Machine learning-aided risk prediction for metabolic syndrome based on 3 years study

Area of Science:

• Cardiology
• Endocrinology
• Public Health

Background:

• Metabolic syndrome (MetS) is a cluster of conditions increasing risks for diabetes and cardiovascular diseases.
• Accurate prediction and identification of MetS risk factors are vital for early intervention.
• Existing prediction models require enhancement for improved accuracy.

Purpose of the Study:

• To develop an improved risk prediction model for metabolic syndrome (MetS).
• To identify key risk factors contributing to MetS onset across different demographics.
• To evaluate the efficacy of novel features derived from longitudinal health data.

Main Methods:

• Utilized a large dataset (67,730 samples) of consecutive physical examination records over three years.
• Developed novel features: differential numerical features (DNF) and differential state features (DSF) from historical data.
• Statistically analyzed risk factors associated with these features concerning age and gender.

Main Results:

• The proposed differential state feature (DSF) significantly enhances MetS risk prediction when combined with numerical examination data.
• The novel feature set demonstrated superior performance compared to state-of-the-art MetS prediction models.
• Identified significant age and gender-specific risk factors for MetS.

Conclusions:

• The integration of differential state features offers a powerful tool for improving MetS risk prediction.
• The proposed prediction scheme shows potential for effective prescreening of MetS occurrence.
• This approach can aid in personalized risk assessment and early preventative strategies for metabolic disorders.