Comparison of prediction model for cardiovascular autonomic dysfunction using artificial neural network and logistic...

Area of Science:

• Cardiology
• Biostatistics
• Machine Learning

Background:

• Cardiovascular autonomic (CA) dysfunction is a growing concern in the general population.
• Accurate prediction models are crucial for early detection and management of CA dysfunction.

Purpose of the Study:

• To develop and compare prediction models for CA dysfunction using artificial neural network (ANN) and multivariable logistic regression (LR) analyses.
• To evaluate the performance of ANN and LR in predicting CA dysfunction in a Chinese population sample.

Main Methods:

• Analysis of a dataset comprising 2,092 individuals aged 30-80 years.
• Development of prediction models using ANN and LR on an exploratory set.
• Validation of models and comparison of their predictive performance, including sensitivity, specificity, and predictive values.

Main Results:

• Univariate analysis identified 14 significant risk factors for CA dysfunction (P<0.05).
• Area under the receiver-operating curve (AUC) values were comparable: 0.758 for LR and 0.762 for ANN.
• Noninferiority was established (P<0.001), with similar sensitivity, specificity, and predictive values between the two models.

Conclusions:

• Both ANN and LR are effective tools for developing prediction models for CA dysfunction.
• The study demonstrates the utility of these computational approaches in cardiovascular health risk assessment.