The primary objectives of this study were to: 1) identify risk factors for subsequent surgery following initial treatment of proximal humerus fractures, stratified by initial treatment type; 2) generate risk prediction tools to predict subsequent shoulder surgery following initial treatment; and 3) internally validate the discriminative ability of each tool.

We identified patients ≥ 50 years with a diagnosis of proximal humerus fracture from 2004 to 2015 using linkable health datasets in Ontario, Canada. We used procedural and fee codes within 30 days of the index fracture to classify patients into treatment groups: 1) surgical fixation; 2) shoulder replacement; and 3) conservative. We used intervention and diagnosis codes to identify all instances of complication-related subsequent shoulder surgery following initial treatment within two years post fracture. We developed logistic regression models for randomly selected two thirds of each treatment group to evaluate the association of patient, fracture, surgical, and hospital variables on the odds of subsequent shoulder surgery following initial treatment. We used regression coefficients to compute points associated with each of the variables within each category, and calculated the risk associated with each point total using the regression equation. We used the final third of each cohort to evaluate the discriminative ability of the developed risk tools (via the continuous point total and a dichotomous point cut-off value for “higher” vs. “lower” risk determined by Receiver Operating Curves) using c-statistics.

We identified 20,897 patients with proximal humerus fractures that fit our inclusion criteria for analysis, 2,414 treated with fixation, 1,065 treated with replacement, and 17,418 treated conservatively. The proportions of patients who underwent subsequent shoulder surgery within two years were 13.8%, 5.1%, and 1.3%, for fixation, replacement, and conservative groups, respectively. Predictors of reoperation following fixation included the use of a bone graft, and fixation with a nail or wire vs. a plate. The only significant predictor of reoperation following replacement was poor bone quality. The only predictor of subsequent shoulder surgery following conservative treatment was more comorbidities while patients aged 70+, and those discharged home following initial presentation (vs. admitted or transferred to another facility) had lower odds of subsequent shoulder surgery. The risk tools developed were able to discriminate between patients who did or did not undergo subsequent shoulder surgery in the derivation cohorts with c-statistics of 0.75–0.88 (continuous point total), and 0.82–0.88 (dichotomous cut-off), and 0.53–0.78 (continuous point total) and 0.51–0.79 (dichotomous cut-off) in the validation cohorts.

Our results present potential factors associated with subsequent shoulder surgery following initial treatment of proximal humerus fractures, stratified by treatment type. Our developed risk tools showed good to strong discriminative ability in both the derivation and validation cohorts for patients treated with fixation, and conservatively. This indicates that the tools may be useful for clinicians and researchers. Future research is required to develop risk tools that incorporate clinical variables such as functional demands.

Researchers and clinicians measuring outcomes following total ankle replacement (TAR) are challenged by the wide range of outcome measures used in the literature without consensus as to which are valid, reliable, and responsive in this population. This review identifies region- or joint-specific outcome measures used for evaluating TAR outcomes and synthesises evidence for their measurement properties.

A standard search strategy was conducted of electronic databases MEDLINE, EMBASE and CINAHL (to June 2015) to identify foot/ankle measures in use. A best evidence synthesis approach was taken to critically appraise measurement properties [COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN)] of identified measures. The review was restricted to English publications and excluded cross-cultural adaptations. Measurement properties collected from each article were coded for validity, reliability, responsiveness, or interpretability. Clinimetric evidence exists for identified measures tested in non-TAR populations, but were not the focus of this review.

The search identified 14 studies to include in the best evidence synthesis with 32 articles providing clinimetric evidence for eight of the measures (one CBO, seven PRO), however only five measures were tested in a TAR population (Foot Function Index, Ankle Osteoarthritis Scale, American Orthopaedic Foot and Ankle Society Ankle-Hindfoot Scale [AOFAS], Foot and Ankle Outcome Score, Self-Reported Foot and Ankle Score). Five studies provided clinimetric evidence in a TAR population and their methodological quality was assessed: (1) Validity—two good quality studies examining different measures provide moderate evidence supporting construct validity (FFI, AOS, AOFAS self-reported items; SEFAS); (2) Reliability—two good quality studies examining different measures provide moderate evidence supporting internal consistency and test-retest reliability (FFI, AOS, AOFAS self-reported items; FAOS, SEFAS); (3) Responsiveness—three poor quality studies, thus unknown evidence for responsiveness; (4) Interpretability—two studies provide interpretability values (AOS, FFI, AOFAS self-reported items; AOS).

This review offers a basis for choosing the most appropriate instrument for evaluating TAR outcomes. Numerous outcome measures were identified with evidence supporting their use in populations with various foot/ankle conditions, but none have strong evidence supporting use in a TAR population. Measures must have adequate clinimetric properties in all patient groups in which they are applied. Evidence supporting or critiquing an instrument should not be based on studies with poor quality methodology, as identified by this review. Further testing in a TAR population would benefit identified measures with emphasis on adequate sample sizes, testing a priori hypotheses, and evaluating their content validity for a TAR population.