Summary

The three questions on overall knee pain, limping when walking, and knee ‘giving way’ were the strongest predictors of subsequent revision within two years.

Abstract

Aims
The Oxford Knee Score (OKS) is a 12-item questionnaire used to track knee arthroplasty outcomes. Validation of such patient reported outcome measures is typically anchored to a single question based on patient ‘satisfaction’, however risk of subsequent revision surgery is also an important outcome measure. The OKS can predict subsequent revision risk within two years, however it is not known which item(s) are the strongest predictors. Our aim was to identify which questions were most relevant in the prediction of subsequent knee arthroplasty revision risk.
Patients and Methods
All primary TKAs (n=27,708) and UKAs (n=8,415) captured by the New Zealand Joint Registry between 1999 and 2019 with at least one OKS response at six months, five years or ten years post-surgery were included. Logistic regression and receiver operating characteristics (ROC) curves were used to assess prediction models at six months, five years and ten years.

Results

Q1 ‘overall pain’ was the strongest predictor of revision within two years (TKA: 6 months, odds ratio (OR) 1.37; 5 years, OR 1.80; 10 years, OR 1.43; UKA: 6 months, OR 1.32; 5 years, OR 2.88; 10 years, OR 1.85; all p<0.05). A reduced model with just three questions (Q1, Q6 ‘limping when walking’, Q10 ‘knee giving way’) showed comparable or better diagnostic ability with the full OKS (area under the curve (AUC): TKA: 6 months, 0.77 vs. 0.76; 5 years, 0.78 vs. 0.75; 10 years, 0.76 vs. 0.73; UKA: 6 months, 0.80 vs. 0.78; 5 years: 0.81 vs. 0.77; 10 years, 0.80 vs. 0.77).

Conclusions

The three questions on overall knee pain, limping when walking, and knee ‘giving way’ were the strongest predictors of subsequent revision within two years. Attention to the responses for these three key questions during follow-up may allow for prompt identification of patients most at risk of revision.