AI in the exam room

Can artificial intelligence help reduce administrative hassles in the doctor’s office?

AAFP Innovation Lab pilot tests a digital assistant that promises major reductions in clerical work, better doctor-patient relationships, and improved practice profits

By Brian Justice


Laptops are not known for their bedside manner. According to the AAFP’s annual member satisfaction survey, the administrative burdens posed by EHRs make them more of an impediment to patient care than a useful tool. A machine stands between — or interferes with, rather — patients and their doctors, who are facing enough troubles these days.

This clerical burden takes up more than half of physicians’ time while, simultaneously, practices are being forced to adapt to population-based care and alternative payment models. Enthusiasm for one’s chosen profession falls as financial risk to the practice rises. Artificial intelligence that does not optimize the character of family medicine impedes the physician and degrades the patient experience.

Andrew Carroll, MD, a family physician in Phoenix, Arizona, voiced a common complaint. “We were looking for a solution to help us get data into the notes. We tried various voice-recognition software options, including Dragon Medical, obviously, and some of the built in Apple tools to dictate directly into a laptop, but they just became untenable,” he said. “So, we were still doing a lot of work after clinic because we were trying to recall the histories and physicals of patients we saw up to 10 hours prior.”

“Today’s EHRs have eroded the experience rather than enhancing it,” said Steve Waldron, MD, MS, and AAFP vice president and chief medical informatics officer. “Our vision is to help family physicians care for patients while using health IT that works for them, not against them.”

Ideally, family medicine is based, and thrives, on profound patient-physician interactions. But technology isn’t going anywhere, and neither is the long-standing and ever-growing need for technology developed and implemented specifically for the family medicine environment.

Pilot program development for dramatic change

In October 2018 the Academy board authorized a 42-month special project to address issues around EHRs. Designed to drive innovation in family medicine, address administrative burden, improve value-based care delivery, and start a long-term conversation with AI and machine learning communities, the goal was to make family medicine a true partner in driving the agenda in that space, especially when it comes to clinical notes.

Clinical notes are now being used in billing, coding, reporting, legal uses, and more, and to address this cumbersome fact of life for the family physician, the Academy partnered with Suki AI for a pilot program.

“We scoured Silicon Valley and other places looking for solutions that had demonstrable improvements; not in the 5% to 15% range…we wanted improvements of up to 70%,” Waldron said. “Suki was chosen for several reasons. The company was founded with the goal of helping physicians spend more time caring for patients and less time on administrative tasks.”

The goal, according to Nathan Gunn, MD, and chief operating officer of Suki, is “to develop a virtual physician’s assistant that can take care of the scut work, and help doctors focus on what they went to medical school to do: take care of patients.”

Creating a solution from the physician perspective

Finding the right IT partner was as crucial to this endeavor as finding the right group of physician partners.

Punit Soni, Suki’s founder and CEO, spent seven years with Google before serving as VP of product management for the company’s Motorola division, but he wanted to build something of his own. He decided to develop a digital assistant for physicians that combines artificial intelligence and voice-enabled technology.

“I was shadowing a family physician in a hospital,” Soni said. “I was amazed to see all the clicking of boxes and how feverishly they were putting everything together.” It seemed to him that a sophisticated group of professionals had been reduced to data clerks.

To ensure that the next generation of health care technology was truly relevant, he determined that the product must be developed not only with the end users in mind, but with physicians actively participating in its creation. “A key aspect of a product like Suki digital assistant is having a keen understanding of doctor workflows,” he said. “How does the doctor practice medicine?”

Digital assistants fall into a new category of products designed to reduce the burden of documentation. They are somewhat akin to Alexa or Siri, using voice recognition, natural language processing and AI that listens, learns, and adapts to a practice. And in practices unable to afford additional human resources, the relief offered by eased clerical and documentation burdens can be immeasurable.

Suki was chosen for several reasons. Its nature aligned with the goals of the initiative and was already being actively and successfully marketed to primary care and family physicians. It is easily adaptable and does not require new hardware; it is an app that can be downloaded and used on an iPhone or, for Android users, the Chrome browser.

The pilot program was conducted on the athenaOne EHR because it was already in use by physicians currently using Suki, which made its adoption and implementation quick and easy, and it came at a reasonable price. To sum it up, it was already integrated, readily adaptable, rapidly implemented, and affordable.

“Athena’s model involves real partnership with the practice to generate revenue,” Waldron said. “They are open to innovation that will help decrease administrative burden while increasing the quality and services that physicians provide in their practices.”

More time with patients, less screen time

Suki does not replace the laptop, but it does end the non-stop typing. The physician tells the Suki digital assistant what to document in the note, and Suki includes the ability to pull information from the patient’s record. Not every bit of data will be available, so the physician will still need to look things up. But the constant typing while trying to record every note during a patient visit will be reduced, if not eliminated.

“While I’m in the room with the patient I can do a prescription or make a referral, then step outside the room and dictate for maybe a minute, getting everything down on Suki’s built-in templates, so I don’t have to dictate the entire exam,” Carroll said. “So, I can go to my next patient right away and my relationship with both patients has been enhanced. I can spend the time with them that I would have spent worrying about the administrivia that I would normally have had to do afterwards.”

Suki also utilizes machine learning, meaning that it is not simply listening to a sound and trying to convert it to a word. Its intuitive capabilities include understanding words and placing them within context. It also understands and incorporates punctuation.

“In the beginning it was a little slow, then it pretty quickly learned how we spoke and how we dictated,” Carroll said, speaking for his three-person office. “Also, two of us have distinct accents, but within a few days Suki was able to recognize that and quickly became pretty much error-free.”

The software is updated in all but real-time. Humans monitor flagged notes, correcting them if necessary and ‘teaching’ Suki in the process. “So, the model is continually being retrained, which they have to be because we are continually learning new things,” Waldron said. “If you had said ‘COVID-19’ in early 2019, it may not have known what that was. Now, because it’s in common use, it knows exactly what it is and how to incorporate it into the EHR.”

Dramatic improvements in both efficiency and quality

The results of the pilot program have aligned perfectly with the goals set by the Academy. Participating physicians recorded a 62% decrease in documentation time per patient, a 51% decrease in documentation time during clinic, and a 70% decrease in after-hours charting.

“I can say that I have literally reduced my work after clinic by about 90%,” Carroll said. “I used to spend three or four hours after work, just finishing notes and stuff. Now, I’m doing it for 15 minutes to 30 minutes and I’m done. It’s really wonderful.” Every participant noted a dramatic increase in their satisfaction with notes completion, time savings, and EHR use for other administrative tasks, and across the board they found it most useful with notes that required anecdotal or complex history and assessments.

“In medicine, we talk about being able to follow someone’s thought process from reading their notes, especially early on when you’re a resident,” said John Abraham, MD, of Tulsa, Oklahoma. “With Suki I can actually document my thought process, and when I go back to my notes I can see why I changed the medicine, or didn’t start one just yet. I feel that if someone had to take care of my patients when I wasn’t around, there would be good explanations and directions for them to follow.”

The platform is available on a subscription basis, with no upfront setup fees or long-term contracts, and users have automatic access to updates and improvements. In fact, early on in the pilot program, user feedback could be incorporated within 15 minutes to an hour. By the end of the program, those changes were made in nearly real-time. “That’s one of the nice things about the subscription model,” Waldron said.

Technology so good, you don’t even notice it

“We were thrilled with the pilot results and excited to roll out Suki to a larger cohort of doctors,” Gunn said, and a Phase 2 study has begun with the participation of 100 physicians. “Our view is that AI does not have to ‘replace’ or take over key diagnosis and treatment decisions from providers to transform health care delivery. Ultimately, the goal is for Suki to function like an outstanding resident, following you on rounds without needing any commands to be issued by the physician. Our mission here is to develop technology so good you don’t even notice it.”

The cohort of doctors in the Phase 1 study was small, but they are extremely pleased across the board.

“I’ve been using Suki for just over a year and it has become indispensable to me,” Abraham said. “I’ve become very accustomed to the flow I’m able to maintain, which helps me prioritize things that I like to get done in a certain order and not take too much work home with me.”

More will be known after the completion of Phase 2, but the Academy’s assumption is that AI-based digital assistants are going to become an essential technology.

“I anticipate big things and more growth, and I hope that other providers have a similar experience,” Abraham said. And whether the ultimate solution comes from Suki or another tech firm, the pilot programs are providing dynamic and robust solutions for family physicians. More importantly, they are driving the development of other product innovations that meet the unique needs of family medicine.