Using Web Interaction to Monitor Parkinson's Disease Progression through Behavioural Inferences on the Web
Julio Vega · 2016 · Proceedings of the 13th International Web for All Conference (W4A) · doi:10.1145/2899475.2899502
Summary
This doctoral consortium paper from the University of Manchester proposes a novel approach to monitoring Parkinson's Disease (PD) progression using passive smartphone data collection and web interaction analysis, replacing the intrusive wearable devices and scripted evaluation tasks used in prior research. Traditional PD assessment relies on clinical scales administered during sporadic clinic visits — an approach the author identifies as subjective, expertise-dependent, prone to recall bias, and unable to capture symptom fluctuations throughout the day. While recent wearable-based approaches have improved objectivity, most require participants to wear multiple uncomfortable devices and perform structured tasks, making them unsuitable for long-term naturalistic monitoring. Vega's approach instead uses smartphones to passively log 29 types of data sources — including sensor data, touchscreen interactions, keystroke patterns, app usage, DOM logs from visited websites, GPS, and ambient data like weather — complemented by external web data sources. From this data, the methodology infers "Latent Behavioural Variables" (LBVs): composite metrics quantifying specific human activities and habits. A "Profile of Living" metric then splits the monitoring period to establish a personal baseline and measure deviations over time, which are correlated with clinical MDS-UPDRS scores.
Key findings
A pilot study with two people with Parkinson's over 83 days collected approximately 290 million data records — 34.5 times more rows and scanning 4 times more data sources than state-of-the-art datasets. The dataset achieved 107,680 records per person per monitored hour across 29 data sources. Six candidate PD-related LBVs were identified: typing patterns, phone usage patterns, climbing and descending stairs, indoor routines, motor activities, and social patterns. The phone usage patterns LBV — based on keystrokes, touchscreen interactions, and DOM logs — could measure hand movement performance during free, naturalistic use rather than controlled tasks. Social media and email communication patterns could potentially link to cognitive and emotional difficulties caused by PD. A focus group with pilot participants revealed they did not find the passive monitoring approach privacy-invasive or burdensome. Critically, participants expressed a wish for a monitoring tool that frees them from actively tracking symptoms, which they said causes them to dwell on the disease — a powerful argument for passive over active monitoring approaches.
Relevance
This research sits at the intersection of digital health and accessibility, proposing that everyday technology interactions — including web browsing patterns, typing behavior, and app usage — can serve as unobtrusive biomarkers for disease progression. For accessibility practitioners, the work has several implications: first, it demonstrates that changes in how people interact with digital interfaces can reflect underlying motor and cognitive changes, suggesting that interaction analytics could inform adaptive accessibility features. Second, the passive sensing approach embodies an important principle in assistive technology design — that monitoring and assessment tools should not add burden to people already managing a chronic condition. The participant feedback about wanting to be freed from symptom tracking resonates broadly with user-centered design for chronic illness. The web interaction component — analyzing keystroke patterns, touchscreen use, and browsing behavior — connects this medical monitoring work directly to web accessibility concerns, as PD motor symptoms like tremor and bradykinesia directly affect web interaction quality.
Tags: health monitoring · Parkinson's disease · smartphones · machine learning · digital health · neurological conditions · passive sensing · mobile accessibility