Aims:
LGIB is common and its incidence is increasing. There have been attempts to create new risk stratification scores to predict major clinical adverse events in LGIB, however none of are widely used. We aimed to identify risk factors associated with adverse outcomes from LGIB and develop and validate a novel scoring system.
Methods:
We retrospectively reviewed patients admitted with LGIB from three centres 2010 – 2017. Adverse outcomes recorded include blood transfusion, endoscopic intervention, CT angiography, surgery, re-bleeding and mortality. Regressional analysis within a machine learning technique identified risk factors for adverse outcomes. Area Under the Receiver Operating Curve (AUROC) were calculated and The BHAM Score was developed.
Validation of the score was conducted by retrospectively reviewing LGIB admissions in a separate centre from the original dataset. Estimates of AUROC were calculated by applying the scores directly to the whole dataset against the BHAM score and the Glasgow Blatchford Score (GBS).
Results:
A total of 473 patients were included for the original dataset (table). The BHAM score consists of: Blood pressure < 90 mmHg (1 point), Haemoglobin (< 72 g/L 14 points, 73 – 95 g/L 10 points, 96 – 117 g/L 7 points, 118 – 139 g/L 4 points), Altered mental state (2 points) and Male (1 point). A total BHAM score gives probabilities of adverse outcomes: ≥12 points ≥90%, 11 points 70%, 9 – 10 points 45%, 8 points 30%, 6 – 7 points 15%, 5 points 6% and ≤4 points < 3%. 181 patients admitted with LGIB were included in validation (table). BHAM score gives AUROC of 0.80 (95% CI 0.72 – 0.87), whilst the GBS gives AUROC of 0.76 (95% CI 0.69 – 0.84) (Figure).
Conclusions:
This validation study has shown that the BHAM score performs well at predicting adverse outcomes of LGIB. It outperforms the GBS and has the advantage of being more simple. A prospective multi-centre study is required to validate the BHAM score further before application in clinical practice.
