Photo Credit: TefiM

The following is a summary of “Primary care prediction of hip and knee replacement 1-5 years in advance using temporal graph-based convolutional neural networks (TG-CNNs),” published in the April 2025 issue of Rheumatology by Hancox et al. 

Researchers conducted a retrospective study to predict primary hip or knee replacement risk 1 and 5 years in advance using clinical codes.  

They used primary care clinical codes from ResearchOne Electronic Health Records (EHRs) (1999–2014) to model patient pathways before hip or knee replacement. They trained and tested models to predict replacement risk 1 and 5 years in advance using temporal graphs, where nodes were clinical codes and edges were the time between visits. Cases were matched to controls by age, sex, and Index of Multiple Deprivation (IMD). Model performance was validated on unseen data using area under the receiver operator curve (AUROC), calibration, and area under the precision recall curve (AUPRC), with recalibration for class imbalance. 

The results showed AUROC of 0.915 (95% CI: 0.914, 0.916) at 1-year and 0.955 (95% CI: 0.954, 0.956) at 5-year for knee replacement, with AUPRCs of 0.353 (95% CI: 0.302, 0.403) and 0.442 (95% CI: 0.382, 0.503). For hip replacement, AUROC was 0.919 (95% CI: 0.918, 0.920) at 1-year and 0.967 (95% CI: 0.966, 0.968) at 5-year, with AUPRCs of 0.409 (95% CI: 0.366, 0.451) and 0.879 (95% CI: 0.833, 0.924). 

Investigators predicted hip and knee replacement risk up to 5 years in advance using a temporal graph-based artificial intelligence (AI) model. They suggested its use for planning preventative treatment or triaging patients. 

Source: academic.oup.com/rheumatology/advance-article/doi/10.1093/rheumatology/keaf185/8105566