Comprehensive Summary
This study introduces a groundbreaking machine learning (ML) model to predict periprosthetic joint infection (PJI) after primary total knee arthroplasty (TKA) in the Asian population, marking a pioneering development in the field. PJI is a severe complication following TKA, leading to increased morbidity, mortality, and healthcare costs. The one-year and five-year mortality rates associated with PJI can reach 10.6% and 25.9%, respectively. As the number of TKA procedures rises globally, so does the economic burden of managing PJI, projected to reach $1.1 billion annually in the United States by 2030. The study analyzed 3,483 TKA cases performed between 1998 and 2021 at Queen Mary Hospital, Hong Kong. Among these, 81 cases (2.3%) developed PJI. The mean patient age was 70.4 ± 10.0 years, and 73% of the cohort were female. The study collected 60 features from patient demographics, surgery-related variables, and comorbidities, with 6 significant predictors identified through univariate and multivariate analysis. These predictors include long operative time (OR 9.07, p = 0.018), male gender (OR 3.11, p < 0.001), ASA score > 2 (OR 1.68, p = 0.028), history of anemia (OR 2.17, p = 0.023), history of septic arthritis (OR 4.35, p = 0.030), and spinal anesthesia as a protective factor (OR 0.55, p = 0.022). To address the challenge of data imbalance, where PJI cases are relatively rare, the researchers employed balanced random forest algorithms. This approach showed superior predictive performance with an AUC of 0.963, balanced accuracy of 0.920, sensitivity of 0.938, and specificity of 0.902. The model outperformed traditional approaches, including gradient boosting machines (AUC 0.931) and logistic regression (AUC 0.728), demonstrating robust and reliable performance in predicting PJI risk. The SHapley Additive exPlanations (SHAP) plots provided both global and local interpretability, highlighting operative time and gender as the most influential factors. The findings offer healthcare professionals a personalized risk prediction tool that enhances preoperative planning and patient counseling. By identifying high-risk individuals early, it enables proactive intervention strategies and informs shared decision-making between clinicians and patients. Moreover, integrating this model into clinical practice can improve patient outcomes and optimize healthcare resource utilization.
Outcomes and Implications
The implications of this study are profound, particularly in the context of personalized healthcare and risk management for TKA patients. The study demonstrates that prolonged operative time (OR 9.07) significantly elevates the risk of PJI, indicating the need for streamlined surgical procedures and efficient intraoperative practices to reduce infection risks. Male patients, with an OR of 3.11, are at notably higher risk, highlighting the importance of gender-specific preoperative assessment and counseling. Among anesthesia types, spinal anesthesia emerged as a protective factor, reducing PJI risk by 45% (OR 0.55) compared to general anesthesia. This insight underscores the potential benefits of using spinal anesthesia when clinically appropriate. Furthermore, patients with an ASA score > 2 (OR 1.68) or history of anemia (OR 2.17) are particularly vulnerable, suggesting that preoperative optimization of comorbid conditions could mitigate risks. Additionally, a history of septic arthritis (OR 4.35) substantially increases PJI risk, necessitating intensive perioperative monitoring for these individuals. Implementing this machine learning model in clinical settings can revolutionize risk stratification and preoperative planning. By predicting high-risk patients preoperatively, healthcare providers can implement targeted interventions, including thorough preoperative optimization, careful intraoperative management, and vigilant postoperative monitoring. Additionally, the model facilitates individualized counseling, helping patients understand their unique risk profiles and empowering them to make informed healthcare decisions. From a healthcare management perspective, this model can enhance resource allocation by prioritizing high-risk patients for more intensive perioperative care. This can lead to reduced hospital costs, minimized postoperative complications, and improved patient outcomes. Future research should focus on multi-center validation and enhancing the model with additional clinical variables to further boost accuracy and generalizability. By leveraging machine learning and advanced predictive modeling, this study establishes a foundational approach to preventing periprosthetic joint infections in Asian populations, promoting evidence-based, patient-centric care while addressing the growing economic burden of PJI management.