Inter- and Intrarater Reliability of the Acromioclavicular Joint Injury Classification on Anteroposterior Radiographs

Previous radiographic reliability studies of the acromioclavicular (AC) joint are from smaller and/or selective patient populations.

To investigate the inter- and intrarater reliability of the Rockwood classification of AC joint injuries on standard anteroposterior (AP) radiographs in an urban population.

Cohort study (diagnosis); Level of evidence, 2.

All suspected shoulder injuries from May 2013 to April 2014 at the Department of Orthopedic Emergency, Oslo University Hospital, were registered by admittance. Patients with anteriosuperior shoulder pain and maximum point tenderness over the AC joint were assessed. In total, 287 consective AC joint injuries were registered of the 2650 shoulder injuries in residents. From the 287 patients with AC joint injury 268 were ultimately included in the present study. On 2 occasions >6 weeks apart and blinded from each other, 2 orthopaedic consultants, 2 orthopaedic residents, and 2 musculoskeletal radiology consultants independently classified the cohort. In the interrater analyses, the raters were also stratified based on clinical experience and specialty. All raters evaluated 2 standardized nonweightbearing AP radiographs for each patient, performed by tilting the beam 15° in the cephalic and caudal direction, respectively. Reliability was assessed based on Cohen kappa (k) values.

Interrater reliability for all raters and for the stratified groups in rounds 1 and 2 was substantial for the Rockwood classification. The interrater reliability improved from round 1 to round 2 for all raters (agreement: 92% vs 94%, k = 0.66 vs 0.73, respectively), orthopaedic consultants (agreement: 90% vs 93%, k = 0.61 vs 0.72, respectively), orthopaedic residents (agreement: 92% vs 94%, k = 0.61 vs 0.76, respectively), and musculoskeletal radiology consultants (agreement: 93% vs 94%, k = 0.70 vs 0.73, respectively). No significant differences in intrarater reliability between the groups were seen for orthopaedic consultants (k = 0.72), orthopaedic residents (k = 0.73), and musculoskeletal radiology consultants (k = 0.79). One of the radiology consultants had the highest intrarater k value (k = 0.84), significantly higher than one of the orthopaedic consultants (k = 0.67) and one of the orthopaedic residents (k = 0.66). The three other raters had substantial intrarater agreement (k = 0.75 to 0.80). The raters had a strong majority agreement (≥4/6 raters) of the Rockwood classification in 82% of the patients in round 1, and in 83% in round 2. Rockwood type II had low majority agreement, with no 5- or 6-raters (only ≤4/6) agreement.

Nonweightbearing unilateral AP radiographs of AC joint injuries expressed substantial inter- and intrarater reliability. There was an improvement in interrater reliability for all groups from round 1 to round 2. One of the raters had an almost perfect intrarater reliability, and the rest had substantial reliability. The majority agreement for Rockwood type II injuries was low.