Objective:

To predict the risk of angina attacks in high temperature environments for patients with stable angina pectoris using machine learning models.

Approach:

• Cohort Description: The derivation cohort consisted of 1,246 individuals with stable angina treated at 43 clinical research centers in China.
• Model Development: Variable selection was performed using the Boruta algorithm, and seven machine learning algorithms were developed and evaluated using metrics such as area under the receiver operating characteristic curve (AUC), sensitivity, specificity, F1 score, calibration analysis, and decision curve analysis.
• Validation: An independent external validation cohort of 120 patients from 5 additional centers was used to assess model generalizability.
• Post Hoc Analysis: A post hoc analysis of a multi-center clinical trial was conducted to validate the model's clinical utility.

Key Findings:

• 14 predictors were identified for the risk factors in the ML model.
• The random forest (RF) model showed strong performance in training and internal validation cohorts.
• External validation confirmed the RF model's predictive performance and robustness.
• Patients identified by the RF model as having high HSA probability showed increased angina frequency during summer months.

Interpretation:

The RF model demonstrated promising predictive performance for identifying patients susceptible to HSA, utilizing various clinical and serological markers.

Limitations:

• Further prospective multicenter validation is required before routine clinical application.

Conclusion:

Further prospective multicenter validation is required before routine clinical application.

Sources:

• Clinical Trial Registration
• Clinical Trials.gov