Agentic AI meets specialty care: How an orthopedic practice is simplifying post-surgical care

Why the practice turned to AI to ease post-surgical monitoring challenges

With staff stretched thin and a patient population that wants answers about their health, Michigan Orthopedic Center found itself in the eye of a perfect storm.

Not only was it dealing with increased call volume to their medical triage line, but the staff answering the phones did not have the same level of training as a spinal surgeon, leading to back-and-forth conversations between staff, providers and patients to extract the right clinical information over the phone.

"So, that's when I started to look at AI and say, 'Well, hold on. We now have this ability to have an AI agent interact with a patient and really drive some of these important questions from the very beginning of the conversation,'" Dorenkamp said.

"And I felt like this was a really good initial use case for natural language processing, because we're not asking the AI to do anything that requires a decision. We're asking it to do a better job of gathering information. So instead of somebody calling and leaving a random phone call saying, 'Oh, I have neck pain, I need Dr. Dorenkamp to call me back,' the AI can say, 'Okay, well, one, did you have surgery? When was your surgery? Where's your neck pain? What level of neck pain is it? Is it trouble breathing, trouble swallowing?'"

Getting this information right away, rather than playing phone tag with patients, can help the practice increase efficiency and  potentially reduce the time each staff member spends on a call, Dorenkamp said.

"Hopefully, in the long term, it's going to decrease the burden on the staff and make their longevity in our clinics better," he added.

The tool, known as the clinical assessment agent, is part of Intelepeer's SmartAgent suite of AI tools built specifically for specialty healthcare practices.

"When you're thinking about somebody calling back because they have knee pain after surgery, you have to understand how to ask the right questions to identify what's there, being very cautious to not give medical advice or give any answers out, and really just ask the right questions," Mark Langanki, chief AI officer at Intelepeer, said in an interview.

For Michigan Orthopedic Center, partnering with a healthcare AI-focused company that understands the value of compliance and governance was crucial for success. Dorenkamp was adamant that the agent they built together would not give medical advice. Rather, it would enable better, faster decisions in post-surgical care.

Both Dorenkamp and Langanki stressed the importance of governance and guardrails when developing and implementing a healthcare AI tool.

"LLMs out of the box are very chatty. They love to give an answer. They don't say, 'I don't know,'" Lankganki said. "So, the agentic framework really is those guardrails, but then it's the context -- what are we talking about here? The more information we can give to the LLM in real-time, either through fine-tuned models or through very good prompting and everything in between, that gives it the ability to reason through what it should ask next when you say something."

Unique AI implementation considerations for specialty care

Dorenkamp stressed that a general-purpose AI tool would not necessarily fit seamlessly into his practice, prompting the team to pursue a more tailored approach.

"Orthopedics is very unique. We deal with a very specific pathology," he said. As such, an effective AI tool must be trained on orthopedic and spine-specific terminology to work effectively in Dorenkamp's practice.

"When I'm vetting vendors for my practice, I'm looking to see if they understand procedural specialties. Do they understand specifically orthopedic surgical prior authorization? Do they understand our CPT codes and our ICD-10 codes and, really, the world and the language we work in? Because that's different than what a GI works in versus what family practice or psychiatry works in," he noted.

Dorenkamp advised specialty practices to conduct proper due diligence when selecting an AI vendor and to focus on full-stack AI companies rather than going through a lengthy contracting process with multiple companies and ending up with tools that can't communicate with each other.

"I lean on partners that have established compliance records and live in this vertical because I feel assured that they're going to have my back if we have a compliance problem," he said.

In addition to ensuring that the tool is a good fit for a specialty practice, organizations must effectively communicate with staff to ensure that the implementation process goes smoothly and doesn't negatively impact morale.

"One of the things I've been very adamant about with our staff is that I have zero intention of replacing any of our staff with this tool," Dorenkamp noted. "Our goal in the next five years is to be staff-neutral. We don't want to have to grow our staff as reimbursement is shifting down, but we're not looking to reduce our staff because we truly need every human in the building here. The goal here is to give these people a tool that allows them to work at the top of their license."

Michigan Orthopedic Center staff have been directly involved in the training process, so they understand the tool fully and are ready to work alongside it. The practice's triage manager was involved in the development, and the staff has been engaged in the design process, Dorenkamp said. workflow design process and has been engaging the staff in

What does AI mean for the future of private specialty practices?

AI is undoubtedly changing healthcare, as providers and administrators across the country embrace the technology's ability to increase efficiency and drive a return on investment.

"I think in five years, we're going to see a completely different healthcare system. I'm probably a little bit of an optimist in this regard because I think that AI and technology is one of the few things that I've seen in the last five years that gives us a glimmer of hope as a private practice," Dorenkamp said.

"We've had so many negative financial things happen where it's been really hard to compete in the market, and you're competing against these behemoths of health systems that have a lot of resources, where we have to be a lot more cost-conscious."

Dorenkamp also said this is the first time he's seen technology adoption be driven by a "physician-up model," wherein providers are enthusiastic about how technology can make their lives easier.

What's more, private practices can be more agile in their adoption because they don't have to get several administrators to sign off on a new product -- they can conduct due diligence with their attorneys and rapidly adopt tools that they think will help their practice.

Dorenkamp emphasized that clinical AI is still a long way from being valuable, but administrative AI tools are already proving their value in healthcare's back offices, whether for appointment scheduling, coding and billing or prior authorizations. He expressed hope that this collaboration with Intelepeer would lead to larger and more effective AI use cases at Michigan Orthopedic Center in the future.

"This is the tip of the iceberg for this project for us. Right now, we're focusing on medical questions. This easily expands into postoperative phone calls, and I'm very hopeful that, downstream, it can decrease ER readmissions, because it's going to give patients easier access to our practice. We essentially have more touchpoints with the patients, and we can make them feel like we're caring for them as a whole person," Dorenkamp said.

"We want to know how patients are doing, how they're responding, and we want to be there for them if they need something. And leveraging tools like this, my hope and dream is that it allows us to do that in an efficient manner and actually increases our human visibility with our patients."

Jill Hughes has covered health tech news since 2021. Her coverage areas include cybersecurity, HIPAA compliance, interoperability, AI and EHRs.