More Art than Science: Impedance Analysis Prone to Interpretation Error

• Published in: Journal of gastrointestinal surgery Vol. 19; no. 6; pp. 987 - 992
• Main Authors: Ciecierega, Thomas, Gordon, Benjamin L., Aronova, Anna, Crawford, Carl V., Zarnegar, Rasa
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.06.2015; Springer Nature B.V
• Subjects: 2014 SSAT Plenary Presentation; Acids; Adolescent; Adult; Aged; Aged, 80 and over; Cancer; Child; Child, Preschool; Diagnostic Errors; Electric Impedance; Electronic Data Processing - methods; Endoscopy; Esophagus; Female; Gastroenterology; Gastroesophageal reflux; Gastroesophageal Reflux - diagnosis; Gastroesophageal Reflux - physiopathology; Hospitals; Humans; Infant; Infant, Newborn; Male; Medical referrals; Medicine; Medicine & Public Health; Middle Aged; Motility; Patients; Physicians; ROC Curve; Software; Surgery; Young Adult; New York; United States > US; Israel; Automated analysis; Gastroesophageal reflux disease diagnosis; Impedance testing; Symptom-reflux association; Interobserver variability
• ISSN: 1091-255X; 1873-4626
• DOI: 10.1007/s11605-015-2809-0

Background Impedance monitoring for reflux evaluation does not have standardized scoring, which can confound interpretation between observers. We investigated the variability of impedance testing interpretation between physicians and computer software. Methods Raw impedance data from 38 patients that underwent impedance monitoring at a tertiary referral center between 2008 and 2013 were collected. Two physicians and computer software each analyzed the same impedance dataset for reflux activity and symptom-reflux correlation. Results Normalized reflux activity interpretations did not differ between physicians and the computer for acid or non-acid reflux. However, for weakly acidic reflux, there was significant difference between physicians ( p  < 0.01) and between physician and computer ( p  < 0.01). In analyzing all reflux, significant variability existed between physicians ( p  < 0.01) but not between physician and computer. Variability in interpretation altered diagnosis in 24 % of patients when comparing between physicians, 18 % of patients when comparing both physicians to the computer, and an additional 24 % of cases when comparing a single physician to the computer. Symptom-reflux correlation differed in 7 % of physician-physician comparisons versus 8 % of computer-physician comparisons. Conclusion Impedance testing analysis is subject to marked variability between physicians and computer software, making impedance prone to interpretation error that can lead to differences in diagnosis and management.