PTH-015 Developing The Recorded Image Quality Index (RIQI) Tool – Measuring Recorded Image Quality, Degree of Representation and Utility. (17th August 2016)

• Record Type: Journal Article
• Authors: Samuel, D; Berrill, JB; Hurley, J; Lee, B; Hawkes, E; Yosief, S; Hawkes, N
• Is Part Of: Gut. Volume 65(2016)Supplement 1
• Page Start: A224
• Page End: A225
• Publication Date: 2016-08-17
• Subjects: Gastroenterology -- Periodicals; 616.33
• DOI: 10.1136/gutjnl-2016-312388.420
• Languages: English
• ISSNs: 0017-5749

Abstract

Abstract : Introduction: Endoscopic images saved on the Electronic Reporting System are the only visible representation of completeness of examination and pathological findings. Together with the endoscopy report these become the only reference for other clinicians not present at the original endoscopy on which to base further decisions. Given the importance of these images, we aimed to develop a systematic scoring system for quality of images recorded at endoscopy and to validate this Recorded Image Quality Index (RIQI) scoring system Methods: We searched the HICCS Endoscopic Reporting System for all colonoscopists performing regular colonoscopy (n = 11). All procedures performed between July and December 2015 were identified (screening cases were excluded). All images and the endoscopy report for the first 10 cases with pathological findings for each colonoscopist were obtained, ordered into folders and data anonymised. A RIQI scoresheet was devised assessing 4 domains (Representation (REP), Image Labelling (LAB), Caecal landmarks (CL) and Image Quality (QUAL)) and rating the utility (U) of the information set further decision-making. 110 image sets were scored by 3 independent assessors. Cohen's kappa values for intra observer variation were calculated each domain. Correlation between domain and utility scores was calculated using Cohen's kappa values for inter-rater agreement (IRA); these informed the score weighting in the final RIQI tool. Results: 110 data sets were reviewed by 3 assessors generating 330 domain scores. IRA for assessors by domain was: REP 0.53, 0.53 & 0.53 (moderate); LAB 0.82, 0.84 & 0.73 (very good); CL 0.44, 0.49 & 0.52 (moderate); QUAL 0.53, 0.39 & 0.44 (moderate). Agreement for utility scores were 0.68, 0.42 and 0.36 (good-moderate). IRA was optimal using 3 point scales (c.f. 4 or 5 point domain rating scales). REP and QUAL domains closely correlated with utility scores (r = 0.68 & 0.64) and were weighted accordingly in the final scoring system. Derived RIQI scores for each assessor correlated closely with clinical utility scores (r = 0.62, 0.63 & 0.73). Conclusion: The RIQI tool provides a method for assessing the quality of image capture across ten procedures with scores in 4 domains. The RIQI score correlates well with clinical utility of the images, with acceptable inter-rater reliability. It shows potential both as an audit and training tool to improve performance in this area of endoscopic practice. Disclosure of Interest: None Declared