Using Cluster Analysis to Identify Patient Factors Linked to Differential Functional Gains After Total Knee Arthroplasty.

BACKGROUND: The basis of poor outcomes following total knee arthroplasty (TKA) is multifactorial. Previous research aimed at predicting outcome following TKA focuses largely on outcomes measured between two specific time points (pre-to post-TKA). Analysis of outcomes measured over multiple time points (trajectory) may expose relationships between patients' characteristics and longitudinal outcome patterns that may otherwise remain obscured. METHODS: The current study analyzed Short Form 36 Physical Component Score (PCS) trajectories of 656 patients composed of 3 time points over a 1-year period. Clusters were constructed utilizing MultiExperiment Viewer hierarchical clustering algorithm. Statistical significance of these clusters was assessed using MeV's built-in bootstrapping method. Patient characteristics of the resulting statistically conserved clusters were summarized and compared using Wilcoxon rank-sum test or chi-squared test as appropriate. RESULTS: Two distinct clusters of outcome trajectory were identified. Cluster 1 included 550 patients (84%) who demonstrated persistent PCS improvement at 6 and 12 months. Cluster 2 included 106 patients (16%) who demonstrated decline in PCS at 6 months followed by improvement at 12 months. Cluster 1 achieved earlier success, greater absolute mental and physical health scores as compared to Cluster 2 (P < .05), and demonstrated higher baseline mental health scores, lower baseline PCS, and a significantly higher proportion of non-Hispanic Whites (P ≤ .05). CONCLUSION: Cluster analysis identified distinct functional outcome trajectories following TKA. Specific differentiating patient factors were associated with differing trajectories. Future studies should focus on this method's ability to inform predictive models regarding patient outcomes.