Predictive Modeling of Acute Hypertensive Disorders in a Real-World Cohort: Integrating Clinical Predictors and...

Area of Science:

• Cardiology
• Emergency Medicine
• Hypertension Research

Background:

• Acute hypertensive disorders, including hypertensive emergencies and urgencies, are common reasons for emergency department visits.
• Distinguishing between hypertensive emergencies and urgencies is critical due to differing clinical management and prognoses.

Purpose of the Study:

• To identify key predictors for differentiating hypertensive emergencies from hypertensive urgencies in patients presenting to the emergency department.
• To evaluate the utility of logistic regression and machine learning models in predicting hypertensive emergencies.

Main Methods:

• Retrospective analysis of 261 patients with acute hypertensive disorders (SBP ≥ 180 mmHg or DBP ≥ 110 mmHg) admitted to an Italian emergency department.
• Application of logistic regression, Elastic Net, and Random Forest models to identify predictors of hypertensive emergencies.
• Categorization of patients based on the presence of acute hypertension-mediated organ damage.

Main Results:

• Elevated troponin I levels independently predicted hypertensive emergencies (OR: 2.82; p < 0.001).
• Machine learning models identified troponin I, age, and SBP as primary predictors, with Random Forest showing high performance (AUC: 0.93).
• An optimal troponin I threshold of 0.12 ng/mL was identified for diagnosing hypertensive emergencies (AUC: 0.66).

Conclusions:

• Elevated troponin I levels, older age, and higher SBP are significant early indicators of hypertensive emergencies in the emergency department setting.
• These findings aid in the timely differentiation between hypertensive emergencies and urgencies, guiding appropriate clinical management.
• Machine learning approaches demonstrate promise in enhancing the predictive accuracy for hypertensive emergencies.