To predict the MDS-UPDRS Part III subscore for patients with Parkinson’s disease (PD) based on accelerometry data collected from a smartwatch during simulated routine tasks. 

PD motor manifestations are variable and highly dependent on the time of assessment in relation to the last dose of PD medications. Tracking these motor symptoms outside the clinic through the use of wearable devices may provide useful information for appropriately managing treatment and monitoring disease progression.

 As part of the Clinician Input Study, a multi-center study sponsored by the Michael J. Fox Foundation, 22 participants with PD performed a series of 13 tasks, simulating routine behaviors, while wearing a smartwatch device. A trained clinician evaluated tremor and bradykinesia of both upper extremities, using a scale of 0-4. Immediately after performing the tasks, the clinician rated the participant’s motor symptoms using the MDS-UPDRS Part III assessment. Each participant performed this procedure once while off-medication and once while on-medication.

Machine learning models were trained to predict the symptom scores for each task based on the smartwatch accelerometry data. Two additional models were trained to predict the Part III subscore, based on symptom scores for the motor tasks, either actual or predicted. Correlations between actual and predicted Part III subscores were evaluated.

Actual Part III subscore was moderately correlated with predicted subscore from clinician task scores (r=.55, p<.001) and predicted subscore from accelerometry-predicted task scores (r=.46, p=.006). The most important tasks for accurate prediction were gross and fine upper body movements, though postural tasks also contributed.

Accelerometry data from a smartwatch device can be used to estimate the severity of symptoms of tremor and bradykinesia in PD with reasonable accuracy. In particular, predicted scores during routine tasks are correlated with overall state of motor symptoms and could be used to monitor motor symptoms outside of the clinic.