Development and validation of predictive models for Crohn's disease patients with prothrombotic state: a 6-year clinical analysis.

BACKGROUND Crohn's disease (CD) is a chronic idiopathic inflammatory disease. Studies show that multiple risk factors during disease progression can lead to a prothrombotic state (PTS), which predisposes the patient to thrombosis. Therefore, predicting PTS can help identify patients at risk of thrombosis. The aim of our study was to classify CD patients through D-dimer levels, and construct a prediction model for PTS. METHODS The clinical and laboratory data parameters were extracted from a retrospective observational cohort. The factors significantly associated with PTS were determined by univariate analysis, and the importance rankings were calculated. Two multivariate models were then constructed using these factors to predict PTS in CD using logistic regression and random forest analysis. RESULTS A total of 744 CD patients were included in the study, of which 116 were in PTS. The significant PTS-related factors were older patients, isolated colonic involvement, penetrating behavior, fever symptom, disease activity, abdominal surgery, lymphocyte counts, hematocrit levels, erythrocyte sedimentation rate, C-reactive protein, hematocrit, mean corpuscular volume levels and albumin. Multivariate logistic regression and random forest models predicted PTS with the accuracy of 89.73% and 90.63% respectively, and the corresponding AUC were 0.76 and 0.84. CONCLUSIONS Two predictive models based on clinical and laboratory variables accurately identified CD patients with PTS with high precision.