Currently submitted to: JMIR mHealth and uHealth
Date Submitted: May 12, 2026
Open Peer Review Period: May 12, 2026 - Jul 7, 2026
(currently open for review)
Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Smartphone Keystroke Monitoring to Predict Progression Independent of Relapse in Multiple Sclerosis: A machine-learning-based evaluation
ABSTRACT
Background:
Progression independent of relapse activity (PIRA) is a major driver of long-term disability in multiple sclerosis (MS), yet its diagnosis remains retrospective, delaying treatment adaptation. Objective, real-time tools to anticipate PIRA are lacking.
Objective:
This study aimed to identify the most effective machine-learning model for PIRA prediction using passively acquired smartphone data and to evaluate its diagnostic performance.
Methods:
We conducted a prospective observational cohort study at a single MS center (November 2021–January 2024). Adults (≥18 years) with confirmed MS, compatible smartphone use, and ≥12 months follow-up were included. A custom mobile application continuously captured typing dynamics, extracting tapping speed as a digital biomarker. Time-series were analyzed using mixed linear regression and ARIMA models. Longitudinal data were transformed into a supervised learning dataset by segmenting the time series into overlapping temporal windows, each treated as an independent sample. The dataset was split into training, validation, and test sets (70/15/15) to develop and evaluate machine-learning models (XGBoost, TimesNet, and Transformers). The primary outcome was prediction of a clinically confirmed PIRA event.
Results:
Ninety-nine patients (median age 47 years, 63% women, median Expanded Disability Status Scale 2.5) completed follow-up; 14 developed PIRA. Baseline characteristics were comparable between groups. Time-series decomposition revealed distinct dynamics (Akaike Information Criterion [AIC]: 69.7 vs 108.0) between patients who developed PIRA and those who did not. A total of 25,440 temporal windows were generated. Among all models, TimesNet achieved the best performance, with accuracy 99.7%, precision 97% (95% CI 94–99), sensitivity 100% (95% CI 95–100), specificity 100% (95% CI 96–100), and F1-score 98%.
Conclusions:
Passive smartphone monitoring enables continuous assessment of motor function in MS. TimesNet outperformed other models for early PIRA detection, supporting its potential as a digital biomarker of progression risk. These findings warrant prospective multicenter validation for individualized disease monitoring.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.