Administrative Database Research Overestimates the Rate of Interval Colon Cancer

Our study reexamines the prevalence of interval colorectal cancer (I-CRC) by manually reviewing CRC cases at a single institution.

In 2% to 8% of patients with CRC, diagnosis occurs during the interval 6 to 36 months after a cancer-free colonoscopy. Rates are often determined by linking the date of colonoscopy with cancer registry information.

We examined all colonoscopies from 1993 to 2011. These examinations were linked with Pennsylvania Cancer Registry data. Matched charts were manually reviewed. We determined whether the CRC was "prevalent" or, for patients with a previous colonoscopy, whether they were interval or noninterval based on time from last colonoscopy. For interval cases, we identified "administrative errors" that could falsely increase the number of reported I-CRC.

Over the study period, 43,661 colonoscopies were performed, with 1147 (2.6%) positive for CRC after excluding cases (n=52) in which patients had IBD, previous surgery, or nonadenocarcinoma malignancy. Prevalent CRCs totaled 1062 (92.6%). Noninterval CRCs (diagnosed over 36 mo from index colonoscopy) were present in 40 (3.5%). There remained 45 (3.9%) potential I-CRC cases. However, after manual review, 21 cases were found to be administrative errors. Therefore, the accurate proportion of colonoscopies that found an I-CRC was 2.1% (95% confidence interval, 1.5%-3.2%).

The prevalence of I-CRC at our institution before adjustment was comparable with previously reported rates. This proportion was 47% lower after adjusting for administrative errors placing our figure at the lower end of reported I-CRC incidence. Reported rates of I-CRC may be falsely elevated due to errors unique to merging administrative databases.