Issue link: https://www.e-digitaleditions.com/i/1501973
18 | Summer 2023 AI and Wearables: A Game Changer in Healthcare The Center for Health Equity using Machine Learning and Artificial Intelligence is a "dry lab" where data scientists spend time working on computers and interacting with data. Much of this data is pulled from patients' electronic health records. But, increasingly— with patient consent— researchers are using data from smartphones and wearables such as electronic watches that can monitor pulse rates and perform simple EKGs. Karnik thinks that, in some measure, most College of Medicine alumni currently in practice are already using AI/ML tools. "Whether they realize it or not, there are algorithms running in the background and pushing information back into the charts and into various workflows in hospitals and clinical settings," Karnik says. "Given that our alumni share our concerns around equity, I think CHEMA's work will be of interest to them because it directly affects issues of equity. Without this research, patient care can suffer, creating further disparities. "Broadly, this technology is changing the practice of medicine," he says. "And the work at CHEMA promises to provide a central way to develop tools that will help ensure that healthcare is delivered in a fair and equitable fashion." "CHEMA is an exciting next step for the college, and we believe that the impact of this initiative will extend across research, clinical practice and teaching domains," adds Bellur S. Prabhakar, MSc, PhD, interim senior associate dean for research and associate dean for technology innovation and training. "We plan to be a leader in the development of new knowledge and the training of the next generation of physicians and translational researchers in the ways that AI/ML technology can be actively woven into practice for the promotion of health equity in underserved communities throughout the country and the world." know when to revisit and reprogram or recalibrate," says Karnik. For example, suppose an algorithm detects when a patient's insulin is not working correctly. The algorithm is deployed in UIC's clinic and is now identifying a lot of people whose insulin is not working, so their type of insulin is changed or a different medication regimen is ordered. As the patients get better, the number of cases in which the insulin is not working drops. The computer isn't going to predict as well because the baseline condition has changed. "In these instances, everything that the computer learned is starting to shift because the prevalence is changing," says Karnik. "So you have to recalibrate your algorithm or develop a whole new one to detect in the new situation when someone's insulin is not working correctly—because now you have changed the treatment or the approach, and the original object is no longer the same." Basically, that is what happened with COVID-19: The initial tools and algorithms that were developed when the pandemic was new and evolving quickly were not adjusted fast enough as more variants developed. Accordingly, the AI and ML should constantly be questioned and evaluated in real time and be agile enough to change the treatment approach. "These tools are not going to be 100% correct," he says. "The computer is not that type of instrument. So we must think in a very sophisticated way about how to use these tools to notice when the information doesn't match the disease or the situation." O P P O S I T E : Niranjan S. Karnik, MA, MD, PhD