A multicentre validation study of a smartphone application to screen hand arthritis

Area of Science:

• Rheumatology
• Artificial Intelligence
• Digital Health

Background:

• Hand arthritis assessment in primary care presents a significant clinical challenge.
• A previously developed screening tool utilizes machine learning algorithms for hand arthritis assessment.
• This study aimed to validate the screening tool's performance among rheumatologists.

Purpose of the Study:

• To assess the diagnostic validity of a machine learning-based screening tool for hand arthritis.
• To evaluate the tool's performance across different rheumatologists in a multicenter setting.

Main Methods:

• 248 new patients from 7 Australian rheumatology practices were enrolled.
• Data collected via smartphone app included hand photographs, a 9-part questionnaire, and wrist irritability.
• Machine learning models analyzed photographic and clinical data to screen for osteoarthritis, rheumatoid, and psoriatic arthritis.

Main Results:

• The tool demonstrated high accuracy in discriminating between hand arthritis types.
• Predictive performance for rheumatoid arthritis: 85.1% accuracy, psoriatic arthritis: 95.2% accuracy, osteoarthritis: 77.4% accuracy.
• Data capture time averaged under 3 minutes, indicating efficiency.

Conclusions:

• The multicenter study validates the screening tool's consistent performance across rheumatologists.
• The smartphone application effectively screens hand arthritis using AI and patient-reported data.
• This tool can assist primary care physicians in improving hand arthritis assessment and management.