Predicting Dissatisfaction with Total Knee Arthroplasty

Total knee arthroplasty (TKA) is a common and effective treatment for end-stage osteoarthritis. However, approximately 20% of patients are consistently found to be dissatisfied or unsure of their satisfaction with their TKA at 1-year post-surgery. Two strategies for reducing dissatisfaction have appeared in the literature. The first aims to reduce the number of dissatisfied patients who will undergo surgery by identifying patients likely to be dissatisfied pre-surgery. The second aims to implement interventions that improve patient satisfaction by generating realistic expectations for outcomes and by association improving met expectations scores. It may be possible to improve the performance of these interventions by identifying dissatisfied patients early in the recovery process. This thesis aimed to determine if dissatisfied patients can be identified pre-surgery or early in the recovery process as well as to better understand the met expectations variable. Four studies were completed for this thesis. Study 1 attempted to predict 1-year post-surgery dissatisfaction using logistic regression and machine learning methods with pre-surgery and surgical variables. Study 2 aimed to identify patients dissatisfied at 1-year post-surgery using logistic regression and classification trees with pre and 3-month post-surgery data. Study 3 expanded on the findings of study 2 by creating a prediction tool to identify dissatisfied patients that can be easily administered in a clinical setting at 3-months post-surgery. This was done by using a pooled index that included five individual questionnaire items drawn from the 3-month post-surgery Knee injury and Osteoarthritis Outcome Score and the Knee Society Knee Scoring System questionnaires. Study 4 investigated the 1-year post-surgery met expectations variable and its relationship with satisfaction. The results of study 1 indicated that pre-surgery and surgical variables were not sufficient to discriminate between satisfied and dissatisfied patients at 1-year post-surgery. The results of study 2 indicated that it is feasible to accurately identify patients who will be dissatisfied at 1-year post-surgery using 3-months post-surgery data. Study 3 found that the pooled index was able to accurately discriminate between satisfied and dissatisfied patients. The results of study 4 indicate that met expectations moderate the relationship between pain and satisfaction. This means that as met expectations scores increase, pain becomes less important for improving satisfaction. Overall, this thesis found that patients dissatisfied at 1-year post-surgery cannot accurately be identified using pre-surgery and surgical variables. However, these patients can be identified at 3-months post-surgery using a simple prediction tool that can be easily administered in a clinical setting. Lastly, this thesis found that met expectations represent an important subjective threshold for patients and it may be reasonable to target unmet expectations to improve satisfaction