The story below is a preview from our July/August 2024 issue. For more stories like it, Subscribe Today. Thank you!
Health care providers are using AI to augment, rather than substitute for, modern medical practices.
Since the release of ChatGPT in the fall of 2022, headlines and discussions surrounding artificial intelligence have largely centered on questions of increased plagiarism and job losses. Alternatively, the larger medical community in Roanoke views the wide-spread adoption of AI differently. In fact, many providers and systems see it as a means by which they can improve patient outcomes. However, they caution that the tool should be viewed as an augment to care, not a substitute.
Both of the Roanoke Valley’s large healthcare systems, LewisGale Medical Center and Carilion Clinic, have started adopting the use of AI in the practice of medicine.
LewisGale Medical Center
“I think we’re in the infancy of where AI can really provide benefit. Five years from now, or even two years from now, we’ll be having a very different conversation,” says Dr. Robert Alphin, chief medical officer at the LewisGale Medical Center.
Despite the uncertainty of the future, he views the use of AI in medicine as falling into one of three categories: providing clinical decision support to providers, increasing access to care and the unburdening of clinicians from administrative tasks.
HCA, the center’s parent company, sees 38 million patients yearly across the country. If it can integrate patient information into an algorithm, he says, it can help guide providers as they make decisions. That scenario, he says, is ideal compared to a local doctor who only has a small data set to turn to for diagnostic and treatment trends. “That’s going to allow patients to have quicker access to care and decisions,” he says.
Carilion Clinic
Carilion has been using forms of AI for quite some time, including predictive models for re-admissions. For roughly six years, it has used a machine learning model to predict the risks of someone deteriorating in a hospital. “It uses a lot of different metrics from our EMR, then gives a visualization of a patient who may be starting to deteriorate so we can do earlier interventions,” Dr. Stephen Morgan, senior vice president and chief medical information officer, says.
Carilion went all-in on AI in 2020 when it started a data science team, which has now grown to encompass three scientists, that is helping it navigate the growing field. “Their role is not only doing research in this area, but also helping to guide the conversation,” he says.
In working with the system’s EMR, or electronic medical records vendor, it has used a model to predict the risk of a patient having sepsis. “That has helped us to reduce mortality within Roanoke Memorial,” he says. Another model, an enhanced version, allows the system to reach out to patients that are at risk of readmission after they’ve left the hospital and do a seven-day follow-up.
The system started using large language models through its EMR vendor, which uses ChatGPT to draft messages that are sent through the patient portal. It also drafts responses for clinicians to send back to patients.
In March, the system adopted DAX, an iPhone-based model used to create clinical notes from patient conversations. Once a patient has consented, the clinician places the phone between them and the patient so that it can record. “Incredible technology,” he says of it. “These types of models will speed up the process for our patients to interact with our health system.”
The system plans to continue to review and adopt various AI developments as time goes on. “We have new ideas all the time. It’s exciting for us,” he says.
Treatment of Stroke Patients/Radiology
Dr. Dan Karolyi, senior vice president and chair of radiology at Carilion Clinic, views the use of AI as being a transformative technology for radiology.
The first use of AI in the radiology department at Carilion was helping physicians select the right imaging exam. Currently, when a physician places an order for an advanced imaging system, like MRI or CT, an AI software is running in the background to match the reason for the exam against what the expert consensus is for what imaging study would be best for the patient.
“If those things don’t match up, it will give feedback to the ordering doctor that there might be a more appropriate exam to order. That physician can either choose to go with that suggestion or choose to go with their original selection as well,” he explains.
The second implementation of AI in radiology is related to stroke imaging, which the system has been doing for roughly five years. When patients come in for a suspected stroke, they get several imaging exams rapidly to determine if they’re having a stroke or if they’re having some other abnormality that is causing the stroke-like symptoms. CT angiograms allow doctors to see if any of the vessels that provide blood to the brain are clogged.
“When we scan those images, the results are rapidly evaluated with artificial intelligence. If the artificial intelligence thinks that there might be a large vessel within the brain that’s occluded, it will alert the entire stroke team through an app on their phone,” Karolyi says. “The message says, ‘We may have an acute stroke patient that has one of their larger central vessels of their brain occluded.’”
This allows the stroke team to look at the images and validate that a stroke is occurring. “Sometimes artificial intelligence doesn’t always get it right, so oversight has to validate that that’s truly the case,” he says. It also allows the team to triage the patient to appropriate therapy, like IV clot-busting drugs.
“In that way, artificial intelligence isn’t really making the diagnosis—it’s making the stroke team aware that there might be a potentially critically ill patient in our system, and it allows us to activate treatment pathways for those patients much more rapidly than we otherwise could,” he says.
LewisGale Medical Center is using AI in a similar fashion. The center utilizes an emergency department AI application for the detection of large strokes and bleeding on the brain.
Upon arrival at the center, patients undergo CT scans, angiograms and geography studies that look at blood vessels. If the application detects a large stroke that can physically be intervened on, a notification pops up on the doctor’s phone within seconds. The application is also linked to transferring facilities so that images can be pushed more quickly. It also lets physicians start the treatment process more quickly while radiology specialists are still manually reading and doublechecking the film to identify possible concerns.
Want to learn more about the role AI plays in augmenting health care, including ethical considerations and insights from LewisGale's Stroke Medical Director Dr. Zach Williams? Check out the latest issue, now on newsstands, or see it for free in our digital guide linked below!
The story above is a preview from our July/August 2024 issue. For more stories like it, Subscribe Today. Thank you!