Abstract

In a hospital environment, patients are monitored continuously by electronic devices and health professionals. Therefore, a large amount of data is collected and stored in electronic health records systems for each patient. Among such data, vital signs are one of the most common and relevant types of information monitored to assess a patient’s health status. Artificial intelligence techniques can be used to analyze and learn useful standards from clinical datasets to provide better evidence to support the decisions of health professionals and thus help to improve patient health outcomes in hospitals. This systematic literature review aims to provide an updated computational perspective of how artificial intelligence has been applied to analyze the vital signs of adult hospitalized patients and the outcomes obtained. To this end, we reviewed 2899 scientific articles published between 2008 and 2018 and selected 78 articles that met our inclusion criteria to answer the research questions. Moreover, we used the information found in the reviewed articles to propose a taxonomy and identified the main concerns, challenges, and opportunities in this field. Our findings demonstrate that many researchers are exploring the use of artificial intelligence methods in tasks related to improving the health outcomes of hospitalized patients in distinct units. Additionally, although vital signs are significant predictors of clinical deterioration, they are not analyzed in isolation to predict or identify a clinical outcome. Our taxonomy and discussion contribute to the achievement of a significant degree of coverage regarding the aspects related to using machine learning to improve health outcomes in hospital environments, while highlighting gaps in the literature for future research.