Radiographic imaging is typically used to diagnose osteoarthritis (OA). However, patients would typically be sent for imaging after they present to a physician because of joint pain. By this time, the condition is likely irreversible. This study aims to determine if human ethomics (i.e. behavior) defined by whole-body kinematics during walking, can be used as a diagnostic biomarker of hip OA. Three-dimensional motion capture was performed on 106 participants with unilateral hip OA and 80 asymptomatic participants (N = 80) during walking. Sixteen sagittal plane joint angle variables were extracted and used as inputs into the prediction model. The categorical outcome was the radiographic severity of hip OA using the Kallgren-Lawrence (KL) scale (0 [no OA], 2, 3, 4[worse]). Functional data boosting was used for statistical modelling with bootstrap resampling. Our ethomics approach to hip OA diagnosis had positive likelihood ratio (LR+) values ranging from 4.79 (95 %CI 3.20, 7.42) to detect the presence of KL3, to 43.95 (95 % CI 14.9, 76.08) to detect the presence of any OA. The present approach had negative likelihood ratio (LR−) values ranging from 0.56 (95 %CI 0.33, 0.79) of 0.07 (95 %CI 0.04, 0.11) to detect the absence of KL4, to 0.07 (95 %CI 0.04, 0.11) to detect the absence of any OA. Human ethomics represents an ideal candidate for OA biomarkers that could overcome many of the logistical challenges of traditional imaging and biochemical biomarkers.

Osteoarthritis

Hip

Diagnosis

Biomarkers

Ethomics

Biomechanics

Kinematics

Machine learning

© 2025 The Authors. Published by Elsevier Ltd.