Project overview
Developed machine-learning models to predict surgical-site infection risk after hip arthroplasty using clinical data from Hospital Universitario Ramón y Cajal collected between 2010 and 2020. The project investigated whether surgical parameters could support earlier risk assessment and improve infection-prevention strategies.
A data-driven approach to identifying infection risk in hip arthroplasty and supporting earlier prevention strategies.
Approach
Preprocessed the clinical dataset and performed feature engineering and exploratory visual analysis to identify relevant variables. Applied one-hot encoding, undersampling, SMOTE and mutual-information analysis before training and comparing logistic regression, decision tree, random forest, XGBoost and multilayer perceptron models.
Outcome
Logistic regression achieved the best overall performance, reaching 94.2% accuracy and 95.5% specificity. Random forest obtained the highest sensitivity at 78%, demonstrating the potential of machine learning to support surgical-site infection risk assessment after hip arthroplasty.