Patient, Procedure Factors May Predict Surgical Site Infection Risk After TJA
In This Article
- 11,882 patients who underwent primary total hip and knee arthroplasty were randomly partitioned into training (n=9,506) and testing (n=2,376) datasets for an artificial neural network model to study and predict surgical site infection
- Variables used to predict surgical site infection included patient and procedure factors, such as BMI, American Society of Anesthesiologists score, smoking status, anesthesia, duration of tourniquet (in cases of TKA) and tranexamic acid usage and dose
- According to results, Charlson comorbidity index and smoking status were the strongest predictors of surgical site infection after total joint arthroplasty
Subscribe to the latest updates from Orthopaedics Advances in Motion
Ingwon Yeo, MD, a research fellow in the Bioengineering Lab in the Department of Orthopaedics at Massachusetts General Hospital, and colleagues randomly partitioned data of 11,882 patients who underwent primary total hip and knee arthroplasty into training (n=9,506) and testing (n=2,376) datasets for an artificial neural network model. Variables used to predict surgical site infection included patient and procedure factors, such as BMI, American Society of Anesthesiologists score, smoking status, anesthesia, duration of tourniquet in cases of knee surgery and tranexamic acid usage and dose.
According to Dr. Yeo's presentation at the 2021 Orthopaedic Research Society Annual Meeting, the researchers applied principal component analysis and logistic regression to reduce the large dimensionality to the seven most statistically influential parameters. To avoid the problem of overfitting associated with the artificial neural network, fivefold validation was employed. The area under the curve (AUC) and receiver operating characteristic analysis were used as metrics predicting surgical site infection in primary total joint arthroplasty (TJA).
Dr. Yeo noted an overall incidence of surgical site infection after primary TJA of 2.7%.
The artificial neural network model had an AUC of 0.78 and the threshold probability was 0.66, which had a sensitivity of 0.76 and specificity of 0.7.
The artificial neural network identified age, gender, Charlson comorbidity index, smoking, alcohol use, diabetes and insurance type as the most influential predictors of surgical site infection. Among these variables, Charlson comorbidity index and smoking were the strongest predictors for surgical site infection.
Read More on Healio Orthopedics Today
Learn more about the Bioengineering Lab at Mass General
Learn more about the Center for Hip & Knee Replacement