ARTIFICIAL INTELLIGENCE FOR HYPERTENSION MANAGEMENT: A SYSTEMATIC REVIEW OF MEDICAL AND ENGINEERING LITERATURES. (April 2021)

• Record Type: Journal Article
• Authors: Tsoi, K.; Lee, H. W. Y.; Yiu, K. K. L.; Leung, C. L. T.; Wong, S. Y. S.
• Is Part Of: Journal of hypertension. Volume 39(2021)e-Supplement 1
• Publication Date: 2021-04
• Subjects: Hypertension -- Periodicals; Hypertension -- Periodicals; 616.132005
• DOI: 10.1097/01.hjh.0000745464.94515.a5
• Languages: English
• ISSNs: 1473-5598

Abstract

Abstract : Objective: The present study aims to review and clarify the current evidence on AI applications for hypertension management from medical and engineering perspectives. Methods: A literature search was conducted on major electronic databases: Embase, MEDLINE, CINAHL with keywords 'artificial intelligence', 'hypertension', and 'blood pressure'. The quality of included studies was assessed with PROBAST tool. Included studies were categorised into (i) software and hardware development, (ii) disease prediction, and (iii) treatment recommendation. Results of individual studies were summarized for comparison. Design and method: A total of 92 articles were included for study, with more than half originated in Asia. Most studies involved blood pressure (BP) measurement with studies in engineering journals concerning the development of hardware and software, whereas medical researchers largely focused on the clinical use of these AI solutions. Electronic signals are the mainstay of BP measurement, although attempts were also made to extract useful information from health records. About a third of the studies were devoted to the identification of, and predicting the incidence and outcomes of hypertension using AI. In particular, studies that aimed to identify hypertension based on electronic signals achieved an AUC between 0.7 and 0.9, very similar to those concerning the incidence and prognosis of hypertension, despite different types of data and algorithms used. Studies on BP profiling and treatment with AI remained scant yet insightful. Conclusion: While a number of novel solutions have been invented for BP measurement and management, closer interdisciplinary collaboration between clinicians and engineers should further be encouraged to streamline BP profiling and treatment, personalised treatments with the aid of AI for example.