Clinical Scorecard: Creation and assessment of a deep learning model for forecasting coronary heart disease in patients with hypertension utilizing 24-hour ambulatory blood pressure monitoring: a retrospective analysis

At a Glance

Key Highlights

• Deep neural network model achieved an AUC of 0.822 in training and 0.796 in validation cohorts.
• Nine predictors identified, including diabetes mellitus and blood pressure control parameters.
• Time in target range (TTR) is a significant predictor for CHD risk in hypertensive patients.
• Model demonstrates superior calibration with the lowest Brier score of 0.172.
• SHAP analysis used for model interpretability.

Guideline-Based Recommendations

Diagnosis

• Utilize 24-hour ambulatory blood pressure monitoring to assess blood pressure control.

Management

• Incorporate time in target range (TTR) as a dynamic predictor in risk stratification for hypertensive patients.

Monitoring & Follow-up

• Regularly monitor blood pressure using ambulatory methods to capture diurnal patterns.

Risks

• High-risk hypertensive patients should be identified early to prevent cardiovascular events.

Patient & Prescribing Data

Hypertensive patients at risk for coronary heart disease.

Consider antihypertensive medications, including calcium channel blockers and β-blockers, in management strategies.

Clinical Best Practices

• Employ machine learning methods to enhance prognostic modeling in cardiovascular diseases.
• Use SHAP values for improved interpretability of predictive models.

Related Resources & Content

• European Society of Cardiology guidelines