Nursing-Focused AI Ambient Scribes with Mark Townsend of Bon Secours Mercy Health

This is part of our weekly executive insights series where Elion CEO Bobby Guelich speaks with healthcare leaders about their tech priorities and learnings. For more, become a member and sign up for our email here.

Role: Chief Clinical Innovation Officer

Organization: Accrete, Bon Secours Mercy Health

Where are you focused as you look at the rest of 2024 and into 2025?

In my role, the problem to solve was engaging our markets and really understanding what they needed help with. From a listening tour we are leading within Bon Secours Mercy Health, we developed 12 innovation priorities, and at least seven of those have turned into active projects. Many of those lend themselves to AI, and all of that then informs the investing pipeline of Accrete, our digital holding company.

What are some of the key priority areas you’re focused on coming out of the listening tour?

• Optimizing patient care delivery

• Patient access

• Patient-provider communication (everything from customizing treatment plans to enhancing patient compliance)

• Patient engagement

You mentioned that for some of these priorities, AI-driven solutions are at least a consideration. Where do you see AI fitting into some of the things you just laid out.

Our use of generative AI has created an interesting conversation internally. We’ve done 60-some introductions of individual technologies into the organization so far this year, which helps us engage in new ways of doing things. While it takes time for us to formalize a relationship, the conversations mobilize our teams to meet our challenges and to create new partnerships.

We’re working hard to be diligent about first considering Epic solutions if we’re looking at a clinical use case. Epic isn’t necessarily synonymous with speed, so we’ve also had to consider where we have the luxury of waiting for Epic to develop these technologies, or where we are going to do it ourselves to meet the expectations of our operators.

One area where can’t rely on Epic is in the category of optimizing care delivery by focusing on ambient AI dictation for nurses. Arguably, this is a white space. There’s not a ready-to-go solution that has this all figured out. I think it’s great that Epic is working on this with the Mayo Clinic and Abridge, and we aren’t sure that we can wait for that relationship to transform this space. Through the leadership of Brian Weirich, our CNO of Innovation and Transformation, we are going to move ahead with a partnership that allows us to help grow this technology into what it needs to be in the near-term, with an aggressive goal of beginning implementation next year.

As you’ve investigated this space, can you share some of the differences in workflows between what you need in a nursing solution versus physicians or APPs?

These solutions largely began with ambulatory patient interactions with physicians or APP’s which follow a very specific format that the clinician can articulate, and which lends itself nicely to generative AI.

For nurses however, you have to first pick a specific use case—we’ve landed on admission or discharge notes—and then you pick a specific unit that needs the most help, and I believe we’ve selected med-surg. If you go to a med-surg nurse doing admissions, they know all the questions they need to ask the patient, but right now documentation is based on discrete documentation.  Said differently, it is based on clicks; click, drop-down, click, click, click. It’s like “death by a million clicks.”

What we envision would be to capture the required discrete data elements during a clinical interaction while a nurse is both talking to a patient, and concurrently taking vitals, for example;  nurses typically know all the questions that need to be answered. Unfortunately, their current workflow often requires them to ask their questions while working with a patient, and then leave the patient room to document afterward, or have a colleague sitting next to them that’s actually documenting in the EHR in real-time. That lends itself to a very different technology than ambulatory dictation because nurses typically document each answer to each specific question into form-based clinical documents. There are solutions being developed that are now moving into this space, but they’re still in their infancy.

Can you tell me a little more about how you’re measuring ROI across all of these projects? On the scribe example specifically, how did you think about ROI there?

That’s a fun one. We picked med-surg because it’s got our highest turnover within the organization, and it traditionally has our highest agency utilization. We aspirationally believe that by investing in the workflows of our teams, we can improve retention and decrease nursing turnover.  We eventually would then decrease our reliance on agency staffing, thereby reducing the spend specific to staffing our acute-care units.

To build on that tall order as we go down the ROI path, if you enhance the clinical workflows of any of our workforces, you see improvements in quality of care. A burned out clinician delivers lower quality care.  If we get this right, then we will see downstream improvements in the quality of care delivered.