Multivariate Time-Series Cluster Analysis for Multiple Functional Domains to Identify Recovery Patterns of Patients With Fragility Hip Fracture After Surgery

Abstract

Patients who have undergone hip fracture surgery have the primary goal of recovering their premorbid level of function across diverse functional domains, including walking ability, balance, cognitive function, emotional well-being, frailty, and activities of daily living. As the speed and level of recovery can vary substantially across functional domains and individuals, the varying recovery patterns of different outcome measures should be considered when designing rehabilitation plans for patients. However, the lack of knowledge of recovery trajectories and their variations in hip fracture patients impedes such efforts. In this study, we develop a multivariate time-series clustering algorithm to analyze the recovery patterns and identify patient groups with similar recovery patterns across multiple functional outcomes. Five distinct recovery patterns were observed that exhibit varying maximum recovery levels and speeds. These findings demonstrated the significance of utilizing multiple outcome measures concurrently to assess the patient’s recovery level. Recovery patterns are identified to exhibit variations across different domains, revealing contrasting trends between walking ability and cognitive outcomes. Furthermore, we present predictions on the trajectory of recovery during the post-acute phase solely based on the acute-phase information. This approach facilitates the early identification of patient groups with an unfavorable prognosis for recovery.

Keywords

• Machine learning
• patient rehabilitation
• pattern clustering
• time series analysis