A Clinical Predictive Model for One-year Colectomy in Adults Hospitalized for Severe Ulcerative Colitis.

Background: Models to predict colectomy in ulcerative colitis (UC) are valuable for identification, clinical management, and follow-up of high-risk patients. Our aim was to develop a clinical predictive model based on admission data for one-year colectomy in adults hospitalized for severe UC. Methods: We performed a retrospective analysis of patients hospitalized at a tertiary academic center for management of severe UC from 1/2013 to 4/2018. Multivariate regression was performed to identify individual predictors of one-year colectomy. Outcome probabilities of colectomy based on the prognostic score were estimated using a bootstrapping technique. Results: Two hundred twenty-nine individuals were included in the final analytic cohort. Four independent variables were associated with one-year colectomy which were incorporated into a point scoring system: (+) 1 for single class biologic exposure prior to admission; (+) 2 for multiple classes of biologic exposure; (+) 1 for inpatient salvage therapy with cyclosporine or a TNF-alpha inhibitor; (+) 1 for age <40. The risk probabilities of colectomy within one year in patients assigned scores 1, 2, 3, and 4 were 9.4% (95% CI, 1.7-17.2), 33.7% (95% CI, 23.9-43.5), 58.5% (95% CI, 42.9-74.1), 75.0% (95% CI, 50.5-99.5). An assigned score of zero was a perfect predictor of no colectomy. Conclusion: Risk factors most associated with one-year colectomy for severe UC included: prior biologic exposure, need for inpatient salvage therapy, and younger age. We developed a simple scoring system using these variables to identify and stratify patients during their index hospitalization.