TechDoctor Announces Rheumatoid Arthritis Wearable Data Study Results at International Conferences, Led by Keio University

TechDoctor Inc., as a participating research and development organization, announced the results of a wearable device study for rheumatoid arthritis patients led by the Department of Internal Medicine (Rheumatology and Collagen Disease) at Keio University School of Medicine. The findings were presented at two international conferences in June 2026.

This study analyzed the relationship between data obtained using Google Fitbit wearable devices—such as physical activity, sleep, and heart rate variability (HRV)—and patient-reported outcomes (PROs). It further demonstrated that by using machine learning, it is possible to objectively estimate a patient's subjective Quality of Life (QOL) and fatigue.

The correlation between the QOL metric 'EQ-5D' and wearable data was presented at the 'EULAR 2026 Congress', while the correlation with fatigue evaluation metrics 'FACIT-F' and 'BFI' was presented at the 'FOCIS 2026 Annual Meeting'.

This research was conducted with support from the Japan Agency for Medical Research and Development (AMED).

■ Background and Research Overview

In rheumatoid arthritis patients, symptoms like fatigue, insomnia, and depression are known to impact daily life, in addition to joint symptoms. These are critical factors directly linked to a patient's Quality of Life (QOL). While questionnaire-based PRO evaluations are important, they are limited by patient memory and response timing, making it difficult to capture fluctuating symptoms objectively in real-time.

In this study, continuous daily life data was collected from 107 rheumatoid arthritis patients using the 'Fitbit Sense2' wristband device. The study analyzed the relationship between this wearable-derived data and the QOL metric 'EQ-5D-5L', as well as the fatigue metrics 'FACIT-F' and 'BFI'. It also verified whether machine learning could estimate these subjective states from objective data. SHAP analysis was also conducted to enhance model interpretability.

■ Key Research Findings

1. Correlation confirmed between wearable-derived data and QOL metric (EQ-5D)
The analysis showed that the EQ-5D utility value correlated with daytime HRV and sleep HRV, among others. A machine learning model built to classify QOL status from wearable data showed high discrimination performance, with an AUC-ROC between 0.75 and 0.89. SHAP analysis identified daytime HRV and sleep duration as key predictors.

2. Correlation confirmed between wearable-derived data and fatigue metrics (FACIT-F/BFI)
The results showed that the FACIT-F score correlated with resting heart rate and nocturnal HRV, while the BFI score correlated with daytime and nocturnal HRV. Machine learning models to classify severe vs. non-severe fatigue groups showed good performance, with an ROC-AUC of 0.88 for the FACIT-F model and 0.82 for the BFI model.

These results indicate that wearable-derived data has the potential to serve as an indicator for objectively and continuously capturing subjective symptoms like QOL and fatigue in rheumatoid arthritis patients.

■ Societal Significance and Future Outlook

This research suggests that biometric data from daily life, acquired through wearable devices, can be used as digital biomarkers to objectively and continuously monitor QOL and fatigue in rheumatoid arthritis patients. This could enable real-time, low-burden tracking of daily condition changes, which has been difficult with traditional questionnaires alone, and is expected to be applied to patient monitoring, disease management, and treatment effect evaluation.

TechDoctor will continue to promote digital biomarker research using wearable data and AI/machine learning, striving to advance data utilization in the medical field and realize patient-centered healthcare.