Cure8 research brief
Improving accuracy of IBD case identification in EMRs can make research and registries more reliable, reduce misclassification, and help ensure studies and quality programs use the right patient groups.
Clinicians, IBD researchers, clinical informaticians, and teams building EMR-based cohorts or registries
This paper (abstract) describes development and validation of multiclass machine-learning models (logistic regression, random forest, XGBoost) trained on 198 EMR-derived features and manually validated cases.
The best-performing model (XGBoost with recursive feature elimination) reduced the feature set and raised the positive predictive value for identifying true IBD cases from ~65% for a single ICD-10 code to over 90%, with high specificity and sensitivity across Crohn's, ulcerative colitis, and non-IBD classes.
The authors used SHapley Additive exPlanations to improve model interpretability and then applied the model to the larger ICD-10–based cohort, proposing a scalable framework for research use.
Abstract-level summary from PubMed; the reported performance reflects the authors' dataset and validation. External validation across different health systems would be needed before broad adoption.
Review the original publication for the complete reporting, methods, and context.
Conflict statement: Declarations. Competing interests: BP is a consultant for AbbVie, Canon, Celltrion, Ironwood, Lilly, Pfizer, Takeda, Johnson & Johnson, Prometheus, and Sanofi, and a speaker for AbbVie, Janssen, Takeda, Lilly. The rest of the author list reports no disclosures. All other authors disclose no conflicts. Ethical approval: The study was conducted under approval of the Houston Methodist Institutional Review Board (IRB protocol #PRO00018644).
This Cure8 brief is based on source text from the linked article. Cure8 is informational only and is not a substitute for professional medical advice, diagnosis, or treatment.
AI could help make IBD assessments (endoscopy, histology, imaging, biomarkers) more objective and faster, which may improve monitoring, treatment selection, and early detection of complications for people with Crohn’s...
Distinguishing Crohn’s from ulcerative colitis can be hard and often requires invasive tests. A reliable blood test would speed diagnosis and help patients start the right treatments sooner. This study is an early step...
Genetic ancestry can influence IBD risk and disease course, and most prior studies focused on European populations.
