Preoperative Risk Factors and Phenotypic Clustering in Periprosthetic Joint Infection: A Matched Case-Control Study

Periprosthetic joint infection (PJI) remains one of the most serious complications after total joint arthroplasty. This retrospective 1:1 matched case-control study investigated preoperative predictors and patient phenotypes associated with PJI in 182 patients (91 cases, 91 controls) undergoing hip or knee arthroplasty between 2013 and 2024. Variables with skewed distributions were log-transformed, and multivariable logistic regression with LASSO regularization identified independent risk factors. Unsupervised K-means clustering was applied to perioperative features to explore data-driven patient phenotypes. Preoperative anemia (OR 5.91, p = 0.026), higher ASA score (OR 1.77, p = 0.033), and surgical delay (OR 1.67, p = 0.024) independently predicted infection, while age and CRP showed non-significant trends. The resulting five-variable preoperative model achieved an AUC of 0.718 (optimism-corrected AUC of 0.661) for infection prediction. Clustering analysis revealed three phenotypes: anemia-dominated, elderly but short-procedure, and high surgery duration with significantly different infection rates (χ² = 23.5, p < 0.001) but similar mortality (p = 0.068). This integrative approach combining regression-based prediction and phenotype discovery enables clinically interpretable, preoperatively applicable risk stratification. The findings identify anemia, comorbidity burden, and surgical delay as key modifiable targets for preventive optimization before arthroplasty. External validation and recalibration to population-level incidence are warranted before clinical implementation.