Abstract / Summary
Background/Objectives: Parkinson’s disease (PD) is clinically heterogeneous, and motor complications are often inadequately captured during routine outpatient assessments. In this cross-sectional observational study, we investigated whether prolonged home-based wearable monitoring could enable digital motor profiling by identifying clinically meaningful patient subgroups through an unsupervised analytical approach. Methods: Seventy-six patients with PD underwent standardized clinical evaluation followed by prolonged home-based monitoring using a single waist-worn inertial sensor. Wearable-derived measures of gait, mobility, and complex motor manifestations, including dyskinesia and freezing of gait (FOG), were analyzed using an unsupervised clustering approach based exclusively on sensor-derived variables. The resulting subgroups were subsequently characterized using clinical and digital outcomes, and cluster stability and interpretability were examined through bootstrap and complementary machine-learning analyses. Results: Unsupervised clustering identified two distinct digital motor profiles that differed primarily in the burden of motor complications rather than overall disease severity. Although disease duration, Hoehn and Yahr stage, and clinical motor phenotype did not differ significantly, patients in one subgroup exhibited significantly greater dyskinesia, motor fluctuations, and dopaminergic treatment requirements, accompanied by a distinct sensor-derived motor profile during daily life. Dyskinesia-related metrics emerged as the strongest contributors to profile discrimination, and the identified clusters were reproduced with high cross-validated agreement by the surrogate classification model. Conclusions: Prolonged home-based wearable monitoring may support clinically meaningful digital motor profiling in PD. Beyond conventional clinical staging, this approach can identify patients with different burdens of motor complications under real-world conditions. Following appropriate validation, objective digital biomarkers could support patient stratification, longitudinal monitoring, and treatment optimization.