A Machine Learning-Based Preclinical Osteoporosis Screening Tool (POST): Model Development and Validation Study.

Area of Science:

• Gerontology
• Medical Informatics
• Public Health

Background:

• Osteoporosis screening is a priority for fracture prevention.
• Existing tools often lack specificity, leading to potential over-diagnosis or missed cases.
• Developing precise, accessible screening methods is crucial for effective healthcare.

Purpose of the Study:

• To create a high-performance, easily accessible preclinical risk screening tool for osteoporosis.
• To utilize a machine learning-based approach tailored for the Hong Kong Chinese population.
• To improve early detection and intervention for osteoporosis.

Main Methods:

• Machine learning models (gradient boosting, SVM, naive Bayes, logistic regression) were developed and compared.
• A validated questionnaire collected risk factors from 800 participants (aged ≥45).
• Bone mineral density (BMD) measured osteoporosis; the best model formed the Preclinical Osteoporosis Screening Tool (POST).

Main Results:

• The POST model identified 7 key predictors: age, gender, education, height loss, BMI, teeth loss, and vitamin D intake.
• The POST achieved an Area Under the Curve (AUC) of 0.86.
• POST demonstrated high sensitivity (0.83), specificity (0.83), and negative predictive value (0.98).

Conclusions:

• The machine learning-based POST is a convenient and accurate tool for osteoporosis prediction.
• POST shows potential for guiding population-based preclinical screening.
• The tool can assist in clinical decision-making for osteoporosis management.