Establishing Governance Capabilities Early in Your AI Journey

Crucial AI governance capabilities include understanding how AI tools are being used and monitoring of AI tools to reduce or eliminate unintended consequences.

Health systems should put a governance structure in place early in their AI adoption process, a top executive at Community Health Network says.

"You need to have the governance in place to make sure that you understand all of the tools that are being used, how the tools are being used, the intended outcome of usage, and how you mitigate bias," says Patrick McGill, MD, MBA, executive vice president and chief transformation officer of the Indianapolis-based health system. "Having a governance structure in place from the beginning is helpful."

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.

CHN recently hired a director of AI and data governance, who will be standing up an executive steering committee to help identify and prioritize AI tools and use cases.

The members of the executive steering committee will include McGill along with the health system's CFO, CMO, chief physician executive, chief information officer, and technical staff from IT and analytics.

"The focus of the executive steering committee will include prioritization of projects; implementation and use of policies, procedures, and traditional governing functions that must be put in place; resource allocation; and oversight of the monitoring of AI tools to reduce or eliminate bias and other unintended consequences of AI," McGill says.

Monitoring is a crucial AI governance function and varies based on various AI tools, McGill explains.

"Some of the monitoring will be through data reports," he says. "Some of the monitoring will be third-party partnerships to monitor the performance of an AI model. A third approach to monitoring will be to understand how third-party partners are using our data—we want to know whether data is de-identified, whether data is going out to a publicly available large language model, and what data the AI model is built on."

Prior to establishing the executive steering committee, the health system began the process of creating AI governance capabilities, according to McGill.

"We have put in policies, procedures, and guardrails for the appropriate use of AI," he says.

Patrick McGill, MD, MBA, is executive vice president and chief transformation officer at Community Health Network. Photo courtesy of Community Health Network.

Community Health Network's AI models

Like most health systems that have adopted AI tools, CHN is using ambient listening and ambient documentation to decrease administrative burden on clinicians.

"We partnered with DAX Copilot for the past year and a half, so we are using that technology across almost all of our outpatient specialties," McGill says. "It is helping people document faster and get out of the clinic faster. They can close their charts efficiently."

The health system recently adopted Secure GPT via a partnership with Qualified Health, which is HIPAA compliant. Clinicians can use the AI model to ask clinical questions and put in clinical information if they need to generate a letter.

"Clinicians are using it for note summarization and rapid look-up of clinical information such as the maximum dose of a certain medication," McGill says. "It is helping clinicians to get quick answers. Clinicians are using it for clinical workflows such as generating a consult letter or generating a summary—any information they might need to make the clinical workflow be more efficient."

One of the newer AI tools that CHN is adopting in a partnership with Notable, McGill says, will help streamline clinical workflows.

"For example, it will read a mammogram, identify whether the patient has an abnormality and needs additional imaging, and initiate a referral for the additional imaging," he says. "If the patient needs additional follow-up such as a breast MRI in six months, the AI tool will identify that need, then generate a referral for that follow-up."

Radiology is the focus for this AI tool now, but there are potentially other use cases with lab reports and procedure reports, McGill says.

Another new AI tool that the health system is piloting can conduct chart summarizations specific to a clinician—what the clinician's preferences are and what the clinician needs to know or wants to know before walking into a room to see a patient.

"This AI tool will automate a lot of manual work, and it is reading the chart to pull out nuggets of information that previously you had to have a human do," McGill says.

AI impact and advice

The health system's clinical care teams are giving positive feedback on the adoption of AI tools, according to McGill.

"AI is taking administrative work off their plate," he says. "It is removing administrative burden and allowing clinicians to focus more on taking care of patients. Often people are resistant to using new technology, but the feedback that we are getting on our AI tools has been welcoming."

In addition to establishing governance capabilities early in a health system's AI journey, McGill offers two pieces of advice to others who are adopting AI tools.

"One, you must include the clinicians in AI adoption from the beginning," he says. "Two, you must be willing to take chances. These tools are new, and they are not perfect, but they are only going to continue to get better. If you are only going to move forward with perfection, the train is going to go past you."

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 HealthLeaders Exchange event, email us at exchange@healthleadersmedia.com.