Construction and Comparison of Machine Learning-Based Risk Prediction Models for Major Adverse Cardiovascular Events...

Area of Science:

• Cardiology
• Endocrinology
• Women's Health

Background:

• Perimenopause involves declining ovarian function and estrogen, increasing cardiovascular disease risk.
• Major adverse cardiovascular events (MACE) encompass heart failure and myocardial infarction.
• Understanding MACE risk factors in perimenopausal women is crucial for preventive strategies.

Purpose of the Study:

• To identify factors influencing MACE occurrence in perimenopausal women.
• To develop and compare prediction models for MACE risk using three distinct algorithms.
• To evaluate the predictive performance of machine learning and logistic regression models.

Main Methods:

• 411 perimenopausal women with MACE were randomly assigned to training (70%) and testing (30%) sets.
• Random Forest (RF), Backpropagation Neural Network (BPNN), and Logistic Regression (LR) were employed to build MACE prediction models.
• Model performance was assessed using accuracy, sensitivity, specificity, and AUC.

Main Results:

• The RF model achieved an AUC of 0.948, BPNN an AUC of 0.921, and LR an AUC of 0.866.
• The RF model demonstrated significantly superior predictive performance compared to the LR model (P=0.023).
• Twenty-six candidate variables were analyzed for their predictive value.

Conclusions:

• The Random Forest model exhibits strong performance for predicting MACE risk in perimenopausal women.
• This model can assist in the early identification of high-risk individuals.
• Findings support the development of targeted interventions to mitigate MACE in this population.