Interobserver variability using a commercially available system of archived digital radiography with integrated...

The purpose of this study was to determine if digital enhancements could improve upon published and interobserver variability for Cobb angle measurements of idiopathic scoliosis using a commercially available PACS system. The study also sought to determine if experience of the observer affected overall variability and to evaluate the time required to measure Cobb angles using 3 different techniques. As the decision for scoliosis treatment requires serial radiographic measurements by 1 or more observers at different times, precise landmark identification and curve measurement should decrease variability and improve accuracy.

Fifty-four consecutive digital radiographs of 49 children with idiopathic scoliosis were collected and archived, yielding a total of 117 curves. Five observers, ranging from a PGY2 resident to a senior level faculty member, measured each radiograph in 3 different ways. Technique A involved measuring the curves as the image first appeared on the computer screen. Technique B consisted of 2 extra steps: enlarging the image until the spine filled the screen and using an edge enhancement tool. Technique C utilized the steps in B and further enlarging each vertebra to adjust each measurement. Each technique was timed for each observer.

Technique C had the lowest variability that was significantly different from technique A. Technique B also had lower variability than technique A. The 2 observers with the greatest experience demonstrated the least intraobserver and interobserver variability. Techniques B and C decreased the variability of less experienced observers. The average time required for techniques A, B, and C was 25, 29, and 40 seconds, respectively. Confounding variables such as obesity did not affect the measurements, but curve location did, with thoracic curves causing greater variability for less experienced observers.

The results demonstrate that less experienced observers using the relatively rapid technique A for digital radiographs are more likely to have clinically significant discrepancies in their measurements, which could affect treatment decisions. Taking 4 extra seconds using technique B significantly decreases variability and improves accuracy in the evaluation and management of scoliosis patients.

I (testing of previously developed diagnostic criteria).