Categorization of changes in the Oxford Knee Score after total knee replacement: an interpretive tool developed from a data set of 46,094 replacements

Objectives: The objective of the study was to create an interpretive categorical classification for the transition in the Oxford Knee Score (OKS) change score (ΔOKS) using the anchor-based method. Study Design and Setting: Registry data from 46,094 total knee replacements from the year 2014/15, were accessed via the Health and Social Care Information Center official website. Data included preoperative and 6-month follow-up OKS and response to the transition anchor question. Categories were determined using Gaussian approximation probability and k-fold cross-validation. Results: Four categories were identified with the corresponding ΔOKS intervals: “1. much better” (≥16), “2. a little better” (7–15), “3. about the same” (1–6), and “4. much worse” (≤0) based on the anchor questions’ original five categories. The mean 10-fold cross-validation error was 0.35 OKS points (95% confidence interval 0.12 to 0.63). Sensitivity ranged from 0.34 to 0.68; specificity ranged from 0.74 to 0.95. Conclusion: We have categorized the change score into a clinically meaningful classification. We argue it should be an addition to the continuous OKS outcome to contextualize the results in a way more applicable to the shared decision-making process and for interpreting research results.