Inter-observer reliability in the interpretation of coronary angiograms.

Abstract

The interpretation of coronary angiograms is indispensable in determining procedure in coronary surgery. The aim of this study was to measure the overall reliability of a group of surgeons in the interpretation of coronary angiograms, surgical procedure and the evaluation of operative risk. Ten coronary angiograms were interpreted by eight cardiac surgeons at four different medical centers. Evaluation of coding discrepancies, in this case of multiple raters applying an ordinal-scale classification scheme (0, 1, 2) with no expert yardstick available for coding, was explored by a two-way random factor analysis of variance. Reliability was substantial for the assessment of stenosis irrespective of the artery (intraclass correlation coefficient (ICC) ranging from 0.92 to 1), and good for the distal part of the artery (ICC ranging from 0.83 to 0.86) as well as for the collateral provision (ICC ranging from 0.75 to 0.94). Agreement between surgeons was good with respect to the number of bypasses to be performed (ICC = 0.88). The number of bypass per patient varied from 2.6 to 3.2 depending on the surgeon. Agreement as to whether or not to bypass was substantial for the right coronary artery (ICC = 0.92), good for the marginal artery (ICC = 0.87) and fair for the left anterior descending artery (ICC = 0.60) and the circumflex artery (ICC = 0.60). There was a higher rate of agreement concerning inferior wall motion (ICC = 0.98) than of the anterior wall motion (ICC = 0.78). Agreement was substantial for ejection fraction (ICC = 0.93), operative risk (ICC = 0.93) and the type of coronary tree (ICC = 0.85). With respect to the overall set of items, no one surgeon disagreed significantly with the rest of the group. Some disagreement regarding anatomy suitable for revascularization exists between surgeons. Surgical assessment of risk is similar. Cardiac surgeons quickly learn to assess risk in a similar manner, even though they might not always graft the same anatomic vessels or assess regional wall motion similarly.