P770 Ascertainment of pediatric inflammatory bowel disease cases from administrative health data in Québec, Canada

P. Jantchou1, F. Conus2, H. Richard2, M.C. Rousseau2

1Sainte-Justine University Hospital Centre, Department of Pediatric Gastroenterology, Montréal, Canada, 2INRS-Institut Armand-Frappier- Institut national de la recherche scientifique-, Epidemiology, Laval, Canada

Background

Administrative databases are useful for estimating population-level disease occurrence. Our objective was to ascertain cases of pediatric inflammatory bowel disease (IBD) by applying two validated algorithms to administrative health data, evaluate agreement, and compare health services utilisation between concordant and discordant cases.

Methods

The Quebec Birth Cohort on Immunity Health was established through linkage of administrative databases and includes 400 611 persons born in the province of Québec (Canada) from 1970 to 1974. Physician consultations (PC) and hospitalisations (H) for IBD were documented in health databases until 2014. Two validated algorithms were used to identify pediatric IBD cases. Firstly, a single-step algorithm was applied [5PC or 2H within 4 years]. Secondly, a two-step algorithm was implemented, first considering whether the person had undergone sigmoidoscopy/colonoscopy before age 18 [yes: 4PC or 2H within 3 years; no: 7PC or 3H within 3 years]. We evaluated the agreement between both algorithms using the Kappa statistic, and compared health services utilisation among concordant and discordant cases using a t-test.

Results

The single-step algorithm generated 527 pediatric IBD cases (0.13%), whereas 480 (0.12%) were identified with the multi-step algorithm. Among the 534 cases identified by either algorithm, 473 (88.6%) were identified by both, 54 (10.1%) only by the single-step, and 7 (1.3%) only by the multi-step algorithm. Kappa was 0.94 (95% confidence interval: 0.92, 0.95), and the proportions of positive and negative agreement were respectively 0.94 and 1.00. The average number of PC and H before age 18 years among concordant and discordant cases was respectively 26.0 and 3.9 (p < 0.0001).

Conclusion

The prevalence of pediatric IBD was similar when applying two different case identification algorithms, few cases were discordant. In the near future, a survey conducted in a subset of the cohort will allow us to compare self-report with ascertainment from administrative databases.