Evaluation of Acute Terminal Ileitis in Hospitalized Patients: Development of a Predictive Model to Distinguish Crohn’s Disease from Other Etiologies

<b>Background/Objectives</b>: Terminal ileitis (TI) is often identified on CT scans in emergency settings. Diagnosing Crohn's disease (CD) as a cause of TI is crucial due to its significant long-term implications. This study aimed to differentiate CD from other causes of acute TI and develop a predictive model for CD diagnosis. <b>Methods</b>: A retrospective case-control study was conducted at Shamir Medical Center including adults diagnosed with acute TI from January 2012 to December 2020. Patients with a history of inflammatory bowel disease or prior intestinal surgery were excluded. Patients were categorized into CD and non-CD groups based on their subsequent clinical course. A logistic regression model was developed and subsequently validated with additional patients hospitalized between 2021 and 2023. <b>Results</b>: Among 135 patients, 37 (27.4%) were diagnosed with CD. CD patients were younger (median age 27 vs. 39 years, <i>p</i> = 0.003), predominantly male (83.8% vs. 51%, <i>p</i> = 0.001), and had higher rates of chronic abdominal pain, diarrhea, anemia, and weight loss prior to hospitalization. Significant laboratory differences included higher platelet counts (<i>p</i> = 0.006) and lower mean corpuscular volume (MCV) (<i>p</i> = 0.001) in CD patients. Radiologic signs of complicated disease were more common in CD (35.1% vs. 4.1%, <i>p</i> < 0.001). The predictive model incorporating gender, abdominal pain history, and MCV showed an area under the curve (AUC) of 0.87, with a sensitivity of 100% and specificity of 63.6% in the validation group of 18 patients. <b>Conclusions</b>: This study identified key predictors of CD in patients presenting with acute TI and developed a predictive model with a substantial diagnostic capability. Use of this model for early identification and treatment of CD may potentially improve patient outcomes. Further prospective validation of this model is warranted.