Excessive clinical documentation is a widespread problem that is causing clinician burnout and even having an impact on patient care – as an American Medical Informatics Association survey recently showed. It revealed the frustration of physicians and nurses about the burden of electronic health record charting and the time and effort needed to complete necessary documentation.
There are technologies that can help. Many health systems are making use of natural language processing tools, generative artificial intelligence and ambient scribes to help their clinicians with EHR-documentation tasks.
But these tools are not plug-and-play. To work effectively, they need to be customized around the specific needs of their clinical end users.
Dr. Dean Dalili, chief medical officer of DeepScribe, says customization with ambient scribes is an essential feature, particularly for medical specialties, because the ability to personalize note-taking reduces clinicians’ time spent editing automated notes.
“There’s less time spent in a physical interface” to create formatting and other individual note-taking preferences, and more time for patients, he said.
In this Q&A, Dalili also discussed ambient intelligence and its benefits to enterprises. The feature compares note-taking to coding standards and creates a reporting structure for looking at clinical visit note quality across points of care.
Q. How does ambient clinical documentation improve care delivery experiences and reduce burnout?
A. To understand the impact of ambient AI documentation, consider the previous standards. Clinical documentation was paper-based and then moved to EHRs as part of [federal] policy.
That transition has been good for quality and safety, particularly around medication safety, and physicians get better insight into the patient’s context as they make a decision.
However, the problem is that EHRs have degraded care delivery because doctors are often more visually engaged with their computers than their patients.
EHRs contribute to burnout – which is really a phenomenon of high levels of mental load. That’s part of being a clinician, but made worse when you’re constantly mode-switching to different types of work and you’re translating an encounter to some sort of visual output in documentation format through a keyboard.
AI clinical documentation allows the provider to directly engage with the patient and have a normal conversation. That conversation becomes the source of information for which the software produces structured documentation that’s comprehensive. In some ways, more comprehensive than when relying on the provider’s memory.
The technology is always listening, and will sometimes catch details that a provider either forgets or isn’t focused on. Those details make it into clinical documentation, resulting in a better-quality experience for the patient and better note quality.
Q. Why are customizable AI scribe tools important for specialty care settings?
A. Customization is important to every doctor but most important to specialists. As doctors, we fall into rhythms.
A one-size-fits-all scribing solution that’s going to produce a note with only a standard structure is not going to serve most providers. There are nuances about how providers like to capture information – that may be elements in the subjective history portion of the note, parts of the physical exam, or in the assessment and plan portion, where the provider groups parts of the treatment plan with each clinical assessment.
There are also formatting elements of the note that each doctor likes. If you’re a geriatrician, you may want to refer to patients as Mr. or Miss; if you’re a pediatrician, you don’t want to refer to a young patient that way and may want a first name only.
Having customization options helps create a note that more likely mimics what the provider prefers, is used to or has historically written. That’s important because when the output matches the provider’s documentation preferences, it requires them to do less editing.
Think about the value proposition of any AI scribing: You don’t have to spend time first having a conversation and then spend more time documenting the conversation. But if you still have to edit that note to make it look the way you want, you still have a lot of extra work. That’s why customization is important for specialties.
Specialty-specific workflows get captured differently than a normal primary care visit. As such, there are different areas of emphasis and detail providers want to capture. It’s important to set up a format not in a generic, one-size-fits-all form that applies to some, but rather to be able to customize the output and structure so the AI is listening for components of the visit that are unique to that specialty.
In oncology, for instance, there’s usually a very long summary of data that will define a patient’s diagnosis and all the data elements that went into identifying their problem. Additionally, specific elements of the plan may be unique to cancer treatments – related not only to medical therapy, but also to social support, nutrition and other things. The note for orthopedics could look completely different and focus on a physical musculoskeletal examination, imaging and so on.
Q. How is the technology different from other AI scribes available in the marketplace?
A. First, we use a unique large language model that includes historical data from clinical visits coded by live scribes, which helps create structured data elements. We train our LLM – unlike an LLM like ChatGPT4, which is trained on the internet.
If you use medical information to fine-tune your LLM, you’re more likely to get accurate medically-related output. If you train the LLM on the entirety of the internet, you get extra noise that can reverberate into the content.
DeepScribe has the largest source of training data based on user notes used to produce highly accurate documentation, which helps with trust and adoption and minimizes the time providers spend on rework or edits.
The second differentiator is the tool offers more than 50 different customization elements, enabling providers to produce work that’s closer to what they would create from scratch, across a broad degree of specialties and users.
The third key differentiator is a new category called ambient intelligence, which is functionality beyond scribing.
This is where the patient conversation can be applied to any kind of structured data, whether it be a coding standard or a clinical-quality standard. From there, the AI can determine whether the conversation met that coding standard or not.
This intelligence also allows us to create a reporting structure where, across a broad spectrum of providers, in an instant, an enterprise can see how physicians perform. It’s the ability to, at the point of care, help determine high-value clinical content and also assess whether or not that content was delivered.
Q. Does DeepScribe integrate into a physician’s existing workflows and tech stack?
A. With ambient AI, the workflow is fundamentally different and does require some adaptation – the physician has to vocalize findings in a way that they may not have done in the past.
Providers can’t simply say, “This doesn’t look good.” Instead they have to say ‘Your left knee looks swollen’” so there is a degree of specificity in the language that providers have to adapt to. This level of detail makes the AI listening more robust.
The degree of integration is EHR dependent and DeepScribe has integrations with Epic, athenahealth, eClinicalWorks and more than one hundred other EHRs, but if a physician wants to connect a proprietary EHR, DeepScribe can also integrate through an API.
Andrea Fox is senior editor of Healthcare IT News.
Email: afox@himss.org
Healthcare IT News is a HIMSS Media publication.