UNC Health's CIO talks generative AI work with Epic and Microsoft

Photo: UNC Health

Brent Lamm is CIO at UNC Health, based in Morrisville, North Carolina. He has quite the seat for the artificial intelligence explosion in healthcare – he’s sitting right in the middle of it.

Lamm and peers at Stanford, the University of Wisconsin, and the University of California at San Diego are part of a pilot project by electronic health records giant Epic, testing Epic’s large language model AI capabilities to enhance the provider experience.

Lamm also is in the midst of some exciting work with Microsoft in the AI space, leveraging generative AI (the kind behind groundbreaking applications like ChatGPT) on a broad range of use cases to assist clinicians and other team members.

Healthcare IT News sat down with Lamm to discuss these projects and his feelings on AI in healthcare as it grows faster than some might like.

Q. You are working with Epic, along with Stanford, the University of Wisconsin and UC San Diego, to pilot its large language model AI capabilities to enhance the provider experience. What is the goal of this project?

A. Our ultimate goals are to reduce clinician burnout and enhance the patient-provider relationship through the use of this new technology that assists with drafting responses to patient communications.

Over the past decade, the implementation and adoption of digital solutions that enable patients to communicate with their care teams more easily between visits have brought tremendous benefits in terms of experience and engagement.

However, a side effect has been an increased workload for clinicians, which has contributed to the current industrywide challenges for care team members in maintaining a healthy personal well-being. This situation only worsened with the pandemic, as the use of digital communication by patients has increased.

We believe the generative AI capabilities Epic is working to develop can play a huge role in helping address these challenges by reducing the cognitive burden of drafting responses. This is similar to how spreadsheet technologies have helped business users dramatically reduce the burden of performing basic calculations for decades, allowing them to spend their time and energy focused on the core essence of the given problem or opportunity.

By offloading the task of crafting a draft communication, the clinician will be able to better focus on the patient and their needs. The positive side effect of this more efficient process should also be an improvement in the patients’ experience and their overall relationship with their care team.

Being an early adopter of generative AI embedded into our Epic EHR allows our team to help drive this necessary transformation faster, while also building trust from our care teams that our providers helped make the end product more effective.

We hope this initial pilot project will be the first of many Epic AI pilots that our team supports. We are excited about the potential benefits and efficiency that AI can offer all our care team members and our administrative Epic users.

Q. More specifically, what will you and your team be doing with the LLM AI in-house to make the EHR experience better for providers? And how will you be getting your providers to understand the AI technology?

A. Our initial work will expose this new capability to approximately 10 of our physicians to safely test this new functionality in a highly controlled environment. We have assembled a team of physicians from various specialties, including primary care, to be part of this initial effort.

Their focus will be to evaluate the usability of the new functionality, particularly the accuracy of the communications drafted by the generative AI.

One key metric we will be examining is how effectively this new technology reduces the time it takes for providers to finalize their communications. The amount of editing required will serve as an important measure of success. Our ultimate goal is to assist Epic in enhancing the technology to the point where only minimal edits are necessary.

Another significant aspect of this work involves evaluating provider satisfaction regarding the tone and style of the draft messages. It is crucial that the generated content aligns with their expectations and preferred mode of communication with patients. We find this aspect of the work particularly promising from a technological and informatics standpoint.

Furthermore, our team is enthusiastic about evaluating and contributing to the development of generative AI with regard to specialty-specific differences and nuances. Our aim is to guarantee that the potential of this technology is accessible and effective for all of our clinicians.

Q. What will you be cautioning your providers about LLM AI? There’s a lot of concern surrounding AI.

A. We are very pleased to see Epic adopting a thoughtful and cautious approach in implementing this technology. One crucial aspect of our work is to properly train and educate the participating providers about this functionality.

The primary caution is to ensure a proper and thorough review of the drafted message prior to hitting send. We know the current state of generative AI is far from perfect and can generate text that is inaccurate or may contain bias, which is commonly referred to as a hallucination.

That issue will improve over time as the AI learns, but users must validate automatically generated drafts carefully. Our intention is to ensure that everyone clearly understands this technology is merely a tool to aid in drafting the initial message, with the providers retaining full control of the final response.

In parallel with this effort, we have launched a formal program with leaders from across our organization to develop and implement UNC Health’s Responsible AI Framework. A committee of multi-disciplinary leaders and clinicians will use this framework to evaluate vendor-developed and homegrown technology solutions that use various forms of AI.

The areas of consideration under the framework are fairness, transparency, accountability and trustworthiness.

Looking ahead, as AI technology continues to advance, we firmly believe maintaining the perspective of AI as “augmented intelligence” is vital and should remain a core principle. Machine learning, large language models and other forms of AI have the potential to revolutionize healthcare for the better, but they should be used as tools to support our greatest asset: our people.

Q. On another front, you’re working with Microsoft on an AI project. Please elaborate about this project and its goals.

A. Yes, we are excited to be working closely with Microsoft to leverage generative AI on a broader range of use cases to assist our clinicians and other team members. Much of this is still very early stage, but we already have a fully functional prototype of an internal general purpose chatbot, similar to the consumer-facing ChatGPT application.

We believe that a local, enterprise solution like this has the potential to unlock the productivity and efficiency gains that generative AI chatbots can offer, while maintaining safety, security and privacy for our users.

One of the first scenarios we are focusing on in leveraging this capability is to assist our care teams in accessing training and education resources more quickly and easily. Frontline care team members frequently reference these materials as protocols and workflows change, and we believe chatbot-style search can offer a far more efficient method for our users to quickly find the specific information they need to best navigate the technology and shift focus back to their patients.

Q. What would you say to your CIO peers at other healthcare provider organizations about working with AI today?

A. First and foremost, we must keep the patient and our care team members at the very heart of this work as it advances. The emerging technology has such tremendous potential to help improve health outcomes, reduce clinician burnout, and bend the cost curve across healthcare.

However, we must not lose sight of our mission to improve the health and wellness of the patients and communities that we serve.

While it may sound paradoxical, I would strongly encourage leaders across the industry to aggressively pursue use of this technology, but do so in a very careful and thoughtful manner. I also believe finding proven partners who share our values related to patient-centered care and protecting data and privacy is critically important.

Given the nature of this technology and the significant resources required to make it work, I expect most organizations will need to align with strategic partners. Ensuring alignment of values is just as important, if not more so, than alignment of technology and architecture.

I look forward to seeing how our peer colleagues and organizations embrace and drive improved outcomes with AI. If we do this right, the possibilities are truly amazing.

Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki
Email him: [email protected]
Healthcare IT News is a HIMSS Media publication.

Source: Read Full Article