Predicting Outcome of Total Knee Arthroplasty by Cluster Analysis of Patient-Reported Outcome Measures

• Published in: The Journal of arthroplasty 2024-10
• Main Authors: Littman, Jake L., Molino, Janine, Olansen, Jon B., Antoci, Valentin, Aaron, Roy K.
• Format: Article
• Language: English
• Subjects: cluster analysis; KOOS; MCID; PROM; total knee arthroplasty
• ISSN: 0883-5403; 1532-8406; 1532-8406
• DOI: 10.1016/j.arth.2024.09.039

Total knee arthroplasties (TKAs) exhibit an 8 to 30% risk of suboptimal outcomes, resulting in persistent symptoms, individual morbidity, and revision surgery, prompting a contemporary focus on risk reduction and outcome improvement. This study introduces hierarchical cluster analysis as a way of preoperatively assessing the likelihood of success/failure of TKA based on several patient-reported outcome measures (PROMs), which have been analyzed both intact and with component questions as individual variables. The study utilized data on 1,433 TKAs from The Miriam Hospital's Function and Outcomes Research for Comparative Effectiveness in Total Joint Replacement registry. Outcomes are expressed as Knee Injury and Osteoarthritis Outcome Score pain and function scores. Criteria for success/failure were developed with an integrative, anchor-based, minimum clinically important difference. Preoperative and postoperative PROMs were studied by cluster analysis. There were three sequential cluster analyses that revealed clusters of patients, based upon preoperative patient responses that were predictive of surgical outcomes. Clusters varied most significantly in their responses to individual component questions of preoperative PROMs. Extracting and combining the clinically meaningful patient-reported component questions yielded a new, and clinically relevant, outcome measure that has the potential to preoperatively predict postoperative outcomes of TKA. In contrast to a single medical, psychological, or social variable, cluster analysis offers the opportunity to develop a whole-patient profile that reflects the contextual interactions of sociodemographic and clinical variables in predicting outcomes. In the context of determining clinical meaningfulness, cluster analysis has one of its major strengths.