Artificial Intelligence (AI) has the ability to play a key role in clinical decision-making, leading to better patient outcomes and a reduction in hospital costs. When people think of the term ‘AI’ the first thing that usually comes to mind is — robotics. While that’s certainly a part of it, AI is an umbrella term that refers to any sort of application where a computer can learn to infer an outcome based on the data it’s presented with. Even in healthcare, AI utilization ranges from identifying lesions in brain scans to automating the billing process.
AI and Clinical Decision Making
Artificial neural networks, an advanced form of machine learning, are used in clinician decision support systems to scour electronic health records in order to identify potential gaps in treatment and help reduce medical errors. Patient monitoring is another field where AI has taken hold, employing data from smartphones and fitness monitoring devices in order to make predictions about a person’s wellbeing in addition to being used in ICUs to monitor vital signs. While AI can predict outcomes on par with clinicians, and in certain areas even better than clinicians, these tools are not meant to replace clinicians, but rather be used alongside them in order to provide the best possible care for the patient. The rise of AI in healthcare delivery has helped tailor treatment towards patient-specific needs which can increase positive outcomes and improve clinical workflow, while also reducing costs.
AI Algorithms During COVID-19
One interesting example has to do with the current COVID-19 pandemic. The number of articles published on the virus has been so overwhelming for researchers that the White House decided to work with companies that develop AI algorithms that can mine the articles’ text using AI techniques such as natural language processing (NLP). These algorithms allow a user to search keywords and quickly find articles relevant to their search, which has the potential to save time, money, and most importantly, lives. Likewise, researchers at Mount Sinai have used computer vision in order to identify possible COVID-19 patients via CT scans. This type of algorithm uses deep learning techniques in order to analyze large amounts of images, along with static patient information, and gives a probability that a given patient has COVID-19. While this type of AI algorithm cannot be used on its own, it aids clinicians in their decision making and speeds up the process.
Machine Learning at HRS
Utilizing AI in telehealth allows providers to make use of data already being collected by telehealth devices, alongside static patient information from EMRs, in order to better predict patient outcomes. Currently at HRS, we are working on a machine learning project that will aid clinicians in their decision making concerning high risk patients. While the majority of patients on telehealth are considered ‘high risk’, we hope to identify trends in patients’ data, including biometric readings, prior readmissions, engagement information, and adherence, that could inform us of whether or not a patient is at risk of having an adverse event that day.
Our current system flags a patient as having a high risk for readmission whenever they take a metric that is outside of the normal threshold. In these cases, the clinician is able to react and communicate with the patient to better understand their symptoms, and decide on next steps. This method, while effective, can be improved by leveraging machine learning techniques. Through machine learning, a high risk of readmission prediction will be made based on multiple variables, both static and longitudinal, rather than a single data point. This allows for a more comprehensive analysis of how the patient is doing and will help clinicians use their time more effectively by directing their attention towards patients who are most at-risk.
According to the Agency for Healthcare Research and Quality (AHRQ), the average readmission cost per patient is approximately $14,400. Moreover, patients with CHF and hypertension, who make up 41% and 22% of HRS’ patient population, respectively, cost $15,900 on average. Through the use of AI, we hope to create a more comprehensive system that will better predict whether a patient is at high risk for a readmission.
References:
AI in Healthcare: 4 Examples in Health Informatics. (2019, August 19). Retrieved July 20, 2020, from https://healthinformatics.uic.edu/blog/ai-in-healthcare-4-examples-in-health-informatics/
Bailey, M. K., Weiss, A. J., Barrett, M. L., & Jiang, H. J. (2019, February). Characteristics of 30-Day All-Cause Hospital Readmissions, 2010-2016. Retrieved July 20, 2020, from https://www.hcup-us.ahrq.gov/reports/statbriefs/sb248-Hospital-Readmissions-2010-2016.jsp
Banerjee, A., Chakraborty, C., Kumar, A., & Biswas, D. (2020). Emerging trends in IoT and big data analytics for biomedical and health care technologies. Handbook of Data Science Approaches for Biomedical Engineering, 121-152. doi:10.1016/b978-0-12-818318-2.00005-2
Das, N., Topalovic, M., & Janssens, W. (2018). Artificial intelligence in diagnosis of obstructive lung disease. Current Opinion in Pulmonary Medicine, 24(2), 117-123. doi:10.1097/mcp.0000000000000459
Hutson, M. (2020). Artificial-intelligence tools aim to tame the coronavirus literature. Nature. doi:10.1038/d41586-020-01733-7
Mei, X., Lee, H., Diao, K. et al. Artificial intelligence–enabled rapid diagnosis of patients with COVID-19. Nat Med (2020). https://doi.org/10.1038/s41591-020-0931-3
Reddy, S., Fox, J., & Purohit, M. P. (2018). Artificial intelligence-enabled healthcare delivery. Journal of the Royal Society of Medicine, 112(1), 22-28. doi:10.1177/0141076818815510