Summary

Mathematically derived variables which predict the likelihood of achieving a high Oxford Knee Score in our large cohort

Abstract

Background

and objectives
The Oxford knee score (OKS) is a 12-item patient reported outcome measure (PROM) specifically designed and developed to assess function and pain after total knee replacement (TKR). With a possible maximum score of 48, the threshold score for what are considered good and excellent outcomes vary in the literature. This study aimed to explore the distribution of 1 year post-operative (postop) OKS scores over a 20 year period at our institute, compare these to similar published cohorts, and determine whether any patient or surgical variables are predictive of the highest scores in our cohort. Our secondary aim was to construct a tool (nomogram) to determine the mathematical relationship between predictive variables and the probability of achieving the highest scores in our cohort.

Methods

A retrospective analysis was carried out on prospectively collected data on patients undergoing single primary or bilateral simultaneous TKR at a single centre from October 2002 until December 2022. Baseline data was available for the majority of the cohort and included demographics, intra-operative findings, and pre-operative OKS. SPSS 29 was used to with descriptive statistics explore the distribution of demographic data and postop OKS, and allow determination of a threshold to permit analysis of outcome as a binary variable (“high score” vs “other”). Chi-squared and t-tests were then used to analyse demographic and clinical data for predictors of high score. Stata 15 was used to construct the nomogram, using logistic univariate and multivariate regression to identify and variables and construct the mathematical model to determine probability function of the binary outcome (high score).

Results

Following exclusions for inflammatory arthropathy and previous surgery (other than arthroscopy), 4126 knees with OKS scores collected at one year (range 11-14 months) post operatively (postop), were available for analysis. 56.5% of the cohort were female with a mean age at surgery of 69.2 years. The mean OKS at 1 year post op was 41.4 (standard deviation of 6.03) and median of 43. 50% of patients achieved a score of 43 and higher, 36% of 45 or higher, 25% of 46 or higher, and 10% of 47 or higher indicating a strong left skew of data. The threshold for binary outcome of high score was chosen as 45 and greater. Following analysis, four variables formed the predictive model and construction of nomogram; intraop achieved flexion (continuous), age at surgery (continuous), preop BMI (categorical) and insert thickness (categorical), with formula: probability of OKS ≥45 = -0.02 * (Age at surgery) - 0.37 * (BMI) + 0.02 * (Intraop flexion) + 0.39 * (Insert thickness) – 1.19.

Conclusions

Our results are consistent with the general literature in the finding of a ceiling effect with OKS, albeit with comparatively higher scores, and further supporting the use of a higher threshold in our cohort for an excellent outcome. The strongest predictors of achieving the highest scores in our group included age at surgery, BMI, intraoperative achieved flexion, and insert thickness, and were used to construct a predictive nomogram. In addition, the results lend support to the use of additional PROMs to mitigate the strong ceiling effect of the post operative OKS.