P1176 - Validation of a Logic Formula to Identify Strictures Among Crohn’s Disease Patients Using CONSTRICT Criteria

Seema Belani, MD, MS¹, Aravinda Ganapathy, MD, MS¹, David Ballard, MD¹, Juan Reyes Genere, MD¹, Richard Tsai, MD¹, Parakkal Deepak, ²
1Washington University School of Medicine in St. Louis / Barnes-Jewish Hospital, Saint Louis, MO; 2Department of Medicine, Washington University School of Medicine, St. Louis, MO

Introduction: The CONSTRICT criteria represent a consensus-based radiologic classification system for identifying small bowel strictures in Crohn’s disease (CD).¹ Widely used in clinical and research settings, the criteria help standardize reporting and guide treatment.² However, applying these criteria at scale is difficult due to variability in radiology report language and volume of imaging data.

Methods: We developed a case-finding logic formula to identify strictures meeting CONSTRICT criteria based on radiology report text. A Montage search identified 8,808 MRE and CT scans from 2015–2024 at Washington University/Barnes-Jewish Hospital and affiliate centers. The formula used inclusion/exclusion terms across three CONSTRICT criteria, flagging scans as CONSTRICT-positive if all three criteria were met (Table 1).

Results: After deduplication, 8,216 unique scans remained; 240 met all three CONSTRICT criteria using the formula. A fellowship trained abdominal radiologist re-reviewed a random sample of 20 flagged scans. The scans represented a cohort of patients that was 55% female, with a median CD disease duration of 19.5 years and predominantly stricturing (80%), ileocolonic (80%) disease phenotype (Table 2). The formula had a 100% positive predictive value (PPV) for identifying scans meeting at least two of the three CONSTRICT criteria, and a 65% PPV for identifying scans meeting all three. Extrapolated to the full dataset, this suggests that 156 of 240 flagged scans would meet the complete CONSTRICT definition.

Discussion: This logic-based approach reduced a large dataset to 240 targeted scans, supporting scalable identification of strictures without labor-intensive manual review. Limitations include: (1) the initial search targeted patients with Crohn’s disease strictures, not small bowel strictures broadly; (2) variability in radiographic descriptors (e.g., unclear bowel diameters or ambiguous terms near surgical anastomoses) may lead to false positives/negatives; (3) non-flagged scans were not re-reviewed, so negative predictive value (NPV), sensitivity, specificity and accuracy calculations could not be assessed; and (4) CONSTRICT lacks strict numeric diameter thresholds, complicating automation.
The logic formula can be embedded into natural language processing (NLP) systems to enable real-time identification of strictures at scale. Further validation in multicenter settings is needed to refine its accuracy and broader applicability.


Figure: Table 1: Constrict Case Finding Formula Logic


Figure: Table 2: Demographics for Reviewed Cohort

Disclosures:
Seema Belani indicated no relevant financial relationships.
Aravinda Ganapathy indicated no relevant financial relationships.
David Ballard: Direct Biologics – Consultant. Merck – Consultant.
Juan Reyes Genere: Boston Scientific – Consultant. Edulis Therapeutics Inc – Consultant.
Richard Tsai indicated no relevant financial relationships.
Parakkal Deepak: Arena Pharmaceuticals – Grant/Research Support, Personal or other fees. Boehringer Ingelheim – Grant/Research Support, Personal or other fees. Bristol Myers Squibb/Celgene – Grant/Research Support, Personal or other fees. Janssen – Grant/Research Support, Personal or other fees. Pfizer – Grant/Research Support, Personal or other fees. Prometheus Biosciences – Grant/Research Support, Personal or other fees. Takeda – Grant/Research Support, Personal or other fees.

Seema Belani, MD, MS¹, Aravinda Ganapathy, MD, MS¹, David Ballard, MD¹, Juan Reyes Genere, MD¹, Richard Tsai, MD¹, Parakkal Deepak, ². P1176 - Validation of a Logic Formula to Identify Strictures Among Crohn’s Disease Patients Using CONSTRICT Criteria, ACG 2025 Annual Scientific Meeting Abstracts. Phoenix, AZ: American College of Gastroenterology.