How an academic medical center integrated AI into its Epic EHR to improve diagnoses
Emergency room physicians are almost always asking themselves, “What’s in front of me right now? Is this independent from the problem the patient has or is it part of the problem?” To answer these questions effectively, the physicians must know a lot about many different diseases.
“To master skin diagnoses, for example, I read a lot of atlases and earlier books from dermatologists, but the problem myself and my colleagues would run into is that we had the information available but not the visual component to answer the question, ‘What is it I’m looking at?’” said Brian Browne, MD, chair of the department of emergency medicine at the University of Maryland.
To overcome these challenges, Browne and his team deployed an imaging-based, artificial intelligence-powered clinical decision support tool from vendor VisualDx. The tool uses more than 100,000 medial images, available through a desktop or a smartphone.
Clinical decision support is a thriving area in the field of health IT. Vendors include Automated Clinical Guidelines, Epic, Information Builders, National Decision Support Company, Outcome Health, Pepid, TransformativeMed and Wolters Kluwer.
“It gives us more than any medical images we could find on the Internet and gets us over the challenge of having so many images to look at but not the informational context of what it actually is that we’re looking at in the ER,” Browne said.
"The best part is that we’re still bringing our expertise as trained physicians to the table when diagnosing so it truly is a tool that helps us to do our jobs more effectively and safely."
Brian Browne, MD, University of Maryland
VisualDx works by using artificial intelligence to compare hundreds of thousands of medical images against the symptoms entered by physicians into the platform. The tool then provides a physician with a few different possibilities for what the disease may be, and it enables users to add in additional symptoms and patient demographics to deliver a more accurate diagnosis, Browne said.
Additionally, the University of Maryland has VisualDx integrated into its Epic EHR such that clinicians can use it via the electronic health record, or on desktop or mobile devices.
“The best part is that we’re still bringing our expertise as trained physicians to the table when diagnosing so it truly is a tool that helps us to do our jobs more effectively and safely,” Browne added.
Browne recently had a case where a woman in her mid-fifties had come into the emergency room with an unusual growth on one of her fingers. She had been trying for weeks to care for the issue at home, which included cutting the growth off and any other means she felt necessary. All of which were to no avail as the growth continued to return and get bigger.
“She finally came into the ER and shared her story with us,” he said. “I had a few suspicions in mind for what the growth may be, but the diagnosis wasn’t immediately clear.”
Browne pulled out his smartphone to I had used the software on my smartphone to confirm narrow the diagnosis down to a few variables.
“After filling in the demographics, sharing a photo and receiving possible diagnoses back, I was relieved to have the second opinion on what it actually was I was looking at,” Browne said.
He had believed he was seeing the first diagnosis recommended, which was much more life-threatening, but it was actually the second diagnosis option presented, a pyogenic granuloma.
“The biggest takeaway from this scenario other than coming to the correct conclusion was that I was also able to build a more trusting relationship with the patient,” he said. “I had been using the tool in front of her and guiding her through the final diagnosis I had come to on the platform. This made her feel much more secure that this is exactly what she had, and it eliminated the feelings she was having of being alone in the diagnosis process.”
Browne was also able to treat her issue within the emergency department and avoid sending her to another physician or running countless, potentially unnecessary tests.
This example highlights what the imaging-based clinical decision support technology brings to the table when it comes to improving care and trimming costs.