Data Science Strategies for Modelling Total Knee Arthroplasty Patient Variability
MetadataShow full item record
Knee osteoarthritis (OA) can present through self-reported patient symptoms and joint-level manifestations, impairing mobility and function. Symptom relief and functional restoration of end-stage OA is typically treated with total knee arthroplasty (TKA) surgery. TKA success it not universal; over 20% of patients do not report satisfaction or clinically meaningful pain and function improvements post-TKA. This thesis aimed to improve our understanding of multidimensional variability among OA patients, and investigate how variability manifests into different TKA outcomes. Four studies investigated relationships between patient-reported measures and knee kinematics and kinetics during walking gait, characterized demographic and knee biomechanical variability (clusters) among OA and TKA populations, and addressed relationships between clusters and functional outcomes after TKA. The first study objective examined demographic and patient-reported factors pre-TKA to two-years post-TKA associated with patient-reported TKA satisfaction using longitudinal analysis. Study 2 examine demographics, pre-TKA knee kinematics and kinetics during gait, and post-TKA gait changes associated with self-reported pain and function improvements one-year post-TKA using regression models. Studies 3 and 4 characterize demographic and knee biomechanical variability among (3) pre-TKA and (4) asymptomatic to post-TKA observations using machine-learning cluster analysis. Studies revealed that (1) pre-operative self-reported symptoms were not predictive of longitudinal satisfaction; however, findings support the ability to identify less satisfied patients as early as six-weeks post-TKA. (2) Frontal and sagittal knee kinematic patterns during gait pre-TKA, and less adduction angle reductions post-TKA were associated with greater self-reported pain and function improvements post-TKA. (3) TKA candidates and (4) patients along the OA continuum can be characterized by demographic and knee biomechanic clusters, separated by features corresponding to clinical OA severity. Collectively, these studies characterized temporal and multidimensional patient variability encompassing self-reported symptoms and knee biomechanics using data science and machine-learning strategies. Knee biomechanics provided important insights into mechanical factors impacting the patient experience, and biomechanical cluster profiling supported the ability to classify patients who may benefit most from TKA. Findings support the utility to reveal novel insights into the patient experience using advanced data science strategies, providing direction for innovations in OA management and TKA care. Presented methodologies are directly applicable to other clinical applications.