As Interest in Ambient AI Grows, HI Professionals Focus on Clinical Documentation

Picture this: at your next provider visit, you notice that your clinician has a voice recognition program on their tablet, and it’s recording your conversation. A few minutes after leaving, you receive an email from the patient portal, notifying you that notes from your visit are ready—and there’s a set of clinical recommendations already prepared.

If that happens, your provider is probably relying on something called ambient listening, which is a form of artificial intelligence technology. Ambient listening AI leverages voice recognition technology to listen to provider-patient conversations in real time, interpret the dialogue, and create medically relevant clinical documentation.

Ambient listening is gaining attention as healthcare organizations start to adopt technology empowered by artificial intelligence (AI). About 30 percent of physician practices already use ambient listening AI, according to the Medical Group Management Association.

While some vendors tout ambient listening as a revolutionary innovation, healthcare leaders, as with any new technology, still have to ask key questions: Will this technology deliver on its promise? And, while the technology is being marketed to physicians and other clinicians, what will the adoption of the ambient listening AI mean for other healthcare constituents such as health information (HI) professionals and coders?

Behind the Buzz

By automatically capturing and transcribing patient–physician interactions, ambient listening technology can potentially alleviate the burden of manual documentation, a problem that is more than just a thorn in the side for clinicians in this age of worker burnout.

For the past several years, physicians have experienced staggering rates of burnout. For example, about half of physicians reported burnout in 2023, according to statistics from the American Medical Association. And much of the stress can be traced to documentation, as a study from the Advisory Board showed that physicians spend 49 percent of their time working in the electronic health record and with other desk work and only 27 percent of their time in direct clinical care.

Meanwhile, staffing shortages among HI professionals also have led to dissatisfaction. According to a survey by NORC and AHIMA issued in October 2023, 76 percent of the more than 2,500 HI respondents reported experiencing burnout. Burnout appears to impact most organizations that employ HI professionals, but was reported at higher rates among groups experiencing HI understaffing, the survey found.

In addition to alleviating burden, ambient listening AI can help improve the physician-patient relationship. With technology handling the note-taking duties, physicians can better focus on patients. At the same time, more sophisticated note taking can translate to less questions about notes for HI professionals.

Early evaluations of ambient listening AI are promising. For example, more than 3,400 physicians used ambient AI during 303,000 patient encounters and they cited the technology’s “capability to facilitate more personal, meaningful and effective patient interactions and reduce the burden of after-hours clerical work,” according to a study published in NEJM Catalyst on February 21, 2024. In addition, patients noted improved interaction with physicians not spending time looking at computer screens.

“The biggest advantage is actually to our patients. They value the face-to-face time they get with their physicians or clinicians,” says Kristine Lee, MD, associate executive director at Permanente Medicine in Oakland, CA, and one of the NEJM Catalyst study’s authors. “For our clinicians using the technology, we’re seeing a decrease in administrative burden of typing up or dictating a note. Instead, they can invest more time in the patient relationship and helping them to stay healthy or take care of their medical problems. The biggest time savings was in ‘pajama time,’ or the time after work hours when clinicians are finishing their documentation at home.”

The clinical documentation process has long placed a burden on providers, health leaders say. Prior to using AI for clinical documentation, physicians had to continually learn about new documentation requirements and to refine their note taking skills. As a result, clinical documentation improvement specialists would train – and retrain – physicians on how to write clinical notes, assessments, and plans, based on continually evolving regulatory criteria.

“Physicians had to learn how to really write the most complete note that reflects the depth of their thinking because that can make a huge difference on all things related to reimbursement,” says Shiv Rao, CEO at Abrige, an AI company based in Pittsburgh, PA.

Better Notes?

The hope, of course, is that ambient listening technology will lead to improved clinical documentation, which will have positive effects on the overall revenue cycle.

“Since AI-assisted documentation should generate fuller notes, it should also facilitate higher-level and/or more accurate coding,” says Steven Lin, MD, director of the Stanford (CA) Healthcare AI Applied Research Team. “The potential advantage of this in fee-for-service settings is higher reimbursement for office visits. In capitated settings, the advantage could be higher risk-adjustment for value-based payments.”

However, the quality of clinical documentation with ambient AI as compared to other methods has not yet been definitely proven, according to Lin.

The quality of clinical documentation is important as it affects patient care, patient safety, and medical errors. Clinical documentation also can impact quality measurement, finance, and research, according to a study published in 2022 in the Journal of Medical Systems.

The fact that notes are being produced in a new way means that additional scrutiny is required. However, ambient listening or “digital scribes” could be limited by increased error rates, according to the research review “Improving Clinical Documentation with Artificial Intelligence: A Systematic Review,” published in the Summer/Fall 2004 issue of Perspectives in Health Information Management.  

AI Homework for HI Professionals

HI professionals must ensure that AI-produced documentation provides what’s needed to optimally treat patients and support revenue cycle processes.

Selecting and then implementing an ambient listening solution that has been developed with clinical documentation concerns in mind can help providers experience greater success. To start, healthcare organization leaders should select an ambient AI listening solution that serves the needs of various audiences including other clinicians, patients, and revenue cycle staff members such as HI professionals and coders, Rao says.   

Perhaps most important, AI listening solutions need to create trustworthy output. HI professionals and coders can more readily trust clinical notes when AI automatically points them to the discreet data that supports the documentation. Right now, part of a coder’s role often requires querying clinicians to substantiate notes in the clinical documentation.

“For example, I might see a patient and then six weeks later, somebody from the coding department is pinging me and paper cutting me queries related to the note,” says Rao, who is a cardiologist. “Did you document something about the condition? Did you really talk about it or discuss it and put enough detail into it? So all those paper cuts, all those queries, clinicians end up being a casualty of war. It is also a lot of work overhead for those revenue cycle departments.”

With ambient listening AI, it’s possible that coders could now “hit play and hear just that snippet of the conversation, the actual information that informed the model's output,” Rao says.

In this way, ambient listening that creates robust clinical documentation can reduce this type of friction by creating a comprehensive and accurate documentation and providing the tools to efficiently audit the note. Reducing “chart chasing” or clinical queries by allowing coders to more clearly see the evidence supporting documentation and thereby more confidently produce the code could have huge implications on the revenue cycle.

As such, it’s important to work with AI technologies that are trained to include all the criteria that can support a code in the clinical documentation, Rao says.

John McCormack is a Riverside, IL-based freelance writer covering healthcare information technology, policy, and clinical care issues.