BMJ Open. 2026 Jan 8;16(1):e109883. doi: 10.1136/bmjopen-2025-109883.
ABSTRACT
OBJECTIVES: Accurate identification of adverse events after colonoscopy is essential for quality assurance in colorectal cancer (CRC) screening. Review of medical records is labour intensive as adverse events are infrequent. The object of this study was to investigate the accuracy of claims data in identifying adverse events after colonoscopy in CRC screening.
DESIGN: Cross-sectional, retrospective.
SETTING AND PARTICIPANTS: Males and females aged 50-74 years were randomised to once-only sigmoidoscopy or biennial faecal immunochemical test in a CRC screening trial at two screening centres in Norway. Participants in the present study underwent follow-up colonoscopy from 2012 to April 2020 after initial positive screening test. We reviewed medical records for adverse events within 30 days following 11 205 colonoscopies.
PRIMARY AND SECONDARY OUTCOME MEASURES: The primary outcome of the study was to assess the sensitivity of claims data from the Norwegian Patient Registry to identify lower gastrointestinal bleeding using emergency contact International Statistical Classification of Diseases and Related Health Problems 10th Revision diagnostic code sets under two definitions: a stringent definition (codes explicitly identifying bleeding) and a broad definition (including suggestive codes). Secondary outcome measures included the sensitivity to identify perforation using a stringent and a broad definition. Additionally, we assessed whether incorporating procedure codes and non-emergency contacts improved accuracy.
RESULTS: 87 cases of lower gastrointestinal bleeding and eight perforations were confirmed. Sensitivity for bleeding differed between the centres (p<0.001). At centre 1, sensitivity was 48.6% (95% CI 31.9% to 65.6%) using the stringent and 89.2% (95% CI 74.6% to 97.0%) using the broad definition. At centre 2, sensitivity was 36.0% (95% CI 22.9% to 50.8%) and 50.0% (95% CI 35.5% to 64.5%), respectively. Combined sensitivity for perforation was 37.5% (95% CI 8.5% to 75.5%) using the stringent and 62.5% (95% CI 24.5% to 91.5%) using the broad definition. Adding procedure codes and non-emergency contacts slightly increased sensitivity but increased false positives.
CONCLUSIONS: Use of claims data underestimated adverse event rates following colonoscopy. Difference in coding practice across hospitals underscores the need for standardised reporting in screening programmes.
TRIAL REGISTRATION NUMBER: NCT01538550.
PMID:41506765 | DOI:10.1136/bmjopen-2025-109883
