S1980 Effectiveness of Emergency Department Risk Calculators in Triaging Need for Urgent Abdominal Imaging for Patients With Crohn’s Disease

Introduction: Recent studies have identified a concerning trend of increasing abdominopelvic computed tomography (APCT) scan use in the management of Crohn’s disease (CD), particularly in emergency departments (ED). While risk calculators have been created to triage APCT usage in the ED, these tools have encountered barriers to widespread use. A recently published survey suggested that an acceptable risk calculator would need a sensitivity ≥95% and negative predictive value (NPV) ≥98%. In this study, we compared the ability of 3 published risk calculators to predict urgent findings in patients with CD in the ED. Methods: We conducted a cross-sectional study of patients with CD who underwent APCT in the ED. The cohort included patients with CD with an established gastroenterology provider who presented to the ED in a large health system with 6 hospitals. We defined urgent radiographic findings as perforations, obstructions, abscesses, or other non-CD acute findings (POANCD). Demographic and encounter characteristics for this cohort were summarized. Three risk calculators were applied to the cohort to determine their test characteristics for identifying POANCD. Results: 300 encounters were reviewed, comprising 228 unique patients. Hospital admission occurred in 187 (62.3%) encounters. Surgery occurred in 16 (5.3%) encounters, with 15 (5.0%) being intra-abdominal. APCT identified 92 (30.7%) POANCD findings. Bowel obstruction was most common, followed by abscess, non-CD acute findings, and perforation. Kerner et al.’s calculator had a 98.9% sensitivity and 96.3% NPV. Govani et al.’s calculator had a 96.7% sensitivity and 89.7% NPV. Riskin Guez et al.’s calculator had a 71.7% sensitivity and 80.3% NPV. Conclusion: The rate of POANCD findings in this study was similar to that reported by Kerner et al. (34.5%) but differed from that reported by Govani et al. (16.8%) and Riskin Guez et al. (44%), likely attributed to differences in population characteristics. When applying the risk calculators to this cohort, the calculators by Kerner et al. and Govani et al. achieved the desired ≥95% sensitivity. No calculators achieved the desired NPV ≥98%, noting that negative case numbers were too small to accurately capture the NPV. The study highlights the potential for application of Kerner et al.'s calculator to triage APCT usage while identifying opportunities for risk calculator optimization.