6 Tips for Clinical AI Adoption in Healthcare

Healthcare leaders should think about business value first in the face of AI’s allure, says Sutter Health’s Kiran Mysore.

Sutter Health’s AI leader says clinicians may be too optimistic about what the technology can do, and they need to understand that there are right and wrong ways to use AI.

“AI is very complex,” says Kiran Mysore, MS, chief data and analytics officer at the northern California health system and a participant in the HealthLeaders Mastermind program on AI in clinical care. “It is rarely a turn-key solution, where you adopt a model and expect it to work.”

“It needs a lot of good, clean data. It needs a lot of talented and skilled professionals to make it work the right way,” he says. “It needs the right workflow integration, and it should impact the point of care. And it needs to be trusted and dependable, which means you must tune the models well so they can predict the right answers.”

HealthLeaders is conducting its AI in Clinical Care Mastermind program through December. The program brings together nearly a dozen healthcare executives to discuss their AI strategies and offerings. As part of the program, each of the panelists are talking with HealthLeaders about the use of AI in clinical care.

Understanding the complexity of AI is one of six pieces of advice that Mysore offers for health systems embracing the technology in clinical care. The other five are:

• You need to lead with the business problem or the clinical care problem you are trying to solve with AI first, before thinking about the technology. In some cases, the answer to the business problem may not even be AI.
   
• In cases where AI is a solution to a problem, be very specific about the outcome you want to drive with AI. Focus on integrating AI into clinical workflows, measuring the outcomes over time, and understanding the improvements you are making against a baseline.
   
• You should try to think about scale on Day 1. Don't wait until an AI pilot is done to think about the next step, because scaling takes a long time. If you don't think about scale and performance on Day 1, you lose momentum with your stakeholders.
   
• Use best practices across the board and from the start. Talk with other healthcare organizations that have adopted AI models to learn from them, so you can capitalize on opportunities and avoid making mistakes.
   
• The biggest pitfall is being too optimistic about AI. We are in the early days of AI initiatives. It is rarely going to work exactly as advertised because every use case in every health system is unique. You must think about challenging each promised AI capability. The pitfall is thinking that AI is a silver bullet, and that it will work for everyone.

Kiran Mysore, MS, is chief data and analytics officer at Sutter Health. Photo courtesy of Sutter Health.

How AI is impacting Sutter Health clinical care teams

Sutter Health has launched AI capabilities in ambient listening to summarize the conversations between clinicians and patients, suggesting responses to patient messages to clinicians’ electronic in-boxes, and helping radiologists work better with images.

"The opportunity is to educate clinical care teams about AI, the right use of AI, and the responsible use of AI," Mysore says. "We need to integrate AI into day-to-day workflows, whether it is a physician, a nurse, or hospital staff. When you do that well, you can see tremendous benefits."

"Ambient listening is saving our physicians valuable time every day in note taking and documentation," he says. "AI is helping with the cognitive burden on clinicians because they are feeling less stress at the end of the day because AI is helping them do things easily."

AI is also impacting clinical care teams in terms of augmentation, Mysore explains.

"I do not see clinical care teams going away because AI is taking over," he says. "What AI does is to make clinical care teams' jobs easier and it frees up time in their day to elevate patients’ experience. AI can help clinical care teams see more patients."

AI governance

Healthcare organizations often do not think about governance as much as they should in the early days of AI adoption, according to Mysore.

"We need to be transparent about what AI can do and cannot do," he says. "For example, if you are using an AI model that is trained on some patient data, we need to be transparent about what patient population was used to train that model. If the model was trained on a patient population that has different characteristics from where it is being used, then its outputs will not be reliable."

Sutter Health is evolving a comprehensive approach to AI governance, with a core AI governance committee comprised of legal and digital team members.

"On the legal side, the committee is focused on whether we are doing the right thing in terms of privacy, security, patient confidentiality, risk, and compliance," Mysore says. "The digital team is focused on whether we are using the right AI models, whether we are building the right technology infrastructure, and how these models are working."

The health system has participation from other groups in the AI governance mix, including a clinical group and a research group.

"We have a core group and sub-groups that are aligned to promote AI governance," Mysore says. "Depending on the need, these groups can engage each other on a frequent basis to get advice, get validation, and check on monitoring of AI models. The goal is to accelerate building the right AI capabilities."

The HealthLeaders Mastermind program is an exclusive series of calls and events with healthcare executives. This Mastermind series features ideas, solutions, and insights on excelling in your AI programs.

To inquire about participating in an upcoming Mastermind series or attending a HealtLeaders Exchange event, email us at exchange@healthleadersmedia.com.