: Background: Acute hypertensive disorders, including hypertensive emergencies (HEs) and urgencies (HUs), are a frequent cause of emergency department (ED) visits. Early differentiation between HEs and HUs is essential, as their clinical management and prognostic implications differ substantially. Methods: We retrospectively analyzed patients admitted to an Italian second-level ED between January and June 2022 with systolic blood pressure (SBP) ≥ 180 mmHg and/or diastolic blood pressure(DBP) ≥ 110 mmHg. Patients were categorized based on the presence of acute hypertension-mediated organ damage (A-HMOD). To identify the main predictors of HEs, we applied both conventional logistic regression and machine learning approaches (Elastic Net and Random Forest). Results: Among 23,678 ED admissions, 261 patients (1.1%) had acute hypertensive disorders, of whom 115 (44%) were diagnosed with HEs and 146 (56%) with HUs. Compared with HU patients, HE patients were older and showed higher SBPand DBP at presentation, along with a greater prevalence of comorbidities such as diabetes, coronary artery disease, and chronic kidney disease (all p < 0.05). In multivariable logistic regression, troponin I levels independently predicted the occurrence of HEs (OR: 2.82; 95%CI: 1.65-4.82; p < 0.001), even after adjusting for confounders. Machine learning analyses confirmed troponin I as the most influential predictor, followed by age and SBP, with the Random Forest model achieving a high predictive performance (AUCROC: 0.93; 95%CI: 0.90-0.96). Elastic Net regression further highlighted troponin I as the most influential variable with the highest standardized coefficient (β = 4.13). As determined by the Youden index, the optimal diagnostic threshold for troponin I was 0.12 ng/mL (AUCROC: 0.66; 95%CI: 0.60-0.72). Conclusions: In patients presenting to the ED, withacute hypertensive disorders, elevated troponin I levels, older age, and higher SBP at admission may serve as early indicators of emergencies.
Predictive Modeling of Acute Hypertensive Disorders in a Real-World Cohort: Integrating Clinical Predictors and Data-Driven Methods / Fucile, Ilaria; Liccardi, Filomena; Manzi, Maria Virginia; Lembo, Maria; Basile, Christian; Santucci, Orlando; Auciello, Stefania; Maniscalco, Mauro; Spedicato, Giorgio Alfredo; Morisco, Carmine; Izzo, Raffaele; De Luca, Nicola; Ambrosino, Pasquale; Mancusi, Costantino; Esposito, Giovanni; Paladino, Fiorella. - In: DIAGNOSTICS. - ISSN 2075-4418. - 15:16(2025). [10.3390/diagnostics15162062]
Predictive Modeling of Acute Hypertensive Disorders in a Real-World Cohort: Integrating Clinical Predictors and Data-Driven Methods
Fucile, Ilaria;Manzi, Maria Virginia;Lembo, Maria;Basile, Christian;Auciello, Stefania;Maniscalco, Mauro;Morisco, Carmine;Izzo, Raffaele;De Luca, Nicola;Ambrosino, Pasquale;Mancusi, Costantino;Esposito, Giovanni;
2025
Abstract
: Background: Acute hypertensive disorders, including hypertensive emergencies (HEs) and urgencies (HUs), are a frequent cause of emergency department (ED) visits. Early differentiation between HEs and HUs is essential, as their clinical management and prognostic implications differ substantially. Methods: We retrospectively analyzed patients admitted to an Italian second-level ED between January and June 2022 with systolic blood pressure (SBP) ≥ 180 mmHg and/or diastolic blood pressure(DBP) ≥ 110 mmHg. Patients were categorized based on the presence of acute hypertension-mediated organ damage (A-HMOD). To identify the main predictors of HEs, we applied both conventional logistic regression and machine learning approaches (Elastic Net and Random Forest). Results: Among 23,678 ED admissions, 261 patients (1.1%) had acute hypertensive disorders, of whom 115 (44%) were diagnosed with HEs and 146 (56%) with HUs. Compared with HU patients, HE patients were older and showed higher SBPand DBP at presentation, along with a greater prevalence of comorbidities such as diabetes, coronary artery disease, and chronic kidney disease (all p < 0.05). In multivariable logistic regression, troponin I levels independently predicted the occurrence of HEs (OR: 2.82; 95%CI: 1.65-4.82; p < 0.001), even after adjusting for confounders. Machine learning analyses confirmed troponin I as the most influential predictor, followed by age and SBP, with the Random Forest model achieving a high predictive performance (AUCROC: 0.93; 95%CI: 0.90-0.96). Elastic Net regression further highlighted troponin I as the most influential variable with the highest standardized coefficient (β = 4.13). As determined by the Youden index, the optimal diagnostic threshold for troponin I was 0.12 ng/mL (AUCROC: 0.66; 95%CI: 0.60-0.72). Conclusions: In patients presenting to the ED, withacute hypertensive disorders, elevated troponin I levels, older age, and higher SBP at admission may serve as early indicators of emergencies.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
