Risk scores in lower gastrointestinal bleeding – have we reached the stars?

Acute lower gastrointestinal bleeding (LGIB) is a common cause of hospital admission and is associated with considerable morbidity, mortality, and health-care resource utilization [1]. The use of clinical risk scores in patient assessment may offer several benefits, including identification of patients at low risk of a poor outcome who could be suitable for outpatient management, and high risk patients at substantial risk of poor outcome who require close monitoring and hospital-based intervention. However, the clinical utility of risk scores hinges on their ability to accurately predict outcomes of high clinical relevance.

A widely used method to evaluate the discriminative ability of binary classification models is the area under the receiver operating characteristic curve (AUROC). A higher AUROC indicates better discrimination, where a value of 1.0 reflects perfect classification and 0.5 corresponds to random chance – akin to flipping a coin. Ideally, a clinically valuable risk score should have an AUROC ≥0.9, be easy to use, and be supported by clinical studies demonstrating improvements in patient outcomes, management, or resource allocation.

“However, even an accurate and well-validated mortality risk score may have limited clinical impact, as it is difficult to prove that interventions based on high risk classification improve outcomes.”

In 2017, Oakland and colleagues developed a risk score to identify LGIB patients who could safely avoid hospital admission [2]. The definition of "safe discharge" included the absence of rebleeding, transfusion, therapeutic intervention, 28-day readmission, or death. The Oakland score consists of seven variables based on patient history, physical examination, and hemoglobin level. It was derived from a prospective cohort of 2336 patients across 143 UK hospitals. A systematic review and meta-analysis including 12 independent cohorts and over 50000 LGIB patients found that the Oakland score outperformed other scores (Strate, NOBLADS, and BLEED) in predicting safe discharge (AUROC 0.86), major bleeding (AUROC 0.93), and transfusion (AUROC 0.99) [3]. Patients with an Oakland score ≤8 have a 95% likelihood of safe discharge, and the European Society of Gastrointestinal Endoscopy (ESGE) recommends using the score to guide decisions regarding outpatient management of LGIB patients [4].

Developing risk scores with AUROCs ≥0.9 for predicting individual hard outcomes such as mortality and rebleeding in LGIB remains a challenge. For these endpoints, most existing scores (BLEED, NOBLADS, Strate, GBS, AIMS65, and Oakland) have AUROCs below 0.80, limiting their practical utility [5] [6].

In 2021, the ABC score was introduced to predict mortality in patients with upper gastrointestinal bleeding (UGIB) or LGIB [6]. It includes eight variables related to age, blood tests, and co-morbidities. The score was derived from an international cohort of 3012 UGIB patients across Europe, the USA, Singapore, and New Zealand, and was validated in 4019 UGIB patients (Spain, Italy, and Israel) and 2336 LGIB patients (UK) [6]. The ABC score demonstrated superior performance in predicting mortality in LGIB patients (AUROC 0.84) compared with the AIMS65 (AUROC 0.75), GBS (AUROC 0.74), and Oakland scores (AUROC 0.69). Patients with low (≤3), medium (4–7), and high ABC scores (≥8) had in-hospital mortality rates of 0.6%, 6.3%, and 18%, respectively [6]; however, large-scale validation of these findings is still lacking. ESGE suggest that an ABC score ≤3 can be used to identify LGIB patients at low risk of death [4].

In this issue of Endoscopy, Dajti and colleagues present the development and validation of the ALIBI score for predicting in-hospital mortality in LGIB [7]. The score incorporates five variables: age, Charlson Co-morbidity Index, hemodynamic instability, in-hospital bleeding presentation, and serum creatinine. The ALIBI score was derived using logistic regression analyses on a prospective nationwide Italian cohort of 1198 consecutive LGIB patients from 15 centers. In the development cohort, the ALIBI score achieved an AUROC of 0.81. External validation was performed using a prospective multicenter cohort of 752 LGIB patients from 12 centers in Italy, Spain, France, Greece, Iran, and Brazil, yielding an AUROC of 0.79. Based on the score, patients were stratified into low (0–4 points; 1% mortality), intermediate (5–9 points; 4.6% mortality), and high risk groups (≥10 points; 19% mortality). The authors should be congratulated for their important work, which may have a significant impact on future LGIB management.

Despite these new data, we are still some distance from having optimal risk scores for LGIB that are accurate, easy to use, and supported by evidence showing that implementation does improve patient outcome or patient management. For identifying low risk patients, the Oakland score appears promising and may enable safe outpatient management. Randomized controlled trials confirming that implementation of the Oakland score safely leads to a reduced number of hospital admissions and improved resource use in real-world settings are eagerly awaited.

For evaluation of LGIB patients at high risk of poor outcome, we need much more evidence. The next step likely involves international comparisons of the performance of the ALIBI score, ABC score, and predictive models developed using artificial intelligence. However, even an accurate and well-validated mortality risk score may have limited clinical impact, as it is difficult to prove that interventions based on high risk classification improve outcomes. This challenge is underscored by the fact that most deaths in LGIB are attributable to underlying co-morbidities, while deaths directly from severe bleeding occur in fewer than 1% of patients [8].

Nevertheless, risk scores indicating a high risk of poor outcomes can still be useful for prognostic assessment, informing patients and their families, and to some degree guiding clinical decisions, such as level of monitoring and timing of colonoscopy.

Publication History

Article published online:
21 May 2025

• References

• 1 Peery AF, Crockett SD, Murphy CC. et al. Burden and cost of gastrointestinal, liver, and pancreatic diseases in the United States: Update 2021. Gastroenterology 2022; 162: 621-644
• 2 Oakland K, Jairath V, Uberoi R. et al. Derivation and validation of a novel risk score for safe discharge after acute lower gastrointestinal bleeding: a modelling study. Lancet Gastroenterol Hepatol 2017; 2: 635-643
• 3 Almaghrabi M, Gandhi M, Guizzetti L. et al. Comparison of risk scores for lower gastrointestinal bleeding: a systematic review and meta-analysis. JAMA Netw Open 2022; 5: e2214253
• 4 Triantafyllou K, Gkolfakis P, Gralnek IM. et al. Diagnosis and management of acute lower gastrointestinal bleeding: European Society of Gastrointestinal Endoscopy (ESGE) Guideline. Endoscopy 2021; 53: 850-868
• 5 Oakland K. Risk stratification in upper and upper and lower GI bleeding: Which scores should we use?. Best Pract Res Clin Gastroenterol 2019;
• 6 Laursen SB, Oakland K, Laine L. et al. ABC score: a new risk score that accurately predicts mortality in acute upper and lower gastrointestinal bleeding: an international multicentre study. Gut 2021; 70: 707-716
• 7 Dajti E, Frazzoni L, Castellet-Farrús S. et al. In-hospital mortality in patients with lower gastrointestinal bleeding: development and validation of a prediction score. Endoscopy 2025;
• 8 Oakland K, Guy R, Uberoi R. et al. Acute lower GI bleeding in the UK: patient characteristics, interventions and outcomes in the first nationwide audit. Gut 2018; 67: 654-662