Registration bias in a clinical quality register.

Background and aims The quality of medical quality registers is poorly defined and lack of trust in data due to low completeness may be a major barrier against their use in quality improvement interventions. The aim of the current observational study was to explore how selective reporting may influence adverse events registered in the Norwegian quality register for colonoscopy (Gastronet). Materials and methods Gastronet's database includes data provided by endoscopists, nurses and patients. All outpatient colonoscopies reported to Gastronet in 2015 were included and compared to the total number of colonoscopies performed in Norway as retrieved from the National Patient Registry. Hospitals were categorized into four groups according to reporting completeness < 50 %, 50 % to 69 %, 70 % to 89 % and ≥ 90 %. The number of recorded adverse events (AEs) and procedure time were analyzed. Multivariate logistic regression models were fitted to explore independent factors for selection bias. Results A total of 22,364 colonoscopies were reported to the National Patient Register of which 15,855 (71 %) were registered in Gastronet. Feedback was received from 11,079 patients (50 %). The frequency of AEs increased from 0.6 % in completeness group < 50 % to 1.6 % in completeness group ≥ 90 % ( P  < 0.001). Long colonoscopy procedure time was associated with low reporting completeness. Patient feedback was associated with older age, cecal intubation success and sedation-free colonoscopy. Conclusion Incomplete registration in a colonoscopy quality register is associated with underreporting of AEs. Longer procedure time, a surrogate marker for time constraint, is associated with low completeness.