Tracking of Systemic Lupus Erythematosus (SLE) Longitudinally Using Biosensor and Patient-Reported Data: A Report on...

Area of Science:

• Digital health
• Machine learning in medicine
• Rheumatology

Background:

• Systemic lupus erythematosus (SLE) is a complex autoimmune disease characterized by immune dysregulation, unpredictable flares, and significant impact on quality of life (QOL).
• Current monitoring methods for SLE can be limited, necessitating innovative approaches for early detection of disease flares.

Purpose of the Study:

• To develop and validate machine learning models for predicting disease flares in SLE patients using a decentralized digital study.
• To assess the utility of patient-reported outcomes (PROs), QOL metrics, and digital biometric data in predicting SLE flares.

Main Methods:

• Recruitment of SLE participants from an online community (LupusCorner).
• Utilized a mobile application for collecting patient profile (PP), PRO, and QOL data.
• Incorporated smartwatch data for digital biometric monitoring.
• Applied feature selection and classification algorithms for data analysis.

Main Results:

• A 26-feature model accurately classified participants based on disease flare risk identified from medical records (MRs).
• Receiver Operating Characteristic (ROC) curves demonstrated significant discrimination between true and false positives.
• A 25-feature Bayesian model achieved high accuracy in time-variant prediction of participant-reported flares.

Conclusions:

• Regular monitoring of patient well-being and biometric data can aid in proactive screening for SLE flares.
• Digital health tools and machine learning show promise for personalized SLE management.
• This study highlights the potential for remote monitoring to improve clinical assessment and patient outcomes in SLE.