P1100 A multivariate time-dependent proportional hazard model to predict development of intestinal failure from Crohn’s Disease

Abstract

Abstract Background Intestinal Failure (IF) is a rare but severe complication of Crohn’s disease (CD). Advancement in therapy has not eliminated incident IF from CD (CDIF), and risk factors for developing IF from CD are not well characterised. The Copenhagen Intestinal Failure Database contains one of the largest, comprehensively described adult CDIF cohort1 from 1973 to 2018, which can be linked with a Danish nationwide CD cohort2 to explore healthcare utilisation patterns associated with developing IF from onset of CD. Methods CDIF patients from Copenhagen Intestinal Failure Database and a nationwide CD cohort without IF were followed from CD diagnosis until IF, death, or censoring on 31/12/2018. Hospital contacts, surgeries, medication use and employment status were extracted from multiple Danish population-based registries. Unplanned hospitalisation or emergency department attendance (inpatient-contact), CD-related abdominal surgery, and stopping work for 6 months or more (in working-age CD patients, excluding full-time study and parental leave) were considered time-dependent covariates and tested for significance (p<0.05) in univariate proportional hazard models. Significant covariates were then modelled together with additional non-time-dependent clinical and demographic variables in multivariate proportional hazard models. Backward variable selection was performed to derive a final predictive model. Results 175 CDIF and 22845 CD patients from whom data was available were followed for 2283.1 and 293536.5 patient-years, respectively. Univariate time-dependent proportional hazard models showed that every instance of unplanned IBD inpatient contact, CD-related abdominal surgery, and stopping work was significantly associated with increased hazard of developing IF, with a hazard ratio of 4.91 (95% CI 3.54-6.80, p<0.001), 29.79 (95% CI 9.70-91.48, p<0.001) and 1.61 (95% CI 1.15-2.26, p=0.005). In multivariate analysis, earlier decade of IBD onset, unplanned IBD inpatient contacts, length of stay of the inpatient contact, CD-related abdominal surgery, complicated surgery, number of previous admissions and surgeries and previous median-long term unemployment increased the rate of developing IF (Table), whereas age and comorbidity level at CD onset, non-IBD inpatient contacts, type of surgery, biologics use, and current employment status were not found to be significantly associated with IF development. Conclusion This is the first study to demonstrate significantly increased risks of developing IF from CD from increased healthcare utilisation patterns.