Development and Validation of a Cost-Effective Machine Learning Model for Screening Potential Rheumatoid Arthritis in...

Area of Science:

• Rheumatology
• Artificial Intelligence in Healthcare
• Primary Care Medicine

Background:

• Diagnosing rheumatoid arthritis (RA) in primary care is challenging due to limited practitioner knowledge and tools, leading to frequent missed diagnoses.
• Delays in RA diagnosis negatively impact patient prognosis and increase healthcare costs.

Purpose of the Study:

• To develop and validate a cost-effective screening model for early RA detection in primary healthcare settings.
• To improve the accuracy and efficiency of RA screening, reducing diagnostic delays.

Main Methods:

• A cohort of 2106 participants was used to develop a screening model with 26 clinical features.
• Ten machine learning algorithms were evaluated, with the best model selected based on performance metrics.
• Feature selection identified key indicators for RA screening, and the model was validated on primary healthcare datasets.

Main Results:

• The Random Forest (RF) algorithm achieved high accuracy (96.20%) in initial model development.
• The validated RF model demonstrated 88.44% accuracy, 98.55% sensitivity, and 85.56% specificity in primary healthcare settings.
• Eleven key features were identified for efficient RA screening.

Conclusions:

• The developed screening model automates prompt RA identification in primary care, enhancing early detection.
• This approach significantly reduces diagnostic delays and associated costs, improving healthcare efficiency.
• The findings support improved RA management and healthcare system responsiveness.