Artificial Intelligence-Derived Risk Prediction: A Novel Risk Calculator Using Office and Ambulatory Blood Pressure

Pedro Guimarães; Andreas Keller; Michael Böhm; Lucas Lauder; Tobias Fehlmann; Luis M Ruilope; Ernest Vinyoles; Manuel Gorostidi; Julián Segura; Gema Ruiz-Hurtado; Natalie Staplin; Bryan Williams; Alejandro de la Sierra; Felix Mahfoud

doi:10.1161/HYPERTENSIONAHA.123.22529

Artificial Intelligence-Derived Risk Prediction: A Novel Risk Calculator Using Office and Ambulatory Blood Pressure

Hypertension. 2024 Apr 25. doi: 10.1161/HYPERTENSIONAHA.123.22529. Online ahead of print.

Authors

Pedro Guimarães^{1

2}, Andreas Keller^{1

3}, Michael Böhm⁴, Lucas Lauder⁴, Tobias Fehlmann¹, Luis M Ruilope^{5

6

7}, Ernest Vinyoles^{8

9}, Manuel Gorostidi¹⁰, Julián Segura⁵, Gema Ruiz-Hurtado^{5

6}, Natalie Staplin¹¹, Bryan Williams¹², Alejandro de la Sierra¹³, Felix Mahfoud^{4

14}

Affiliations

¹ Chair for Clinical Bioinformatics, Saarland University, Saarbrücken, Germany (P.G., A.K., T.F.).
² University of Coimbra, Coimbra Institute for Biomedical Imaging and Translational Research, Institute for Nuclear Sciences Applied to Health, Portugal (P.G.).
³ Department of Neurology and Neurological Sciences, Stanford University, CA (A.K.).
⁴ Department of Internal Medicine III, Cardiology, Angiology, Intensive Care Medicine, Universitätsklinikum des Saarlandes, Saarland University, Homburg/Saar, Germany (M.B., L.L., F.M.).
⁵ Cardiorenal Translational Laboratory and Hypertension Unit, Institute of Research i+12, Hospital Universitario 12 de Octubre, Madrid, Spain. (L.M.R., J.S., G.R.-H.).
⁶ CIBER-CV, Hospital Universitario 12 de Octubre, Madrid, Spain. (L.M.R., G.R.-H.).
⁷ Faculty of Sport Sciences, European University of Madrid, Spain (L.M.R.).
⁸ La Mina Primary Care Center, University of Barcelona, Spain (E.V.).
⁹ IDIAP Jordi Gol, Barcelona, Spain (E.V.).
¹⁰ Department of Nephrology, Hospital Universitario Central de Asturias, REDinREN, Oviedo, Spain (M.G.).
¹¹ Medical Research Council Population Health Research Unit, Clinical Trial Service Unit and Epidemiological Studies Unit, Nuffield Department of Population Health, University of Oxford, United Kingdom (N.S.).
¹² University College London (UCL), Institute of Cardiovascular Science, National Institute for Health Research, UCL Hospitals Biomedical Research Centre, United Kingdom (B.W.).
¹³ Department of Internal Medicine, Hospital Universitario Mútua Terrasa, Universidad de Barcelona, Spain (A.d.l.S.).
¹⁴ Institute for Medical Engineering and Science, MIT, Cambridge, MA (F.M.).

PMID: 38660828
DOI: 10.1161/HYPERTENSIONAHA.123.22529

Abstract

Background: Quantification of total cardiovascular risk is essential for individualizing hypertension treatment. This study aimed to develop and validate a novel, machine-learning-derived model to predict cardiovascular mortality risk using office blood pressure (OBP) and ambulatory blood pressure (ABP).

Methods: The performance of the novel risk score was compared with existing risk scores, and the possibility of predicting ABP phenotypes utilizing clinical variables was assessed. Using data from 59 124 patients enrolled in the Spanish ABP Monitoring registry, machine-learning approaches (logistic regression, gradient-boosted decision trees, and deep neural networks) and stepwise forward feature selection were used.

Results: For the prediction of cardiovascular mortality, deep neural networks yielded the highest clinical performance. The novel mortality prediction models using OBP and ABP outperformed other risk scores. The area under the curve achieved by the novel approach, already when using OBP variables, was significantly higher when compared with the area under the curve of the Framingham risk score, Systemic Coronary Risk Estimation 2, and Atherosclerotic Cardiovascular Disease score. However, the prediction of cardiovascular mortality with ABP instead of OBP data significantly increased the area under the curve (0.870 versus 0.865; P=3.61×10^-²⁸), accuracy, and specificity, respectively. The prediction of ABP phenotypes (ie, white-coat, ambulatory, and masked hypertension) using clinical characteristics was limited.

Conclusions: The receiver operating characteristic curves for cardiovascular mortality using ABP and OBP with deep neural network models outperformed all other risk metrics, indicating the potential for improving current risk scores by applying state-of-the-art machine learning approaches. The prediction of cardiovascular mortality using ABP data led to a significant increase in area under the curve and performance metrics.

Keywords: artificial intelligence; blood pressure; machine learning; neural networks, computer; risk factors.