JSES - 2026-10-12 - Journal Article
Predictive Modeling of Clinically Apparent Shoulder Periprosthetic Joint Infection (PJI).
Hsu JE, Ricchetti ET, Tashjian R, Namdari S, Collins AP, Fedorka C, Kim HM, Zhang C, Presson AP, Whiston AJ, American Shoulder and Elbow Surgeons (ASES) Revision Arthroplasty / Periprosthetic Joint Infection (PJI) Multicenter Group
Topics
Key Takeaway
A seven-variable logistic regression nomogram predicted clinically apparent shoulder PJI (sinus tract or intra-articular purulence) with an apparent concordance statistic of 0.90 in a prospective multicenter cohort of 861 revision shoulder arthroplasties.
Summary Depth
Choose how much analysis to show on this article page.
Summary
This study used prospective multicenter data from the ASES Revision Shoulder Arthroplasty and PJI Research Group to develop a predictive model for clinically apparent shoulder PJI, defined by sinus tract or intra-articular purulence as major criteria. Among 861 consecutive revision cases, 97 (11%) met the primary outcome; logistic regression identified seven independent predictors: erythema, unexpected wound drainage, elevated ESR/CRP, radiographic humeral osteolysis, synovial neutrophils ≥70% or positive frozen section, intraoperative humeral loosening, and ≥2 positive cultures with the same organism. The final model achieved an apparent C-statistic of 0.90 (95% CI 0.87–0.94), and a nomogram was constructed to estimate individual PJI probability.
Key Limitation
The model lacks external validation, meaning its C-statistic of 0.90 reflects apparent discrimination in the development cohort only and true performance in independent populations is unknown.
Original Abstract
INTRODUCTION
The first International Consensus Meeting (ICM) definition of shoulder periprosthetic joint infection (PJI) introduced a tiered framework of PJI categories based on weighting of minor criteria for those cases not having major criteria of Definite infection. The selection of minor criteria and weightings was based on expert opinion and limited available evidence. The purpose of this study was to use prospective, multicenter data to develop a predictive model for clinically apparent shoulder PJI to inform a more evidence-based understanding of minor criteria and PJI categories.
METHODS
Prospective data were collected on 861 consecutive revision shoulder arthroplasty cases within the American Shoulder and Elbow Surgeons (ASES) Revision Shoulder Arthroplasty and PJI Multicenter Research Group. Presence of a sinus tract communicating with the prosthesis and intra-articular purulence were identified as pathognomonic indicators of infection (major criteria) and used as the primary outcome for multivariable analysis. Logistic regression modeling was performed, with model specification guided by information criteria. Variable selection emphasized optimal model fit while balancing complexity, focusing on: (1) markers of host inflammatory response, (2) universally available and commonly utilized diagnostic tests, (3) minimizing the total number of variables, and (4) preserving model usability despite missing data.
RESULTS
Among 861 cases, 97 shoulders (11%) met the outcome of interest (sinus tract or intra-articular purulence). The final model incorporated seven variables spanning clinical findings, laboratory values, imaging, intraoperative assessment, and microbiology. These variables were erythema, unexpected wound drainage, elevated erythrocyte sedimentation rate (ESR) or C-reactive protein (CRP), radiographic humeral osteolysis, synovial neutrophils ≥70% or positive frozen section (≥5 polymorphonuclear cells in ≥5 high-powered fields), intraoperative humeral loosening, and ≥2 positive cultures with the same organism. The apparent concordance statistic in the development cohort was 0.90 (95% CI 0.87-0.94). A nomogram was subsequently developed to estimate the probability of PJI, defined by the presence of a sinus tract or intra-articular purulence.
CONCLUSION
Using prospective, multicenter data, this study proposes a seven-factor nomogram to estimate the probability of clinically apparent shoulder PJI. The model showed strong apparent discrimination for clinically apparent infection in the development cohort. External validation will be necessary to confirm performance, and consensus will be required to establish weighting and diagnostic thresholds for tiers of non-definite shoulder PJI.