Sensor-derived turning measures for differentiating earlier-stage Parkinson's disease and characterizing motor severity.

INTRODUCTION

Turning is commonly impaired in Parkinson's disease (PD), but whether distinct sensor-derived turning measures support earlier-stage discrimination and severity assessment remains unclear. We examined their complementary clinical value.

METHODS

Participants were 79 healthy controls, 51 with earlier-stage PD (Hoehn and Yahr [H&Y] stages 1-2), and 62 with moderate-stage PD (H&Y stage 3). Thirteen sensor-derived motor features were examined using prespecified group comparisons, associations with MDS-UPDRS Part III, and age- and sex-adjusted regression. Random forest classifiers used 10-fold cross-validation.

RESULTS

Turning duration was longer in earlier-stage PD than in controls (4.43 ± 1.48 vs 2.73 ± 0.73 s; P < 0.001) but did not differ between PD strata (P = 0.059). Turning velocity decreased across the three groups (216.16 ± 41.20, 143.29 ± 45.80, and 105.46 ± 40.16°/s; both pairwise P < 0.001) and was inversely associated with MDS-UPDRS Part III (r = -0.349, P = 0.0001). These findings persisted after adjustment for age and sex (standardized β = 1.220 for turning duration, -1.243 and -0.843 for the two turning velocity comparisons, and -0.323 for its association with MDS-UPDRS Part III; all P ≤ 0.0003); turning duration remained nonsignificant between PD strata (P = 0.255). Motor-feature models outperformed age- and sex-only models (AUC: 0.943 vs 0.575 and 0.766 vs 0.606).

CONCLUSION

A brief standardized turning task yielded complementary information: turning duration showed potential for differentiating earlier-stage PD from healthy controls, whereas turning velocity reflected motor severity. These measures may support quantitative PD assessment, pending external and longitudinal validation.