With the technology of EHRs has come a massive amount of documentation requirements for doctors.  Can AI technology be part of the solution?

According to Dr. Matthew Fradkin a pediatrician at Swedish Medical Group in Seattle, the answer is, well, possibly.

Dr. Fradkin described his attempts to use so-called “AI scribes” in a recent HIMSS20 Digital Event.

“I want to find ways to make caregiving easier for our providers,” Dr. Fradin explained his underlying motivations, “and by finding ways for our providers to care for patients the way they want to in a large healthcare system, pushing towards standardization for population health advances. Or basically how do we prevent provider burnout using various digital tools and technology.”

Part of making caregiving easier is by helping providers get a handle on administrative “pain points,” or, more particularly, the burden of documentation for which providers are primarily responsible.

“We want ambient technology to provide an accurate note where providers don’t change a thing in how they deal with patients in the clinical room or in their head,” he said, describing his decision to consider AI and machine learning scribes and to work with the digital innovation team at his health system to look into appropriate pilot projects.

The tech that Fradkin decided to pilot was an AI-based medical scribe that is able to train and learn a provider’s individual style and preferences over time.

“Part of the digital innovation core of Swedish and Providence St. Joe's is the accelerated pilot process. This is a set framework that allows providers to investigate possible new technologies to help with their clinic experience,” he said. “This pilot allowed us to follow that framework and to evaluate possible vendors in a virtual arena, come up with KPIs or key performance indicators and streamlining the path through IT, security, legal or any other red tape that prevents pilots from occurring in large health system.”

At the beginning of the pilot, Fradkin explained, providers weren’t responsible for actually training the machine learning algorithm. Rather, there was “a real-person reviewer offsite, reviewing what the ML (was) coming up with after a visit, and correcting it and making adjustments in the background based on my personal and organization-wide templates already in the system.”

The system also allowed providers to choose how they would like to use it, with some using it for the simple transaction notes, while others used a “snip-it” mode, where they said a phrase that helped the system choose a certain path.

The upshot?

“During and after the pilot I was able to go from seeing 16 to 17 patients a day to 23, with a range of 22 to 30 with the entirety of the pilot – where, for return on investment with the particular product we used, I would have to see one patient extra a week, four extra patients a month, to pay for the product,” Fradkin said. 

Looking ahead, Fradkin noted, “Venturing into this new realm of AI-driven scribes is an exciting one, and one that needs to provide caregivers flexibility and responsiveness – the same factors parents expect from their providers in this new digital age of medicine.”