Objective: To evaluate early real-world use of a bounded conversational assistant that helps people with Parkinson disease interpret wearable and patient-reported data and identify when clinician follow-up may be warranted.

Background: People with Parkinson disease increasingly generate continuous wearable and app-based data, but most lack practical tools to interpret change between visits. In this model, questions are routed across three bounded personas for education, personalized data interpretation, and guidance when clinician outreach may be warranted.

Method:

A conversational assistant was embedded within a digital Parkinson disease platform integrating smartphone and smartwatch measures with medication, exercise, symptom, and other patient-reported data. To keep responses bounded and clinically relevant, the system used three vetted routing personas: education, personalized interpretation, and clinician-directed guidance. Routing logic, prompts, and safeguards were extensively tested before deployment. Analytics captured chats per user, conversational turns, and topic categories; surveys and output review also assessed value, response accuracy, and when clinician follow-up was suggested.

Results:

Users completed 5 chats on average and 3 conversational turns per interaction. Common topics were symptom interpretation and pattern recognition (35-40%), medication questions (20-25%), and emotional processing or reassurance (10-15%). Three usage archetypes emerged: episodic, pattern-seeking, and transactional. Review found no inaccurate responses, and clinician follow-up was suggested in many conversations. Survey responses suggested that users often initiated chat during symptom change or uncertainty about medication effects, and many reported better understanding of symptom patterns and greater reassurance between visits.

Conclusion: Early use suggests that a bounded conversational assistant can help people with Parkinson disease better understand symptom fluctuations and patient-generated data between visits while preserving a pathway to clinician follow-up when needed. More broadly, this model may support guided self-exploration of wearable and self-reported data in a clinically grounded environment and could inform future tools for safer patient-facing interpretation, earlier escalation, and stronger between-visit communication.

« Back to 2026 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/early-use-of-a-conversational-assistant-to-interpret-wearable-and-patient-reported-data-in-parkinson-disease/