Predicting periprosthetic joint Infection: Evaluating supervised machine learning models for clinical application

Abstract

Background and objectives: Periprosthetic joint infection (PJI) is a serious complication that can occur after joint arthroplasty, such as hip or knee replacement surgeries. It involves the invasion of the periprosthetic space by pathogens, leading to severe inflammation and often requiring complex medical intervention. PJI is associated with significant morbidity, increased healthcare costs, and a reduced quality of life for patients. This study aims to evaluate the performance of multiple supervised machine learning models in predicting PJI using clinical and demographic data collected from patients who underwent joint arthroplasty.

Methods: Eight supervised machine learning models-Logistic Regression, Random Forest, XGBoost, Artificial Neural Network (ANN), k-Nearest Neighbors (KNN), AdaBoost, Gaussian Naive Bayes (GNB), and Stochastic Gradient Descent (SGD)-were trained and tested on a dataset of 27,854 patients. Models were evaluated using accuracy, precision, recall, specificity, F1 score, and area under the ROC curve (AUC).

Results: Random Forest and XGBoost showed the best overall performance, with high accuracy and balanced metrics across all evaluation criteria. KNN also performed strongly, particularly in minimizing misclassifications. GNB and SGD yielded weaker results, with higher error rates.

Conclusion: Random Forest, XGBoost, and KNN are the most promising models for clinical implementation in PJI prediction. Their robust performance may support earlier diagnosis and improved patient outcomes in orthopedic care.

Translational potential statement: This study demonstrates that machine learning models-particularly Random Forest and XGBoost-can accurately predict periprosthetic joint infection (PJI) using structured electronic health record data. By integrating these models into preoperative assessment workflows, clinicians may be able to identify high-risk patients earlier, personalize prophylactic strategies, and reduce infection-related morbidity. The implementation of these predictive tools has the potential to enhance clinical decision-making, improve surgical outcomes, and optimize the use of healthcare resources in orthopedic practice.

Keywords: Artificial intelligence; Classification metrics; Machine learning; Orthopaedics; Periprosthetic joint infection; Predictive modelling.

Graphical abstract

Fig. 1

Schematic representation of the diagnostic algorithm of periprosthetic joint infection (PJI); WBC = white blood cells, ESR = erythrocytes sedimentation rate; CRP = C reactive protein; FIB = fibrinogen; PMN = polymorphonuclear leukocytes.

Fig. 2

Pearson correlation heatmap of relevant numerical and binary variables prior to model training. The plot shows pairwise Pearson correlation coefficients, with red indicating strong positive correlations and blue indicating strong negative correlations. Variables include demographic, laboratory, and engineered clinical features.

Fig. 3

System architecture for ML-based PJI prediction, illustrating the pipeline from dataset extraction and feature selection to preprocessing, model training (with hyperparameter tuning), and final evaluation.

Fig. 4

Schematic representation of the evaluation process applied to all machine learning models, connecting the FOISOR dataset to ROC-AUC analysis and confusion matrix generation.

Fig. 5

Receiver Operating Characteristic (ROC) curves for all eight machine learning models evaluated for PJI prediction. Curves are based on test set performance, with area under the curve (AUC) used to assess discriminatory ability.

Fig. 6

Confusion matrices for seven machine learning models used in PJI prediction, showing true positive (TP), true negative (TN), false positive (FP), and false negative (FN) counts based on the test set.

Fig. 7

Mean absolute SHAP values for the top 25 features contributing to the Random Forest model's prediction of septic cases. Higher values indicate greater impact on model output. Features are ordered by importance for class 1 (PJI-positive).