Autonomous medical coding company CodaMetrix recently announced the formation of a Coding Quality Council, a governance body made up of subject-matter experts from five large health systems. To better understand the council’s goals, we spoke with Jamie Noorlander, CodaMetrix’s director of coding applied research, and Monica Watson, corporate coding director at Allegheny Health Network and adjunct instructor at University of Cincinnati.

The council includes coding leaders from CU Medicine, Mayo Clinic, Henry Ford Health System, Allegheny Health Network, and Oregon Health & Science University.

As the corporate coding director for Pennsylvania-based Allegheny Health Network, Watson oversees coding operations for the entire network, which includes inpatient, outpatient, emergency room, and the professional organization.

As the director of coding applied research at CodaMetrix, Noorlander is responsible for the company’s coding quality program. She works  with the company’s machine learning team to help monitor AI model quality and improvements. She previously worked in coding at academic medical centers. CodaMetrix says it partners with over 30 leading health systems across 27 states – collectively representing more than $191 billion in net patient revenue, 30 million patients, and 120,000 physicians

Healthcare Innovation: Could you talk about why you thought it was important to create this council? What are some issues with coding quality or trust that the company identified and that resonated with these five health systems?

Noorlander: From my perspective, starting with CodaMetrix five years ago, I was working a lot with the early customers, and just personally realizing that all of us across the medical coding industry agree that 95% is our quality standard, but how we were defining it and measuring it was really different across all of the health systems we were working with. Even where I came from, we had different ways of auditing, and the focus was different depending on whether providers were coding or whether it was human coders. A lot of times the scope of audits and the cadence was different, too, depending on the resources a health system has available. 

As an AI vendor, we wanted to measure our quality as well, and we wanted to hit that 95% standard, but we were finding that the way we were measuring it might not match the way our customers were measuring it, so we were seeing a gap that there is no set standard for measuring quality that works across all health systems, no matter how the codes are being provided, whether AI or provider medical coding professionals. We saw there was a need. We and other AI vendors are trying to determine how we are going to report out quality that is standard across all health systems, and even AI vendors and payers. 

Watson: I've been a coder for 25 years, and the same thing has always been true. You can put five coders in a room and more than likely we'll have five different answers, but we want to have accurate outcomes. Our passion is to have accurate coding, because we tell the patient's story. We, don't get to give hands-on patient care, but the coding we provide is the way we give our care by telling their story accurately. To do that, we need to have standard sets of definitions. To Jamie's point, if we don't know what we are calculating, if we don't know what the expected outcomes are, if we don't know what the metrics look like, we don't know how we're being measured. It's very difficult to ensure that a technology knows how to achieve it, because we have to tell that technology what to do.

HCI: If this isn’t addressed, could it lead to disputes between payers and providers, or compliance gaps in meeting requirements?

Watson: We have that now. We deal with that more than anything today. I think if nothing else, having AI tools gives us a standard set of opportunities to level the playing field a little bit. Having these standards in place, we can say these health systems and these AI tools are coming together to define something we all know as an industry standard and that we all feel comfortable with. We need to work together, because we're not competitors. We’re ultimately doing this for the better of our patients. So my hope is that it would improve our relations, and instead of continuing to be adversarial the way they are today.

Noorlander: If we can get broad adoption of a quality framework that can expand to payers, then we're all working from the same quality measurement standard and understanding. I think that we really could help reduce a lot of that friction between the coding and the payer worlds. As Monica said, it’s definitely an issue now, but as AI is being adopted on the healthcare side, a lot of the insurance payers are also using AI.

HCI: Jamie, you mentioned the goal of having broad adoption of this quality framework, and you're starting with five big healthcare organizations, but are they all CodaMetrix customers? How can you get other organizations involved? 

Noorlander: Actually, Monica’s organization isn't currently a CodaMetrix customer. We really value having some outside voices and hope to expand to other healthcare systems. I think our network of medical coding professionals is a pretty tight group, so we are looking to each other to identify other medical coding professionals who are forward-thinking and who know that AI is the future and want to be part of how we're going to define quality. Obviously this is not just for CodaMetrix, but we want this to be something that can be adopted across all the health systems, and eventually payers as well.

Watson: We’re at a crossroads in the industry where we need to upscale our coding professionals. Our health information students who are coming into this industry need to see that there are pathways they can follow, so we want to leverage national organizations and we want to leverage each other's expertise to push the path forward and give these frameworks an opportunity to grow and expand.

HCI: Monica, have you seen just in the past couple of years at Allegheny Health Network an influx of AI into the coding practices and have there been some issues you've had to work through to implement it?

Watson: Absolutely. Even before coming to Allegheny. I’m also a professor at University of Cincinnati. We're building it into a lot of our university curriculum. There is a large retiring workforce, even before AI started to gain traction, so that coupled with that technological factor is a big struggle. So we start bringing these tools in, and those coding professionals become very uncomfortable and very concerned about learning new technology. Then you have these new students and they aren't really sure about the future longevity of their position. They don't understand that this isn't going to take their job, it's actually a great tool in their toolbox. So, we have had some of that change, and I'd say it probably started 10 years ago, or so, with the computer-assisted coding, natural language processing.

HCI: But are you starting to see some efficiency gains so that people are able to do more with AI assistance?

Watson: Absolutely, and it isn't just doing more. We have a lot of muscle memory as coders. We’re trained to look for specific things and we get into a place of complacency. Having these tools available gives us an opportunity to get out of that comfort zone and trigger a different part of that brain to have a little bit more of constructive and critical thinking that we lose over time when we fall into that complacency mentality. So we are definitely seeing efficiencies, seeing a lot of good questions, seeing a lot of ‘aha’ moments and noticing the accuracy that comes with it.

Noorlander: From the CodaMetrix perspective, I think we've seen that customers that have a base of medical coders and a strong medical management who is really excited about the AI adoption and sees the potential, they tend to have a lot more success because they're willing to work closely with us, and they're not just trying to find errors with AI, but they  want to help improve the workflows and how the AI models perform. Even just having that mind shift from a customer perspective has made the adoption and the models more successful on our side with those customers. 

HCI: I understand that this council is meeting quarterly to review the quality framework. Has that already starting happening?

Noorlander: We've had a couple of meetings so far, so we are just starting with the foundation of our quality framework. We really want to vet it. We're starting with our radiology service line — that’s our most mature service line — looking for some volunteer medical coders to participate in a white paper study so we can prove that this quality framework makes sense, and it's working, and also get feedback. We know this is fluid, and we will iterate and grow it over time, so we want to get feedback, and we want to improve the quality framework. We need medical coders and professionals willing to participate and give us feedback, so that we can do that.

I hope that we can see some success in getting adoption of this idea of a standard quality framework. I think it's important, too, that this isn't just about AI, it's about all coding platforms, whether it's a provider, human coder, or AI. Just having a standard way of measuring and benchmarking consistently is important.