Prediction of cardiovascular risk using machine-learning methods. Sex-specific differences.

Area of Science:

• Cardiology
• Data Science
• Public Health

Background:

• Machine learning (ML) offers advantages over traditional scoring systems for assessing cardiovascular risk factors (CVRFs) and major cardiovascular events (MACE).
• Real-world data (RWD) availability enables training ML algorithms for clinical practice.
• ML models can potentially improve personalized cardiovascular risk prediction.

Purpose of the Study:

• To evaluate major cardiovascular event (MACE) risk using XGBoost and Random Forest ML algorithms.
• To apply these algorithms to real-world data (RWD), stratifying by sex.
• To compare the performance of XGBoost and Random Forest in predicting MACE.

Main Methods:

• The study included 52,393 subjects with a follow-up from 2018 to 2020.
• Three models were generated for each ML algorithm (XGBoost, Random Forest), incorporating age and combinations of blood tests, CVRFs, and medication adherence.
• Cardiovascular risk factors (CVRFs) and medication adherence were analyzed.

Main Results:

• A total of 581 major cardiovascular events (MACE) occurred; incidence was 1% in women and 1.3% in men.
• Hypertension and hypercholesterolemia were the most prevalent CVRFs. Treatment adherence varied, being highest for antihypertensives and lowest for antidiabetics.
• Age was the primary contributor to MACE risk, followed by adherence to antidiabetics. Both ML algorithms showed similar performance.

Conclusions:

• Machine learning (ML) models effectively assess cardiovascular risk using real-world data (RWD).
• Medication adherence is a significant predictor of major cardiovascular events (MACE).
• ML application supports personalized prevention strategies in primary care.