Designing a Machine Learning Model to Predict Cardiovascular Disease Without Any Blood Test
Healthcare in the USA is struggling with alarming levels of hospital readmission. Cardio Vascular Disease (CVD) has been identified as the most frequent cause. While the factors related to high hospital readmission are complex, according to prior research, early detection and post-discharge management has a significant positive impact. However, the widening gap between the number of patients and available clinical resources is acutely aggravating the problem. A solution that can effectively identify well patients at risk of future CVDs will allow focusing limited clinical resources to a more targeted set of patients, leading to more widespread early detection, prevention and disease progression management. This in turn, can reduce CVD-related hospital readmissions. Moreover, if the patient data required by such a solution can be collected without any blood test or invasive procedure, the addressable patient population can be vastly expanded to include home care, remote, and impoverished patients while delivering cost savings of the invasive procedures. Using a Design Science Research (DSR) approach, this research has led to the design and development of a machine learning based predictor artifact capable of identifying patients with future CVD risks. The performance of this predictor artifact, as measured by the area under the receiver operating characteristic (ROC) curve, is 0.859. The sensitivity or recall is 85.9% at probability threshold of 0.5. The significant differentiating feature of this artifact lies in its ability to do so without any blood test or invasive procedure.
KeywordsCardiovascular disease Design science research Machine learning Healthcare
Co-author Dr. Chatterjee was funded in part by a fellowship from the Schoeller Research Center of Nuremberg, Germany. We would also like to thank Dr. Luanda Grazette of USC Keck School of Medicine for helping us with the knowledge of cardiology domain.
- 1.World Health Organization. Cardiovascular Diseases (CVD) Fact Sheet. https://www.who.int/en/news-room/fact-sheets/detail/cardiovascular-diseases-(cvds)
- 3.World Health Organization: Global strategy on human resources for health: Workforce 2030 (2016)Google Scholar
- 4.Colantonio, S., Martinelli, M., Moroni, D., Salvetti, O., Chiarugi, F., Emmanouilidou, D.: A decision support system for aiding heart failure management. In: 2009 Ninth International Conference on Intelligent Systems Design and Applications 2009, pp. 351–356. IEEE (2009)Google Scholar
- 6.Sourla, E., Sioutas, S., Syrimpeis, V., Tsakalidis, A., Tzimas, G.: CardioSmart365: artificial intelligence in the service of cardiologic patients. Adv. Artif. Intell. 2012, 2 (2012)Google Scholar
- 7.Zhang, J., Goode, K.M., Rigby, A., Balk, A.H., Cleland, J.G.: Identifying patients at risk of death or hospitalisation due to worsening heart failure using decision tree analysis: evidence from the Trans-European Network-Home-Care Management System (TEN-HMS) study. Int. J. Cardiol. 163(2), 149–156 (2013)CrossRefGoogle Scholar
- 9.Choi, E., Schuetz, A., Stewart, W.F., Sun, J.: Using recurrent neural network models for early detection of heart failure onset. J. Am. Med. Inform. Assoc. 24(2), 361–370 (2016)Google Scholar
- 14.World Health Organization: Types of Cardiovascular disease. https://www.who.int/cardiovascular_diseases/en/cvd_atlas_01_types.pdf
- 16.ASCVD Risk Estimator Plus. http://tools.acc.org/ASCVD-Risk-Estimator-Plus/#!/calculate/estimate/
- 18.Sleep Heart Health Study. https://sleepdata.org/datasets/shhs