Predicting whether patients will achieve minimal clinically important differences following hip or knee arthroplasty

Area of Science:

• Orthopedic surgery outcomes research
• Health informatics
• Machine learning in healthcare

Background:

• A significant portion of patients undergoing knee arthroplasty (KA) or hip arthroplasty (HA) do not achieve a minimal clinically important difference (MCID), indicating a lack of meaningful improvement.
• Patient-reported outcome measures (PROMs) are crucial for assessing treatment success after arthroplasty.

Purpose of the Study:

• To evaluate the predictive performance of machine learning (ML) models, logistic regression (LR), and pre-surgery PROM scores in forecasting whether patients undergoing HA or KA achieve an MCID.
• To compare the predictive capabilities of ML against LR and pre-surgery PROM scores.

Main Methods:

• Minimal clinically important differences (MCIDs) were determined using the change difference method in a cohort of 1,843 HA and 1,546 KA patients.
• Various ML algorithms (artificial neural network, gradient boosting machine, LASSO, ridge, elastic net, random forest) and LR were employed, alongside pre-surgery PROM scores, to predict MCID achievement for EQ-5D-5L, EQ-VAS, HOOS-PS, and KOOS-PS.

Main Results:

• The predictive performance for the best models varied, ranging from 0.71 for HOOS-PS to 0.84 for EQ-VAS in the HA sample.
• Machine learning demonstrated statistically significant superiority over logistic regression and pre-surgery PROM scores in two out of six prediction tasks.

Conclusions:

• Predicting MCID achievement in arthroplasty patients is feasible with reasonable accuracy.
• While ML models show promise and can outperform traditional methods, this advantage was observed in a limited number of cases.