Refining a conversion factor for accurately estimating the adenoma detection rate for both high and low polyp detectors

Abstract

Introduction: Determining the adenoma detection rate (ADR) is one key performance indicator for ensuring quality colonoscopy. However, it is time‐consuming and laborious to determine, as often the pathology and the endoscopy data are not electronically interfaced, despite being increasingly relevant to current high‐quality practice. Francis (2011) proposed using an adenoma to polyp detection rate quotient (APDRQ) and the individual endoscopist polyp detection rate (PDR) to estimate the ADR. This worked in that study with low PDRs; however, other studies were unable to replicate this, although they have found that proximal PDR correlated well.

Aim: Our aim was to determine whether PDR predicts ADR and if this is the same for high PDRs.

Methods: Colonoscopy records from July 2014 to September 2017 were retrospectively examined at two regional hospital endoscopy units in Australia. Patients with inflammatory bowel disease, polyposis syndromes, preoperative indications, failed examination due to inadequate bowel preparation, anastomosis, obstructing mass, and failed intubation were excluded. The Francis calculation was completed as they described and compared with the actual ADR for all endoscopists. The proposed modification to the Francis calculation was replacing the ADR target of 25% with the Proximal PDR and then calculating the APDRQ from this (APDRQ = Prox‐PDR ÷ PDR). The group average weighted APDRQ was then calculated (0.6349), and the individual's PDR was multiplied by the weighted APDRQ.

Results: A total of 10 571 colonoscopies were included. The mean number of completed colonoscopies per endoscopist was 352. The average age of patients was 58 years and 45% were male. Including all endoscopists, the average PDR for all indications was 60.6% (SD, 14.1%) and the average ADR was 39% (SD, 10.8%). Table 1 presents the PDR, Prox‐PDR, Actual ADR, Francis (2011) ADR, and modified calculated ADR. While the Francis ADR did have a strong positive correlation with actual ADR (r = 0.8461, P < 0.00001), it under‐represented the actual ADR by a mean of 13.5%. In comparison, the modified calculation also had a strong correlation (r = 0.846, P < 0.00001) and only differed from the actual ADR by a mean of 0.006%. This correlation increased as the required minimum scope numbers increased to only those with >250 (r = 0.9639, P < 0.00001). Moreover, the ADR for all endoscopists except one was within the confidence interval of modified calculation with a 3.9%.

Conclusion: It is possible to accurately estimate ADR for an individual endoscopist by using a proposed modification to the Francis calculation, by incorporating the Prox‐PDR. In this study, this result is highly correlated to their actual ADR. This may be invaluable in estimating ADR before obtaining histology confirmation. The strength of this proposed modification is that it is applicable for high and low polyp detectors.