Assessing Quality of Colonoscopy

Abstract

Purpose: At present, Withdrawal Time (WT) during Colonoscopy (C) is used as an indicator of quality, irrespective of whether clear or blurred images are seen. Moreover, the start of the WT phase is subjective and determined by the endoscopist making true inter-endoscopist and -institutional comparisons impossible. Aims: To develop a completely automated method to measure WT and to assess the fraction of the WT that consists of clear images. Methods: We have developed a completely automated, real-time image analysis system that detects when an endoscope is inside a patient and records the entire procedure as a digital video file. Using this system we have created a database of about 3000 anonymous endoscopies at Mayo Clinic Rochester (MCR). 75 video files obtained from two rooms over 4 consecutive days in 2007 were randomly selected from the database. 50 files represented Cs in an intact colon; from these 10 (3 smallest, 4 median and 3 largest) were selected based on size for review by a panel of 6 colonoscopists at MCR (MD). A review panel was created to document video frames marking (a) entry into cecum, (b) maximal extent and (c) extubation of colon. Algorithms were developed to determine maximal extent of intubation based on cessation of forward movement and clear or blurred state of the digital video image (Machine). Results: The MDs and Machine agreed on timing of extubation for all patients. For timing of “entry into cecum” and “maximal extent”, the 6 MDs and Machine had perfect agreement for 4/10 patients. In the remaining 6 patients, although MDs disagreed about one or both of these time points, Machine placed “maximal extent” between or at one of the MDs' coding of “entry into cecum” and “maximal extent” time points. The WT for the 10 Cs was 10.8 ± 10.5 and 7.0 ± 10.4 Min (Mean ± SD, range 2.8–36.4 and 1.7–35.5) for MDs using these time points respectively as the start of WT. Mean WT for Machine was 10.0 ± 9.6 Min (range 2.5–34.6). When blurry frames were removed, mean Machine WT was 5.1 ± 4.9 Min (range 0.8–13.3); 6/10 Cs had a Machine clear WT below 3 Min. Conclusion: (1) Algorithm-based, automated quality assessment of C is a novel, promising quality control method. (2) MD and Machine come to similar WT conclusions when human reviewers are in complete agreement regarding “entry of cecum” and “maximal extent” time points. (3) When human reviewers do not come to complete agreement regarding insertion time points, Machine always provides a time point that is similar to some of the reviewers. (4) WT as currently documented at MCR does not reflect the actual time that clear images of the mucosa are seen and provides a false sense of quality. Large studies that correlate Machine-based measures of quality with patient outcomes are needed.