CT Angiography in Suspected Lower Gastrointestinal Bleeding: Utilization, Yield, and Downstream Intervention in a National Emergency Department Cohort

Computed tomography angiography (CTA) is frequently conducted in the emergency department (ED) evaluation of suspected acute lower gastrointestinal (GI) bleeding, although guideline recommendations emphasize its use in patients with ongoing hemorrhage or hemodynamic instability. We evaluated multicenter CTA utilization, diagnostic yield, and downstream intervention in suspected lower GI bleeding.

We conducted a retrospective cohort study of adult ED encounters for suspected lower GI bleeding across 140 hospitals within a large United States health care system between March 31, 2024 and March 31, 2025. Patients with isolated upper GI bleeding,trauma activation, or transfer from non-system facilities were excluded. Two cohorts were defined: Group A based on chief complaint and Group B based on discharge diagnosis. Hemodynamic instability was defined as systolic blood pressure less than 90 mmHg, heart rate greater than 100 beats per minute, or packed red blood cell (PRBC) transfusion within four hours of arrival. Appropriate use was defined as CTA performed in the presence of hemodynamic instability. Positive imaging findings were identified using a natural language processing classifier applied to radiology reports. Intervention within 24 hours was defined as embolization, endoscopy, surgery, or transfusion of two or more units of PRBCs. Multivariable logistic regression was used to identify predictors.

Among 22,740 encounters, 16,642 (73.2%) underwent CTA. Appropriate use was 4.8% (95% CI 4.1-5.5) in A and 5.5% (5.0-6.1) in B. Diagnostic yield was 5.2% (4.5-5.9) in A and 5.5% (4.9-6.0) in B. Intervention occurred in 16.2% (15.0-17.4) in A and 20.0% (19.0-20.9) in B. Site-level correlation between appropriate use and diagnostic yield was weak (rho 0.25, p=0.003). Angiodysplasia was the strongest predictor of positive CTA (OR 2.19-2.34). Instability and low hemoglobin predicted intervention (OR 2.84-3.46; OR 6.43-6.76). Among patients with positive CTA, 80.7% had no recorded intervention and were less often unstable than those with recorded intervention (6.5% vs 34.8%). Secondary analysis suggests that embolization may be incompletely captured in structured data.

CTA use was high with modest diagnostic yield. Physiologic severity better predicted intervention than imaging positivity. Findings support selective, severity-guided CTA use.