This study evaluated the reliability and comprehensiveness of the Unified classification system (UCPF), Wright & Cofield, Worland and Kirchhoff classifications and related treatment recommendations for periprosthetic shoulder fractures (PPSFx). Two shoulder arthroplasty specialists (experts) and two orthopaedic residents (non-experts) assessed 20 humeral-sided and five scapula-sided cases of PPSFx. We used the unweighted Cohen's Kappa (?) for measuring the intra-observer reliability and Krippendorff's alpha (a) for measuring the inter-observer reliability. The inter-rater reliabilities for the Wright & Cofield and Worland classifications were substantial for all groups. The expert and non-expert groups for UCPF also showed substantial inter-rater agreement. The all-rater group for the UCPF and the expert and non-expert group for the Kirchhoff classification revealed moderate inter-rater reliability. For the Kirchhoff classification, only fair inter-rater reliability was found for the non-expert group. Almost perfect intra-rater reliability was measured for all groups of the Wright & Cofield classification and the all-rater and expert groups of the UCPF. All groups of the Kirchhoff and Worland classifications and the group of non-experts for the UCPF had substantial intra-rater reliabilities. Regarding treatment recommendations, substantial inter-rater and moderate intra-rater reliabilities were found. Simple classification systems for PPSFx (Wright & Cofield, Worland) show the highest inter- and intra-observer reliability but lack comprehensiveness as they fail to describe scapula-sided fractures. The complex Kirchhoff classification shows limited reliability. The UCPF seems to offer an acceptable combination of comprehensiveness and reliability.
