A 50-year-old patient schedules their first screening colonoscopy. If a polyp is detected during the procedure, it will be removed. Fortunately, most are benign, but some are pre-cancerous and if left alone will go on to become colorectal cancer—the second leading cause of cancer death in the US.
In a new paradigm, the colonoscopist uses real-time artificial intelligence (AI) as a “second set of eyes” to help detect polyps. The AI algorithm uses a neural network model patterned after the human brain and has been trained with millions of images. The AI analyzes each frame in the live video and alerts the doctor by drawing a box around a potential polyp, which due to its size and color, may have been easily missed. With combined vigilance, the detection rate goes up and the risk of cancer in the interval between colonoscopies goes down.
Artificial intelligence is anything but artificial, and Joe Cummings, Ph.D., Vizient’s technology program director, believes AI technology is already transforming the delivery of patient care for the better. Dr. Cummings’ identifies and evaluates high-impact, innovative emerging technologies for use within members’ performance improvement initiatives via Performance Improvement Collaboratives and Member Networks. To support a growing need to understand and address this disruptive technology, Dr. Cummings regularly produces TechFlash educational reports to help members think “outside the box” when it comes to awareness and understanding of new technology.
TechFlash reports assess AI that’s disrupting health care
Four TechFlash reports have been produced this year with a specific type of AI technology as the main theme. Each report includes a technology overview and status, technology significance, current practices and alternatives, clinical evidence summary, financial considerations, patient selection criteria, future developments and recommendations.
Highlights in the reports provide evidence-based learning for members. For example:
- AI for real-time detection of polyps during colonoscopy (September 2021) —improves quality metrics. Real-time detection and analysis of polyps may improve the rate of colorectal cancer.
- AI for automated detection of diabetic retinopathy (August 2021)—improves access to screening. Rapid and convenient access may prevent patient blindness.
- AI-enabled ischemic stroke detection, triage, and notification (June 2021)—reduces workflow delays and improves patient outcomes. Impacts morbidity, mortality, functional and quality of life outcomes associated with stroke.
- Wearable sensors and AI algorithms for ambulatory ECG analysis (April 2021)—improves patient satisfaction, comfort and convenience. Can detect AFib early and prevent potential strokes. Wearables with AI may be as accurate as human readers.
“The high impact technology topics evaluated ultimately come from the membership,” Cummings said. “Much of my reporting is based on a systematic review of the clinical evidence for an identified performance improvement opportunity. The evaluation of that evidence, which is detailed in the TechFlash report, helps members decide whether or not to implement new technology.”
Dr. Cummings has also evaluated several laboratory instruments that can help with the rapid diagnosis and treatment of sepsis and presented that information to PI Collaborative members for consideration and adoption. The early identification of sepsis presents a huge opportunity for performance improvement. “It’s a high-cost condition and a major challenge for hospitals. It’s imperative that providers improve their performance in sepsis and AI may be able to help here as well.” Dr. Cummings insights on this technology will be available in an upcoming Knowledge on the Go podcast.
AI is transforming health care
Since 2018, there have been more than 100 AI-based medical technologies approved by the Food and Drug Administration. They include both standalone AI-based software and software incorporated within a medical device. At a rate of three-to-four approvals per month, AI use is ramping up significantly to be truly disruptive and transformative in health care.
Dr. Cummings noted more than 130 members are using the new AI stroke detection technology thanks to Centers for Medicare & Medicaid Services reimbursement. Using data from the Vizient Clinical Data Base, Dr. Cummings found that 63 members were using the technology in Q4 2020 when reimbursement first started. Within six months, the number has doubled.
AI’s impact on health care delivery is vast and already far-reaching. In the detection of diabetic retinopathy—a major cause of preventable blindness in the U.S.—AI technology can improve health equity and patient access to care. Currently, not everyone gets tested, yet if detected early by screening, the chance to do lifestyle modifications and treatments can prevent or prolong the onset of blindness.
Imagine a scenario where a diabetic patient can go to their primary care physician and instead of worrying about access to a retina specialist to do further testing, the physician in the primary care office has the technology to take a picture of the patient’s retina. The AI then reads the scan and does a diagnosis automatically and within a minute tells the primary care physician the patient has more than mild diabetic retinopathy that needs to be followed to prevent blindness. All done in real-time in the same office.
“This technology is available right now for the primary care setting,” Dr. Cummings said. It’s a clear example of not only preventing a serious health issue but improving access as well as health equity across the U.S. That’s exciting.”
AI adoption reality check
Hospitals considering early adoption of AI look at working through implementation barriers like cost, reimbursement mechanisms, proof of patient safety and efficacy, privacy and cybersecurity as well as the benefits of workflow efficiency and quality improvements. Being better positioned to address these barriers for AI as it matures puts organizations in front especially with health care consumers whose preference is always the “least invasive” and “most convenient” approach to care.
AI is clinical technology and that comes with associated risks. Many are strategizing how to best cut thru the hype, integrate clinical decision-making with supply chain, medical staff, and IT and determine what is beneficial to the organization and patient care. Once the technology is available, appropriate health care use is driven by the clinical value equation—what outcome does it improve compared to the cost?
“Members should be thinking outcomes,” Cummings said. “I always say, look long term. In a value-based model, you’re not buying an AI medical device, you’re buying outcomes.”