Description
Our model aimed to predict minimal clinical important difference (MCID) in patients receiving platelet-rich plasma (PRP) treatment for knee osteoarthritis (OA) at 6 months using machine learning models. We used data from 191 patients who received PRP treatment and evaluated various pre-treatment predictors.
The results showed that the Explainable Boosting Machine (EBM) model effectively predicted achieving MCID in knee function scores at 6 months, with an AUC-ROC of 0.84 and F1 Score of 0.75. Other models performed similarly.
Key findings included that high pre-injection KOOS, male gender, and older age were negative predictors of achieving MCID, while a higher PROMIS Mental score was the most important positive predictor. Patients with Kellgren-Lawrence grade II OA were more likely to achieve MCID. The study concluded that the EBM model was effective in predicting outcomes using pre-treatment factors, but external validation is needed.
