Derivation, validation, and comparison of a new prognostic scoring system for acute lower gastrointestinal bleeding

Abstract

AbstractObjectivesLower gastrointestinal bleeding is a common presentation with little data concerning risk factors for adverse outcomes. The aim was to derive and validate a scoring system to stratify risk in lower gastrointestinal bleeding and compare it to the Oakland score.MethodsA total of 2385 consecutive patients (mean age 65 years, 1140 males) were used to derive the score using multivariate logistic regression modeling then internally and externally validated. The Oakland score was applied and area under receiver operating characteristic (AUROC) curves were calculated and compared. A score of <1 was compared with an Oakland score of <9 to assess 30‐day rebleeding and mortality rates.ResultsRebleeding was associated with age, inpatient bleeding, syncope, malignancy, tachycardia, hypotension, lower hemoglobin and mortality with age, inpatient bleeding, liver/gastrointestinal disease, tachycardia, and hypotension. The area under the receiver operating characteristic curves was 0.742 for rebleeding and 0.802 for mortality. A score <1 was associated with rebleeding (0.0%–2.2%) and mortality (0%). The Oakland score had a significantly lower area under the receiver operating characteristic curve for rebleeding of 0.687 but not for mortality; 0.757. A score <1 was associated with a lower 30‐day rebleeding risk compared to an Oakland score <9 (4/379 vs. 15/355, p = 0.009) but not mortality (0/365 vs. 1/355, p = 0.493).ConclusionsOur score predicts 30‐day rebleeding and mortality rate with low scores associated with very low risk. The Aberdeen score is superior to the Oakland score for predicting rebleeding. Prospective evaluation of both scores is required.