NEW HAVEN SURVEY OF JOINT DISEASES. XI: OBSERVER VARIABILITY IN THE ASSESSMENT OF X-RAYS FOR OSTEOARTHROSIS OF THE HANDS

Wright, E. C. and R. M. Acheson (Dept. Epidemiology and Public Health, Yale Univ. School of Medicine, New Haven, Conn. 06510). New Haven survey of joint diseases. Xl: Observer variability in the assessment of x-rays for osteoarthrosis of the hands. Amer. J. Epid., 1970, 97: 378–392.—-X-rays of the hand from the New Haven survey of joint diseases were read independently by three observers, two of whom also read a set of films twice. There were considerable differences between the readings of the three observers: the percentage of diseased joints varied from 6% to 15% and all three readers agreed on only 68% of the joints. (The pairwise agreements were 74%, 70% and 87%). The replicate readings of the two observers who read the films twice showed less variability. The reader upon whose readings the survey results are based considered 18.8% of the joints to be diseased the first time and 18.2% the second time. His second reading agreed with his first in 80% of the joints and was within one grade of his first reading in 98% of the joints. It was found that the agreement of the observers depended on the proportion of normal joints in the sample and on the frequency with which each observer detected very early disease. An analysis of the agreement between observers for each grade showed that the grade 1 (doubtful disease) caused the most difficulty, with the grades 2 (minimal) and 3 (moderate) causing more difficulty than either 0 (normal) or 4 (severe). It was felt that the reliance of the grading system on several parallel but independent manifestations of the disease might contribute to this disagreement.