BACKGROUND AND AIMS Performing an endoscopy out of hours confer significant burdens on limited health-care resources. However, not all on-call endoscopies lead to therapeutic interventions. The purpose of the present study was to analyze predictive factors for performing therapeutic intervention in patients with suspected gastrointestinal bleeding.

METHODS We reviewed and analyzed electronic medical records regarding on-call endoscopy that were prospectively collected for quality control. The subjects were patients with suspected gastrointestinal bleeding who underwent on-call endoscopies at night, on weekends and on holidays between April 2013 and January 2017 in Seoul National University Bundang Hospital. To determine predictive factors for performing therapeutic intervention, the following variables were analyzed: symptoms, patient status, coexisting disease, laboratory findings and medications. To clarify the association between the likelihood of therapeutic intervention in on-call endoscopy and AIMS65 score, the included variables were divided by cutoffs.

RESULTS A total of 270 patients (male: 72.6%, mean age: 62.6 years) with suspected gastrointestinal bleeding had on-call endoscopies and 153 (56.7%) patients had therapeutic intervention. Gastroscopy, colonoscopy and both endoscopic techniques were performed in 215, 42 and 13 patients, respectively. In the multivariate analysis, hematemesis (p < .001, odds ratio [OR], 2.484) and prolonged prothrombin time-international normalized ratio (PT-INR) (p = .033; OR, 1.958) were correlated with performing therapeutic intervention in on-call endoscopy. AIMS65 score with a cutoff of 2 was associated with the likelihood of intervention (p = .043).

CONCLUSIONS Hematemesis and prolonged PT-INR were predictive factors of therapeutic intervention when on-call endoscopy was performed in patients with suspected gastrointestinal bleeding.