New scoring systems for lower gastrointestinal bleeding can predict prognosis at least as well as established scoring systems, according to a recently published validation study in a cohort in South Korea.
Although the incidence of lower gastrointestinal bleeding (LGIB) has been increasing, until recently, clinicians had largely been relying on prognostic scoring systems initially developed for upper GIB (UGIB), such as the Rockall Score (RS), the Glasgow Blatchford Score (GBS) and the AIMS65 score. However, in the past few years, novel LGIB-specific scoring systems have been developed, including the age, blood tests and comorbidities (ABC) score, the Oakland Score, and the SHA2PE score. As these new systems are yet to be externally validated, researchers at Chungnam National University Hospital, in Daejeon, South Korea, performed a validation study of the prognostic predictions of these new systems and compared the accuracy of the new and established systems.
The study included a retrospective cohort of adults who presented to the emergency department (ED) of Chungnam National University Hospital and were admitted with confirmed hematochezia between 2016 and 2021. Study participants also had to have signs of LGIB without UGIB, small bowel bleeding or varix bleeding. For each of the six scoring schemes (RS, GBS, AIMS65, ABC, Oakland and SHA2PE), predictive performance was calculated for both 30-day mortality and prolonged (≥10 days) hospital stay using the area under the receiver operating characteristic curve (AUROC).
Over the study period, of nearly 3,800 patients who visited the ED for hematochezia, 963 met the inclusion criteria. The median age of participants, 56.5% of whom were men, was 69 years. Cancer was the most common comorbidity (19.5%), followed by liver cirrhosis (12.7%), renal failure (11.2%) and heart failure (10.9%). Nearly 20% of patients were taking antiplatelet therapy and about 10% of patients were taking anticoagulation therapy at baseline.
The 30-day all-cause mortality was 3.6%, and the bleeding-related mortality was 2.3%. Study participants were in the hospital for a median of six days (range, 1-148 days).
All six scoring systems had AUROC significantly greater than 0.5 for predicting 30-day mortality (P<0.001), but the AIMS65 (0.854) and ABC (0.835) scores had significantly higher AUROC than the other systems. At the AIMS65 system cutoff value of 2.5, the 30-day mortality prediction had a sensitivity of 68.5% and a specificity 85.5%. For the ABC score, at its cutoff value of 5.5, the system predicted 30-day mortality with a sensitivity of 85.7% and a specificity of 72.0%. For predicting prolonged hospital stay, the SHA2PE scoring system was best (AUROC, 0.753; P<0.001).
In addition, the researchers evaluated whether certain patient demographic or clinical variables were associated with the risk for 30-day all-cause mortality. In their multivariable model, the researchers found that the odds of mortality were increased in patients with blood urea nitrogen of 30 mg/dL or higher (odds ratio [OR], 5.402; 95% CI, 2.11-13.81), albumin 3.0 g/dL or lower (OR, 3.91; 95% CI, 4.16-73.14), international normalized ratio greater than 1.50 (OR, 3.54; 95% CI, 1.67-7.47), or systolic blood pressure less than 100 mm Hg (OR, 3.16; 95% CI, 1.14-8.74).
Although the single-center design of this study may limit the generalizability of the results to some extent, the researchers considered their findings to illustrate the promise of LGIB scoring systems as prognostic predictors, with certain scoring systems potentially better suited for predicting mortality versus prolonged hospitalization.
—Natasha Albaneze, MPH
