Background: Adenoma detection rate (ADR) is a widely touted quality measure for colonoscopy. However, there are no guidelines for minimum numbers of procedures to include to ensure reliable ADR estimates.
Objective: We sought to illustrate how a confidence interval (CI)-based approach can suggest minimum numbers for ADR calculations and provide a reasonable method for comparing ADRs and the mean number of adenomas per procedure (MAP) when relying on limited numbers of procedures.
Design: Mathematical modeling and use of real world clinical inputs.
Setting: Academic medical center.
Patients: Adults presenting for screening colonoscopy. METHODS AND MAIN OUTCOME MEASUREMENTS: We calculated 95% CIs for theoretical ADRs of 15% to 40%, with varying sample sizes, using the formula p ± 1.96√[p(1 - p)/n], where p is the ADR point estimate and n is the number of procedures. We then compared the ADRs and 95% CIs among 17 endoscopists to determine whether CIs offered important additional information. We also calculated MAPs with 95% CIs using the formula x ± 1.96(sd/√n), where x is the MAP point estimate and sd is the standard deviation of the number of adenomas detected per procedure.
Results: Large numbers of procedures (eg, 500) are needed to provide narrow CIs for typical ADR estimates. Although 10 of the 17 endoscopists had an ADR below the group's combined mean ADR of 34%, only 3 endoscopists had CIs failing to contain 34%. Likewise, whereas 9 endoscopists had MAPs below the group's combined mean MAP of 0.66, only 4 had CIs failing to contain 0.66.
Limitations: Clinical examples come from small numbers of physicians and patients.
Conclusions: Large sample sizes are required for a reliable assessment of an endoscopist's ADR. When calculating ADRs and MAPs, 95% CIs account for uncertainty and better reflect endoscopist performance.
Copyright © 2013 American Society for Gastrointestinal Endoscopy. Published by Mosby, Inc. All rights reserved.