P070 MEDICAL THERAPY BEFORE PEDIATRIC CROHN’S DISEASE DIAGNOSIS IS ASSOCIATED WITH DIAGNOSTIC DELAY, BUT DECREASED RISK OF PERIANAL FISTULA IN ADMINISTRATIVE CLAIMS DATA

Abstract

Crohn’s disease (CD) commonly causes perianal fistulas (PF), which are difficult and costly to treat. With 10% having PF present at CD diagnosis, delay in diagnosis is associated with worse outcomes. Furthermore, >20% develop PF later, with few reliable predictors, and preventive strategies are nonexistent. We sought to characterize the variation in CD diagnosis to identify potential predictors of PF development. We hypothesize that timely diagnosis and treatment is associated with lower rate of PF development. We used OptumInsight Clinformatics Data Mart to identify patients with inflammatory bowel disease (IBD) diagnosis between 5-24 years during 2001-2014. Diagnosis required 3 CD claims; index date was 1st occurrence of any IBD claim. We required 6-month run-in period without IBD claims, and 2-years continuous follow-up after index date. IBD symptoms were defined by claims for common presenting symptoms associated with IBD. We searched for symptom claims prior to index date. PF was identified by a previously validated claims-based case definition, based on diagnosis (fistula, abscess), or procedures (seton, fistulotomy). We characterized IBD symptoms, and medication use prior to index date. Time from the 1st IBD symptom to index date, and PF development were evaluated. Of 3404 patients identified with CD, 2597 (76%) had ≥1 symptom claim within 90 days before or at index date (mean age 16.7 ± 0.2 years, 48% female, and 77% white). Median time from 1st symptom claim to index date was 110 (interquartile range [IQR]:516) days, with 9% having 0 day and 32% >1 year. Overall, 159 (6%) patients developed PF prior to or at index date, with a median 76 (IQR:413) days from 1st symptom claim to PF. Males had 2.14 (CI:1.46-3.16) greater likelihood of PF development than female. 1895 (73%) patients started medications prior to or at index date and before PF development if any, including aminosalicylates (5ASA; 38%), corticosteroids (36%), antibiotics (45%), immunomodulators (8%), and anti-tumor necrosis factor (anti-TNF; 1%). After adjusting patient characteristics, initiation of 5ASA, steroid, or immunomodulators/anti-TNF was associated with lower risk of PF development (odds ratio [OR]:0.05 (95% confidence interval [CI]:0.02-0.14), 0.29 (CI:0.16-0.52), 0.07 (CI:0.01-0.53), respectively). Initiation of 5ASA, steroid, and antibiotics were associated with 59, 102, and 286 days longer time from the 1st symptom claim to index date respectively (p<0.0001). Immunomodulators/anti-TNF were not associated with time difference from the 1st symptom claim to index date (p=0.07). This is the first report of medication use prior to pediatric CD diagnosis in a large commercial claims database. We found that pre-diagnosis therapy was associated with delay in diagnosis, but with decreased risk of PF at CD diagnosis.