diverticular bleeding (22%). Thirty-four patients (29%) presented with severe bleeding, 19 (16%) developed recurrent bleeding and 48 (41%) needed blood transfusion. Endoscopic therapeutic intervention was needed in 18 patients (15%) and 4 (3%) underwent surgery.

For predicting severe bleeding, OS (AUROC 0.90) and GBS (AUROC 0.84) showed highest discriminating power; both were statistically superior to NOBLADS and AIMS65 (respectively, p=0.002; p=0.001 and p=0.004, p=0.001). OS, GBS and NOBLADS were similar when predicting rebleeding (AUROC 0.75 vs 0.74 vs 0.71). Regarding haemostatic intervention, OS, GBS and NOBLADS were identical (AUROC 0.68 vs 0.62 vs 0.60). However, OS was statistically more accurate than GBS and NOBLADS (AUROC 0.94 vs 0.85 vs 0.79) for predicting blood transfusion needs (respectively, p=0.001 and p< 0.0001). Mortality was better predicted by AIMS65 than GBS, NOBLADS and OS (AUROC 0.87 vs 0.68 vs 0.64 vs 0.54).

Conclusions Our results favour the use of OS in predicting severe bleeding and transfusion needs, and the use of AIMS65 for predicting mortality. However, in predicting rebleeding and need for haemostatic intervention OS, NOBLADS and GBS were similar.