Use of machine learning to predict hypertension based on BMI and routine lab data

Area of Science:

• Biomedical Informatics
• Cardiovascular Disease Research
• Machine Learning in Healthcare

Background:

• Hypertension diagnosis relies on traditional metrics, often requiring invasive procedures.
• Early detection of hypertension is crucial for preventing severe cardiovascular complications.
• Integrating machine learning with routine clinical data offers a novel screening approach.

Purpose of the Study:

• To evaluate the efficacy of a Random Forest machine learning model in classifying hypertension.
• To determine if non-invasive data like BMI and laboratory characteristics can predict hypertension.
• To assess the potential of machine learning for enhancing hypertension screening.

Main Methods:

• Trained multiple machine learning models using clinical variables from 150 participants.
• Key variables included fasting glucose, low-density lipoprotein (LDL), and body mass index (BMI).
• Utilized a Random Forest algorithm for classification analysis.

Main Results:

• The Random Forest model achieved the highest predictive accuracy at 88.0%.
• The model demonstrated the feasibility of using readily available clinical data for hypertension classification.
• Significant variables included fasting glucose, LDL, and BMI.

Conclusions:

• Machine learning, specifically Random Forest, can effectively classify hypertension using non-invasive data.
• This approach holds potential for early hypertension detection and improved patient screening.
• Machine learning integration can enhance the effectiveness of cardiovascular disease screening programs.