PMO-259 Predicting failure of medical management in IBD patients- data from the 3rd round UK IBD audit

Introduction The value of admission clinical parameters and therapies to predict inpatient treatment failure is poorly defined. We used data collected in the 3rd Round UK Inflammatory Bowel Disease (IBD) Audit to determine whether number of previous hospital admissions with IBD and outpatient drug therapy predicted surgery after a failure of medical treatment. Methods Retrospective patient data from 198 UK sites were collected using an online form with up to 40 patients per site. Inclusion criteria were age >17 years, admission date from September 2010–August 2011 and a discharge diagnosis of IBD. Relevant ICD10 and OPCS codes were provided. Data items collected included number of admissions to hospital with IBD in the previous 2 years and outpatient drug therapy on admission. Our outcome was surgery due to treatment failure and a logistic regression model was fitted using both forward and backwards stepwise modelling to find independent predictors. Results There were two statistically valid models which could be used to predict surgery. In both models the number of admissions in the last 2 years was a statistically significant predictor of surgery. Two admissions doubled the risk (OR 1.97, 95% CI 1.17 to 3.32) and three or more admissions increased it further (OR 2.62, 95% CI 1.61 to 4.26). In one model the number of different drugs being taken (corticosteroids (CS), immunosuppressives (IS), anti-TNF (aTNF)) was a predictor and in the other the use of CS and the use of aTNF were both significant independent predictors instead of number of drugs. Other treatments were looked at but were not independent predictors. Taking one drug on admission was associated with a doubling of risk (OR 1.93 95% CI 1.15 to 3.24), two drugs trebled (OR 3.01 95% CI 1.78 to 5.10) and three or more quadrupled (OR 4.10 95% CI 2.08 to 8.05) the risk. Alternatively CS doubled the risk and aTNF trebled the risk with OR 1.73 and 2.96 respectively. Conclusion This data shows that the number of previous admissions to hospital can be used as an easy prognostic indicator to better inform patients of their risk of requiring surgery for a failure of medical management after admission to hospital. In addition there is relationship between the risk of surgery and taking CS and aTNF, this probably reflects disease severity, but might also be used to guide patient management. Further analysis is being carried out to find other predictors. Competing interests None declared.