Smartphone-based Remote Digital Monitoring for Parkinson’s Disease and Dementia
Kinan Muhammed1, Francis Chmiel1, Siddharth Arora2, Katarina Gunter3, Masud Husain3, Michele Hu3
1Neuhealth Digital, 2Somerville College, 3Nuffield Department of Clinical Neurosciences, University of Oxford
Objective:
Develop an objective, multi-modal digital platform (Neu Health) to remotely measure changes and predict clinical outcomes in patients with Parkinson's disease (PD) and dementia using a smartphone application and interactive dashboard.
Background:
PD and dementia are becoming increasingly prevalent and have significant impact on morbidity and mortality. Challenges include medication management, limited clinical resources, and therapy access disparities, particularly in rural, minority, and low-income communities. Innovative solutions to objectively measure disease progression, assess patient risk, and optimize medication management are needed.
Design/Methods:
A software-based platform was developed that utilizes smartphone sensors to quantify signs and symptoms remotely. Digital assessments included voice, balance, gait, finger tapping dexterity, reaction time, rest and postural tremor, and behavioural cognitive tests. 689 healthy controls and over 600 patients with either idiopathic PD or Alzheimer’s disease completed the assessments. Unified Parkinson’s Disease Rating Scale (MDS-UPDRS) and Addenbrookes Cognitive Examination (ACE-III) were also performed. Clinical scores and risk of future disease outcomes were estimated using smartphone features, heuristic analysis, and machine learning models. Key metrics were displayed on a clinician dashboard.
Results:
Digital motor features captured from the smartphone assessments showed significant correlations with corresponding clinical questions from the MDS-UPDRS (r 0.11 to 0.51, p<0.05). The tests were sensitive to daily fluctuations in PD symptoms and the dashboard could be used to view clinical characteristics and monitor response to medication. Cognitive tests also significantly correlated with ACE-III (r>0.52, p<0.05). Smartphone features could be used to estimate clinical scores and predict future outcomes, including risk of falls, disease progression and cognitive decline, up to 18 months in advance (AUC > 0.7).
Conclusions:
Remote smartphone assessments can estimate clinical scores and risk outcomes that are clinically interpretable. A reliable digital platform integrating assessment and predictive capabilities could improve neurodegenerative condition management. Further research should be performed to assess impact on clinical care.