Application of machine learning algorithms to predict heat-sensitive angina (HSA) attacks: a multicentric...

Area of Science:

• Cardiology
• Environmental Health
• Machine Learning in Medicine

Background:

• Studies confirm a link between high temperatures and increased angina frequency in heat-sensitive cardiovascular disease patients.
• Existing cardiovascular risk models lack specificity for identifying susceptibility to heat-sensitive angina (HSA) exacerbation during hot weather.

Purpose of the Study:

• To develop and validate a predictive model for identifying patients susceptible to heat-sensitive angina (HSA) exacerbation.
• To create a decision-support tool for individualized heat-related cardiovascular risk assessment.

Main Methods:

• A derivation cohort of 1,246 stable angina patients was used to develop machine learning models, with variable selection via the Boruta algorithm.
• Seven machine learning algorithms were evaluated, with a random forest (RF) model showing strong performance.
• External validation in a separate cohort of 120 patients assessed model generalizability and robustness.

Main Results:

• A random forest (RF) model incorporating 14 predictors demonstrated strong predictive performance in training, internal, and external validation cohorts.
• Patients identified by the RF model as having high HSA probability showed increased angina frequency during summer months.
• The model's robustness was confirmed through external validation and post-hoc analysis.

Conclusions:

• An RF model utilizing region, MPA, DBP, BMI, constipation, body fat indicators, and seven serological markers shows promise for predicting HSA susceptibility.
• The developed RF model can serve as an exploratory decision-support tool for personalized heat-related cardiovascular risk assessment.
• Further prospective multicenter validation is necessary before routine clinical application of the model.