Lessons learned and best practices for an effective, impactful clinical AI strategy

The healthcare sector has met artificial intelligence with a mixture of excitement and apprehension. In a world where clinicians are overwhelmed by data, many view predictive AI solutions as necessary tools that can be paired with human expertise and judgment. When done ‘right’, these tools can dramatically improve health outcomes, reduce costs, and increase productivity to alleviate critical labor challenges such as staffing shortages.

During the Becker's Hospital Review 9th Annual CEO + CFO Roundtable, an intimate roundtable discussion was held around exploring best practices for implementing an effective clinical AI strategy. Rishi Sikka, president of Sutter Health in Sacramento, Calif., moderated the session with four experts who shared their experiences and insights:

• Suchi Saria, PhD, CEO and Founder, Bayesian Health; John C. Malone endowed chair and Associate Professor of Computer Science, Statistics and Health Policy, and the Director of the Machine Learning and Healthcare Lab at Johns Hopkins University
• Dan Durand, MD, Chief Clinical Officer, LifeBridge Health in Baltimore
• Julie Yoo, General Partner, Andreessen Horowitz
• Michael Ries, MD, Director of adult critical care and the eICU, Advocate Aurora Health in Downers Grove, Ill., and Milwaukee

Four key takeaways:

1. The best AI tools acknowledge the complexity of healthcare data. In recent years, the field of machine learning and AI has grown dramatically. In healthcare, one of the most promising applications is early identification and treatment of serious conditions. The journey to this point, however, has been emotional. "In healthcare, people got excited about AI and then outsiders came in who didn't appreciate how complicated and messy healthcare data is. Now the field has matured, we have technical approaches that are developed with messy health data in mind, and we are seeing exciting results -- decision support models that are 10x more accurate than alerts you typically see in your EMR," Dr. Saria said. Read more about Bayesian’s clinical decision support model results.
2. When selecting an AI solution, organizations should focus on areas with a defined clinical need. A common failure is buying an AI technology that wasn't designed for the organization's most pressing use case. "A general solution may require a ton of work and maintenance in the long run to stay on top of the organization's evolving use cases," Ms. Yoo said. Many organizations partner with brand-name vendors with no healthcare expertise. "In the beginning, you get a beautiful press release. Five years later, you have no results to show and you're exhausted because you had to educate the vendor about medicine," Dr. Saria said. Read more on the essential checklist for predictive AI solutions.
3. Involving clinicians leads to win-win AI strategies. When creating clinical AI solutions, Advocate Aurora Health engages key stakeholders, including IT and clinicians. "AI isn't a fixed product that you're investing in. It must integrate with what you are doing clinically," Dr. Ries said. "Involving clinicians creates win-win situations." A leader from a general acute care hospital system in Tennessee agreed, noting, "We need transparency in our processes, so there is trust that we've involved clinical leaders in the formulation of our technology solutions."
4. "Invisible technology," not more checkboxes and buttons, is the goal. Enterprise technology evaluation processes often emphasize various features and functionality. But a result is that healthcare organizations end up with solutions that require 17 clicks to get the right answer. On the other hand, some of the best consumer technology is completely invisible to end users. Rather than micromanaging vendors and focusing on specific features, the panel recommended focusing requests for proposals on the usability and benefit of AI solutions. Read more on 7 common mistakes when evaluating healthcare predictive tools.

Looking ahead, predictive AI platforms will sit on top of EMR systems and provide validated, rigorous, high-quality, workflow-integrated solutions in multiple clinical areas. This clinical operating system will enable organizations to scale clinical excellence across sites.

"In five years, we'll look back and say, 'I can't believe we were using EHRs before because they were billing and frustration systems. Now we are actually using the data to change clinical actions and outcomes,'" Dr. Durand said. Lifebridge Health is one system working directly with Bayesian Health deploying its AI-based clinical decision support platform to help diagnose and treat pressure injury, sepsis, and patient deterioration.

The Bayesian Health platform is a demonstrated leader in the space based off of a decade researching, building, and validating state-of-the art AI/machine models to fuel its platform. In addition, Bayesian works with leading systems to customize workflow integration, deploy comprehensive performance reporting systems, and enable a lightweight high-quality setup process for minimal disruption.

To learn more about Bayesian Health, click here.