Abstract Background Inflammatory bowel disease (IBD) is a phenotypically heterogeneous group of disorders. The Montreal classification system is the standard framework for defining IBD phenotypes, but its clinical application can be inconsistent, particularly for location and behaviour in Crohn’s disease (CD), and extent and severity in ulcerative colitis (UC) (1). Accurate phenotyping is essential for risk stratification, treatment selection, and disease monitoring. Automated phenotyping using structured electronic medical record (EMR) data offers scalability and may reduce subjective variability. However, current models are not universally validated and may not generalise across institutions (2). This study compared automated phenotypes with clinician-entered phenotypes within Crohn’s Colitis Care (CCCare), Australasia’s largest IBD-specific EMR and clinical quality registry, to identify areas for improvement in automated methods. Methods Data were extracted in November 2025. Eligible participants required both automated and clinician-entered phenotypes derived from histology, endoscopy, radiography, or surgery. Participants with missing data were excluded from analysis. Inter-rater reliability (IRR) between automated and clinician-entered phenotypes was assessed using unweighted Cohen’s kappa, with 95% confidence intervals estimated via bootstrap resampling with 2,000 iterations. Kappa coefficients were interpreted according to conventional benchmarks (3). Results There were 2,749 eligible participants included across 17 centres (mean age 44.7 years SD 23.4; 50.1% female; median disease duration 10.9 years IQR 5.9-17.8; 63% CD). For CD, IRR was substantial for endoscopy-derived location (κ = 0.76 95% CI: 0.73 – 0.79, p 0.001) and behaviour (κ = 0.66 95% CI: 0.61 – 0.70, p 0.001). Moderate reliability was observed for radiography-derived behaviour (κ = 0.47 95% CI: 0.34 – 0.60, p 0.001). All other CD phenotype features derived from radiography and surgery showed only slight IRR. For UC, IRR was substantial only for endoscopy-derived extent (κ = 0.79 95% CI: 0.75 – 0.82, p 0.001). All other UC phenotype features derived from endoscopy, radiography, and surgery showed only slight IRR. Conclusion Automated phenotyping in a large Australasian multicentre real-world IBD registry demonstrated substantial inter-rater reliability in most aspects of endoscopy-derived clinician-entered Montreal phenotypes. Reliability was lower for radiography and surgery-derived phenotypes. Further work is underway to refine and validate these models across all diagnostic modalities. Future steps include evaluating the relative accuracy of each phenotyping approach and auditing the clinician- and site-level factors contributing to variation. References: 1. Ukashi O, Amiot A, Laharie D, Menchén L, Gutiérrez A, Fernandes S, et al. Inter-Rater Disagreements in Applying the Montreal Classification for Crohn’s Disease: The Five-Nations Survey Study. United European Gastroenterol J. 2025;13(5):685-96. 2. Alzoubi H, Alzubi R, Ramzan N, West D, Al-Hadhrami T, Alazab M. A Review of Automatic Phenotyping Approaches using Electronic Health Records. Electronics. 2019;8(11):1235. 3. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics. 1977;33(1):159-74. Conflict of interest: Mr. Kwan, Trevor: No conflict of interest Wu, Rodger: No conflict of interest Wilson, William: No conflict of interest Caquilpan, Victor: No conflict of interest Chan, Patrick: No conflict of interest Rivas, Consuelo: No conflict of interest Connor, Susan Jane: Ad Boards: Abbvie, Amgen, BMS, Celltrion, Eli Lilly, Ferring, GSK, Janssen, Organon, Pfizer, TakedaSpeaker Fees: Cornerstones Health, Dr Falk, Ferring, Janssen, Sandoz, Sydney IBD School, Takeda Educational Support: DrFalk, Sandoz, Takeda Research Support: Abbvie, Agency for Clinical Innovation, Amgen, BMS, Chiesi, Celltrion, DrFalk, Ferring, Janssen, Medical Research Future Fund, Pfizer, South Western Sydney Local Health District, Sydney Partnership for Health, Research and Enterprise, Takeda and The Leona M and Harry B Helmsley Charitable Trust Andrews, Jane Mary: Speaker’s fees, research support, Ad Boards, industry consultation within last 3 years: Abbvie, BMS, Celgene, Celltrion, Falk, Ferring, Fresenius Kabi, Gilead, Janssen,J & J, MSD/Organon, Nestle, Novartis, Pfizer, Sandoz, Shire, Takeda, Vifor, Royal Adelaide Hospital Research Fund, The Hospital Research Fund 2020-2025, The Helmsley Trust 2020-2026.
Kwan et al. (Thu,) studied this question.