How AI Solutions Alleviate Administrative Burden and Empower Clinical Staff – But Are We Ready?

8 minute read

Johnny Hecker, Consensus’s Executive VP of Operations, Shares His Thoughts on an Article About the Risks of AI in Healthcare – and Describes How the Proper Use of AI Technology Can Create Massive Benefits Both for Providers and Their Patients

If you’ve read the 2023 Los Angeles Times article titled AI may be on its way to your doctor’s office, but it’s not ready to see patients, you could be forgiven for being concerned about visiting a medical facility using this technology.

The article addresses one of the most talked-about topics in healthcare IT today: the use of Artificial Intelligence to streamline providers’ operations and improve care. But unfortunately, the author goes down only one single  path, focusing on a subset of AI, and broadly concludes that AI in general poses serious risks in healthcare.

A few of those risks discussed in the article include:

  • An AI program inaccurately summarizing the notes of patient visits.
  • A tool like OpenAI gleefully leaking confidential patient data when asked for it.
  • AI bots mysteriously “hallucinating” non-existent details about patients’ medical records and then using that made-up data to suggest diagnoses and treatment plans.

We agree. Scary stuff. But are those even the roles that the healthcare industry should be asking AI to play at the current maturity of the technology? If not, how can AI add value to providers’ operations today?

To answer those questions, Consensus Cloud Solutions editor sat down with me to talk about this confusion.

Extractive AI – Not the Much-Hyped Generative AI – Is Today’s Game Changer in Healthcare Tech

Interviewer – Jeff Solis:

The LA Times article takes a defensive and even nervous view of what could happen as artificial intelligence tools gain more traction in healthcare. Do you think that fear is warranted?

Johnny Hecker:

If things were to play out exactly the way the article envisions, I’d sure be worried. I wouldn’t want to explain my symptoms to a robot that might mischaracterize key details of our “conversation” in the summary notes – or worse, actually hallucinate details that didn’t come up at all and have nothing to do with me.

But we need to step back and define our terms because I think the LA Times article started off with an inaccurate premise about AI and then followed that premise to its logically worrisome conclusion.

Artificial intelligence can easily be perceived as  computer-based sentient beings and popular language and generation models like Chat GPT or DALL-E 2 encourage that perception.. But at the end of the day, the technology is really just a type of software, one that uses algorithms and large data sets to learn how to perform specific tasks with increasing competence.

The LA Times article is focused on the specific branch of the technology, called generative AI. That refers to artificial intelligence software that can generate new content based on the inputs it receives. 

If you’re asking an AI tool to listen in on your doctor-patient visits and then write summaries of those appointments (one of the use cases discussed in the article), that’s generative AI. Or, if you feed patient data into your AI tool and ask for a diagnosis (another use case from the article), that’s also generative AI.

I agree with the author: It’s really unclear when that technology is ready for primetime. But that’s why I think the article misses the mark, talking unfavorably about a branch of AI that’s simply not yet commercially ready and thus collectively condemning AI.

Interviewer – Jeff Solis:

Okay, so you’re implying there’s another branch of AI technology, one that’s ready today for the healthcare tech stack. Please tell us what that is, how it differs from generative AI, and how it can improve operations for healthcare organizations.

Johnny Hecker:

Readily deployable and truly impactful technology for healthcare right now is “extractive AI” facilitating desperately needed interoperability and reducing administrative burden.

Like generative AI, it uses advanced tools and capabilities such as Large Language Models, Natural Language Processing, Robotic Process Automation, and even the more basic Optical Character Recognition. But with extractive AI, you’re not asking the software to generate new content, only to expertly read, understand, and extract data from existing sources and then move that data to where it needs to be.

And because so much data still moves between healthcare providers in unstructured formats – such as handwritten notes or faxes – there’s a significant opportunity for providers and the entire healthcare ecosystem like payers, pharmacies, etc. to increase efficiencies, save time, and improve patient care by implementing extractive AI solutions.

Interviewer – Jeff Solis:

Can you describe a concrete example of how extractive AI could benefit a healthcare organization?

Johnny Hecker:

Imagine you’re a nurse on duty at a hospital. A primary care physician has just faxed the records of a patient recently admitted to your facility, and those records arrive as a 20-page document on the fax machine at your nursing station.

That’s a fairly common scenario. And for most hospitals, the workflow that follows is manual, slow, frustrating, and prone to error.

As the nurse who drew the short straw and has to process the inbound fax, you’ll have to pull the document off the fax machine, review its contents, and then open your EHR to begin populating the patient’s record with the details in the fax. If we’re talking about a new patient, you’ll need to first create a new electronic health record and then add the fax’s data to the relevant fields.

Among the problems with this workflow is that healthcare faxes often contain handwritten notes from the provider, illnesses described using different terminology, and a general lack of data standards that make transcribing the information from fax to EHR very inefficient and time-consuming.

A study published by the National Institutes of Health, for example, found that hospital nurses spend an average of 162 minutes per 12-hour shift entering data into the EHR. That’s 22.5% of every shift sitting in front of a screen – and not with patients.

Of course, not every one of those 162 minutes reflects time spent manually entering data into a patient record after reading it on a fax document. But you can see the problem.

Interviewer – Jeff Solis:

Actually, that inefficient workflow can lead to several problems, right? What are the other negative effects a healthcare organization might face by continuing with its legacy processes of manual data entry?

Johnny Hecker:

In addition to wasting the time of highly productive clinicians, manual EHR data entry can also contribute to delayed patient care. Keep in mind, this slow and inefficient process doesn’t even start until someone at the healthcare facility grabs the fax and begins reviewing its contents. At a busy hospital, that fax might sit unnoticed for hours on the fax machine. If the document contains critical information about a patient, those delays can undermine the patient’s care and health outcomes.

Also, as I alluded to earlier, manually copying data from one source to another introduces the risk of human error. Let’s return to that busy hospital nurse, who might be reviewing one patient fax after another as quickly as possible and logging the information into the EHR. Isn’t it plausible that this nurse could make a mistake when typing in the name of a patient or medication – especially considering the fact that the provider might have scribbled critical information or simply ‘urgent’  on the fax using handwriting?

Finally, let’s consider the overall morale and job satisfaction of your hospital’s clinical staff. Leaving your nurses to manually input every paper-based patient record into the EHR creates a high-stress burden on those professionals. I read in a 2023 study by AMN Healthcare that 78% of nurses want their employers to “reduce the documentation burden” and say doing so would play an important role in lowering their work-related stress.

When you consider the fact that healthcare organizations are facing the massive problems of clinician burnout and shortages of good candidates, you can see how implementing an extractive AI solution – one that significantly reduces the manual-data-entry burden – can help an organization attract and retain quality people and increase their job satisfaction.

In fact, you could make a strong case that implementing the right AI can directly help your healthcare organization overcome the worker shortage issue.

Let’s say you add the right AI solution into your hospital’s IT environment, and as a result allow your nurses to take only a portion of the 22.5% of their shifts previously spent on data entry and use that time for more productive work, like caring for patients. Isn’t that essentially the same as adding that amount of additional  nursing staff  who can spend time with your patients?

Interviewer – Jeff Solis:

Okay, this artificial intelligence capability sounds extremely valuable for a healthcare organization looking to improve its operations. But what about the fears discussed in the LA Times article? Does extractive AI pose any of those same risks?

Johnny Hecker:

Fortunately, no. There is no hallucination. In fact, the solutions we provide deliver a confidence score in the rare case of doubt to prevent errors from happening. This is a proven concept for healthcare deployable and in use as we speak. 

Bottom line: If you run a healthcare organization and want to take advantage of the massive benefits that AI can offer you today – improved workflows, added efficiencies, cost savings, happier employees, better patient outcomes – you can do so easily by implementing the right extractive AI solution, like our Consensus Clarity technology.