Crohn Disease (CD), Dynamic Prediction Model
Conditions
Keywords
Crohn Disease (CD), dynamic prediction model, multivariate functional principal component analysis, Random survival forests, Bayesian joint models
Brief summary
This single-center retrospective cohort study aims to develop and internally validate a dynamic prediction model that uses longitudinal data from routinely collected blood tests to continuously assess the risk of Crohn's disease (CD)-related intestinal surgery, and to construct a simplified tool for clinical application. Patients with a confirmed diagnosis of CD and serial routine blood tests available during long-term follow-up will be included. Longitudinal trajectories of laboratory markers will be characterized, and their association with CD-related intestinal surgery will be evaluated. A full-variable dynamic prediction model will be built using dynamic random survival forest methodology, and a parsimonious model incorporating only the core laboratory markers will be developed via Bayesian joint model. The goal is to establish a practical, non-invasive, and dynamic risk assessment framework to support the transition from reactive to proactive long-term management of CD.
Interventions
None listed
Sponsors
Study design
Eligibility
Inclusion criteria
* Confirmed diagnosis of CD * Regular follow-up at this center for more than 5 years
Exclusion criteria
* Incomplete clinical data * Fewer than 2 blood tests during the follow-up period
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Crohn's disease-related intestinal surgery | Follow-up began at diagnosis and ended at death, loss to follow-up, or December 31, 2026, whichever occurred first. | CD-related intestinal surgeries included intestinal resection, ostomy creation, and similar procedures. The onset of a CD-related intestinal surgery was defined as the time point when it was first detected on imaging during follow-up. |