Reliability of radiographic assessment of acromial morphology

The most widely used radiographic classification system for acromial morphology identifies three distinct acromial shapes: type I (flat), type II (curved), and type III (hooked). The purpose of this study was to measure the interobserver and intraobserver reliability of determinations of acromial morphology as defined by this system. Between 1990 and 1992, one hundred twenty-six supraspinatus outlet radiographs were obtained from 126 patients by technicians from Triangle Orthopaedic Associates in Durham, N.C. Six fellowship-trained shoulder surgeons independently reviewed each radiograph and classified it as type I, II, or III on the basis of established guidelines. Two surgeons classified each film a second time in random order. Analysis of variance was performed to obtain coefficients for interobserver and intraobserver reliability. Consensus ratings were then used to classify the 126 radiographs into consensus type I, consensus type II, or consensus type III groups. Percentages of type I, II, and III individual ratings within each consensus group were determined. The intraobserver reliability coefficient was 0.888, interpreted as good to excellent reliability. The interobserver reliability coefficient was 0.516, interpreted as poor to fair reliability. Of the 126 radiographs, 26 (20.6%) were rated as consensus type I, 76 (60.3%) were rated as consensus type II, and 24 (19.1%) were rated as consensus type III. The reliability of observer ratings was lowest when delineation between acromial types II and III was required. The low interobserver reliability makes comparisons of studies by different authors difficult to interpret and obscures the true incidence of acromial morphologic types. It also questions reported correlations between acromial type and shoulder pathologic conditions. It is concluded that a system that incorporates more objective classification criteria and acknowledges the continuous nature of acromial morphologic types may improve interobserver reliability and validate the system's use in making clinical and surgical judgments.