Classification and stratification of patient pain archetypes following total knee arthroplasty: a machine learning approach
Authors' conclusion
Machine learning identified distinct postoperative pain trajectories after TKA, and early pain data predicted later pain patterns.
Study context
- Population
- Primary unilateral total knee arthroplasty (TKA), 2021–2024
- Participants
- 17,200 TKAs
- Exposure
- Low pain vs high pain
- Low painn = 7,082reference
- High painn = 6,678
- Outcome
- Postoperative pain archetypes (NRS 0-10, 0-72 h) and prediction of high-pain cluster membership (accuracy, ROC-AUC)
- Follow-up
- 0-72 hours postoperatively
Methodological detailssingle-centre, exploratory, machine learning, 80% training (n=13,760) / 20% testing (n=3,440) random split, LSTM with contrastive learning and K-means clustering; LASSO feature selection from 107 variables; XGBoost model
- Single-centre
- Exploratory
- Machine learning
- 80% training (n=13,760) / 20% testing (n=3,440) random split
- LSTM with contrastive learning and K-means clustering; LASSO feature selection from 107 variables; XGBoost model
Main finding
As reported in the abstract
Pain archetypes and XGBoost prediction of high-pain cluster
XGBoost accuracy for high-pain cluster membership
64%
XGBoost ROC-AUC
0.68
Observational data: associations, not effects.
Further findings
Concordance of clustering using first 12 or 24 h Numeric Rating Scale (NRS) data with 72-h results
- Concordance, first 12 h
- 61.3%
- Concordance, first 24 h
- 62.4%
- Moderate concordance
Original publication
Classification and stratification of patient pain archetypes following total knee arthroplasty: a machine learning approach. Reg Anesth Pain Med. Published 15 April 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →