A question you often hear about artificial intelligence in healthcare is how long will it be before we see widespread adoption. In a recent panel discussion, John Halamka, MD, MS, president of the Mayo Clinic Platform, responded with an oft-quoted quote from author William Gibson: “The future is already here; it’s just not evenly distributed yet.
Halamka said that right now at the Mayo Clinic, 14 algorithms are run on every 12-lead ECG taken. “We print the results of all the predictive models on the ECG itself – typical ECG frequency, rhythm intervals. We can actually tell you your ejection fraction. Do you have pulmonary hypertension? Do you have cardiomyopathy hypertrophic? Will you have A-fib in three years? It’s on the ECG itself, so there’s no cognitive load. There’s no button to press, no app to download. or distraction outside of the EHR.”
Halamka said that “over the next six quarters, we’re going to see more and more of these kinds of things being integrated into the EHR workflow itself, so this human augmentation, which is exactly the right term, will just be there.”
The Sept. 26 meeting hosted by Permanente Medicine also included Ed Lee, MD, executive vice president of information technology and chief information officer at the Permanente Federation. He is also an Associate Executive Director at Permanente Medical Group in Northern California.
Lee said he likes to refer to AI as augmented intelligence rather than artificial intelligence because he views the technology as a set of tools that help and augment a doctor’s ability to care for their patients. “It’s kind of like other ways we support physicians with clinical decision support tools; AI happens to be more advanced and more complex than other types of decision support,” he said.
Lee described a number of ways they use AI at Kaiser Permanente, including use cases related to natural language processing, computer vision, and predictive analytics. “With natural language processing, we analyze our patients’ emails and categorize the emails according to the topics the patient is writing about, which allows us to ensure that the most appropriate member of the team of care is addressing the message and every member of the team is practicing at the top of their reach. And of course, that helps our patients get quick answers to their health issues,” he said.
“We are also interested in computer vision, analyzing diabetic retinal images. We know that diabetes is one of the leading causes of blindness, and given the number of diabetic patients we have, using a tool that helps us determine whether or not a patient has diabetic retinopathy can identify retinopathy earlier, giving us the best chance of preventing someone from going blind.
Finally, in the area of analytics, Kaiser Permanente has developed a number of algorithms to help it stratify COVID-positive patients so they can better anticipate which patients are most at risk of developing more severe symptoms. . “We also have our advanced alert monitoring program, which helps us keep tabs on our hospitalized patients in real time, and predicts which patients are at risk of deteriorating and may need to be transferred to intensive care,” Lee explained. . “It gives us the ability to intervene before our patients get worse. And in the case of our early warning monitor program, we’ve estimated that we save hundreds of lives a year, and that’s actually a pretty conservative estimate. With all of these examples, AI is augmenting the care of our physicians and teams, and when combined with clinical judgment, we are creating the potential for significantly improved outcomes for our patients as well as efficiencies. for our clinicians and our healthcare system as a whole. .”
Halamka noted that all of this work starts with well-organized data. This involves data from EHR, imaging, telemetry and patient-reported outcome data, organized longitudinally and then made available to investigators using what he calls an AI factory. “That’s what Mayo did; it is very hard work. We actually took a few years to clean the data,” he said. “We then organize the data so that it is not episodic, but longitudinal. And we’ve anonymized it so there aren’t a lot of IRB sensitive human subjects or privacy issues with the use of the data, but it’s still stored in a secure cloud container with tools in more for all of our clinicians and all of our Mayo Clinic investigators to create models.
He emphasized that creating a model is not enough; one needs to test the model, validate the model, and use data even outside the training set. “We have also established a variety of national and international collaborations to test the models to ensure they are fair, unbiased and useful.”
Lee noted that AI’s work on risk prediction is important because it can affect “not just individual patients with the outcome of what we predict; it can, in fact, affect entire populations and entire communities. This is the power of what we have discussed here, where we can positively contribute to the health of very many patients. With the limited resources we have in the health sector, we need to focus our efforts on how to get the best value for money. We know that hospital readmission is costly for the healthcare system and costly for the patient. Sepsis is a case where a patient can become seriously ill and eventually die. How can we predict this type of outcome and prevent it from happening? »
Halamka was asked about the challenges of data interoperability and how they relate to advances in AI. One of the challenges to overcome, Halamka said, is not really related to technology. It is the ability to bring together multiple organizations to work together for the benefit of society. He described a few different promising models that are developing. One is an open source product from Verily called Terra. “He says Kaiser can put his data in this secure container; Mayo can put it in this secure container. Mayo and Kaiser can’t see each other’s data, but we can develop algorithms for both. And there are four or five other technologies that enable this kind of collaboration without necessarily requiring centralization or giving up control,” he said. “So I’m very optimistic about where we’re headed. We have achieved interoperability where interoperability is needed for structured data and for unstructured elements we have found secure computing technologies that promote discovery without compromising privacy or reputation.
Lee was asked about a KP group focused on quality assurance and algorithms. “Whenever we release these tools, we need to make sure they still do what we intended. to do,” he said. “Maybe they will work for the subset of the population that the algorithm was developed on, or maybe it works at first, but as more data is collected and more information are available, the algorithm should really evolve and continue to develop. We make sure to provide the fairest care possible and to ensure that the algorithms always do what they are supposed to do. It really involves validation, revalidation , then revalidation again. If you don’t look, you’ll never find a bias. You need to make sure this is built into the process in which these algorithms are used, developed, and maintained. As we continue to develop many of these tools here, that’s how we look at it. We’re making sure it doesn’t stop once things are in place. This is just the beginning.”
#Kaiser #Permanente #Mayo #Clinic #executives #detail #approaches