Multidimensional analysis of preoperative patient-reported outcome measures identifies distinct phenotypes in patients booked for total knee arthroplasty: Secondary analysis of the SHARKS registry in a metropolitan hospital department

BackgroundTraditional research on total knee arthroplasty (TKA) relies on preoperative patient-reported outcome measures (PROMs) to predict postoperative satisfaction. We aim to identify distinct patient phenotypes among TKA candidates, and investigate their correlations with patient characteristics. MethodsBetween 2017-2021, 389 patients with 450 primary knee cases at a metropolitan public hospital were enrolled in a clinical quality registry. Demographics, clinical data, and the Veterans Rand 12 (VR-12) and Oxford Knee Score (OKS) were collected. Imputed data were utilised for the primary analysis, employing k-means clustering to identify four phenotypes. ANOVA assessed differences in scores between clusters, and nominal logistic regression determined relationships between phenotypes and patient age, sex, body mass index, and laterality. ResultsThe sample comprised 389 patients with 450 primary knees. Phenotype 4 (Mild symptoms with good mental health) exhibited superior physical function and overall health. In contrast, patients in phenotype 2 (Severe symptoms with poor mental health) experienced the most knee pain and health issues. Phenotype 1 (Moderate symptoms with good mental health) reported high mental health scores despite knee pain and physical impairment. Patient characteristics significantly correlated with phenotypes; those in the Severe symptoms with poor mental health phenotype were more likely to be younger, female, have a higher BMI, and bilateral osteoarthritis (P<0.05). ConclusionsThis multidimensional analysis identified TKA patient phenotypes based on common PROMs, revealing associations with patient demographics. This approach has the potential to inform prognostic models, enhancing clinical decision-making and patient outcomes in joint replacement. Significance and InnovationsThis study leverages the power of machine learning to simultaneously analyse multiple patient-reported outcome measures, which is not utilised in traditional research in total knee arthroplasty Four distinct phenotypes were identified, and they demonstrated significant associations with patient demographics This method has potential for developing prognostic models in joint replacement, with the ability to improve clinical decision-making and patient outcomes.