Pitfalls and Perils of Using Administrative Databases to Evaluate the Incidence of Inflammatory Bowel Disease Overtime

T he inflammatory bowel diseases (IBD), consisting of Crohn’s disease and ulcerative colitis, affect approximately 0.6% of North Americans. A systematic review of population-based studies on the incidence of IBD suggests that IBD has become a global disease. The systematic review also demonstrated that since 1980s, the incidence of IBD significantly decreased in only 6% of ulcerative colitis studies and in none of the Crohn’s disease studies. Thus, the literature suggests that the incidence of IBD has predominantly been rising or stable in regions throughout the world. In this issue of Inflammatory Bowel Disease, 2 articles evaluated the incidence of IBD in 2 provinces of Canada. In the first article, Bitton et al used administrative health databases from Quebec (hereafter called the Quebec IBD Cohort) to demonstrate that the incidence of Crohn’s disease and ulcerative colitis decreased significantly from 2001 to 2008. In contrast, Benchimol et al used administrative health databases from Ontario (hereafter called the Ontario IBD Cohort) to demonstrate that the incidence of IBD increased in children and adults younger than 64 years from 1999 to 2008. The divergent results may be explained by fundamental differences between Ontario and Quebec patients with IBD including differences in: genetic susceptibility; environmental exposures such as diet and smoking; socioeconomic factors; and practice patterns of health care providers and health care delivery. Alternatively, methodological differences between the study designs may explain these opposing results. At the surface, the methodology used by both of these studies seems relatively similar. Both studies used linked administrative health databases, overlapping time periods and follow-up, previously validated definitions of IBD, and washout periods to define an inception cohort. However, a deeper evaluation of the methodological approaches highlights intrinsic differences in study design that likely explain the differential findings in incidence overtime. This editorial will explore the advantages and disadvantages of using administrative health databases for IBD epidemiology research, identify differences in study designs between these 2 articles, and recommend standardized reporting of IBD epidemiology studies using administrative databases. ADVANTAGES OF ADMINISTRATIVE DATABASES IN STUDYING IBD Administrative databases are timeand cost-efficient resources that have been widely used to study IBD, e.g., evaluating incidence, health care resource utilization, surgical outcomes, environmental risk factors, and pharmacoepidemiology. Administrative databases, like the ones used in Ontario and Quebec, allow linkage across databases and thus, patients with IBD can be followed seamlessly between outpatient and inpatient management. Furthermore, many of these databases have existed for decades, which allow researchers to track large cohorts of patients with IBD throughout long follow-up periods. These data sets allow for the evaluation of uncommon outcomes (e.g., mortality), time-specific outcomes (e.g., cumulative risk of surgery), and time-evolving outcomes (e.g., change in the incidence overtime). The data are recorded independently, which minimize selection biases and maximize external validity. Despite these advantages, administrative databases have inherent challenges that can bias the interpretation of findings.