Predicting osteoarthritis in adults using statistical data mining and machine learning

Area of Science:

• Rheumatology
• Epidemiology
• Data Science

Background:

• Osteoarthritis (OA) is typically seen in older adults but can affect younger individuals.
• Risk factors for OA in adults aged 20-50 require further investigation.

Purpose of the Study:

• To develop a predictive model for OA risk factors in adults aged 20-50.
• To compare the performance of various machine learning models for OA risk prediction.

Main Methods:

• Utilized data from 19,133 participants (aged 20-50) from the National Health and Nutrition Examination Survey.
• Employed supervised machine learning models including logistic regression, deep neural networks (DNN), and support vector machines.
• Identified demographic and personal characteristics associated with OA.

Main Results:

• Key risk factors identified: female sex, older age, smoking, higher body mass index, high blood pressure, race/ethnicity, and physical/mental limitations.
• The model achieved a 75% area under the receiver operating characteristic curve for predictive performance.
• Deep neural network (DNN) models demonstrated the best predictive performance.

Conclusions:

• Female sex, older age, smoking, elevated BMI, hypertension, specific race/ethnicities, and functional limitations are significant risk factors for OA in younger adults.
• Machine learning, particularly DNNs, can effectively identify individuals at risk for early-onset OA.