Reliability of measurements of the fractured clavicle: a systematic review
BACKGROUND A relative indication for surgical treatment of midshaft clavicle fractures is shortening ≥2.0 cm. A standard method for determining shortening with routine clavicle radiographs has not been established. This study evaluated the interobserver and intraobserver reliability when measuring shortening of midshaft clavicle fractures in adolescents. METHODS We identified all clavicle radiographs of simple midshaft clavicle fractures in adolescents from 2006 to 2010. Thirty-two radiographs were chosen following a power analysis for 7 observers. Each film was measured twice by each evaluator using 2 separate methods. Method 1 was the evaluator's method of choice to determine shortening on the digital radiographs. Method 2 was standardized. Intraclass correlation coefficient and confidence intervals (CI) were calculated to determine interrater reliability, and average differences between the 2 time points with 95% CI were calculated to determine intrarater reliability. RESULTS Interrater reliability for method 1 was 0.771 (95% CI, 0.655-0.865) and 0.743 (95% CI, 0.604-0.851) at the 2 time points for fair agreement. Interrater reliability for method 2 was 0.741 (95% CI, 0.629-0.842) and 0.685 (95% CI, 0.554-0.805) at the 2 time points, for fair and poor agreement, respectively. Neither method was statistically superior to the other. For method 1, the SD for the measurements averaged 3.1 mm. For method 2, the average SD was 3.0 mm. Intrarater reliability for method 1 was 2.62 mm average difference between the 2 time points (95% CI, 2.24-3.00), and for method 2 it was 3.34 mm average (95% CI, 2.88-3.80). Method 2 had a significantly greater difference at the 2 time points than method 1 (P=0.027). CONCLUSIONS There is only fair agreement among observers when measuring the shortening of clavicle fractures in adolescents on digital clavicle radiographs by either method described. However, as the average difference among measurers was only 3 mm, this is unlikely to influence clinical decision making. A lack of standardization of measurement in previous studies on clavicle fracture treatment may not represent a significant problem. LEVEL OF EVIDENCE Level III diagnostic study.