New gen AI system from AWS reduces paperwork, physician burnout

NEW YORK – Amazon announced the launch of its AWS HealthScribe at a company summit here in July. It introduced an AI-powered system that’s designed to save hours of time for physicians by automatically listening to encounters with patients and turning out full transcripts or summaries containing the most meaningful data.

The system uses generative AI, allowing users to engage with it in plain English. AWS says the solution will significantly reduce physician burnout by cutting the amount of time doctors spend taking notes and filling out forms.

“There’s a lot of time spent by clinicians on paperwork,” said Tehsin Syed, general manager of health at AWS. “It can take them twice as much time to fill out forms as the time they spend face-to-face with the patient. We want to reduce that.”

“We’re allowing them to focus more on their patients, as many of them are trying to do their paperwork at the computer at the same time they’re talking to patients.”

HealthScribe can even tell if more than one person is with the patient and captures their conversations, too. Often, a son or daughter – or a friend or caregiver – will accompany an elderly parent to an appointment to help. The system parses out each speaker and produces an accurate transcript of the conversation.

The Large Language Model used by HealthScribe has an extensive medical vocabulary and identifies medical terms in the encounter, such as specific medications, tests and procedures.

When using the HealthScribe summaries, doctors can track the source of every line in the original transcript, giving them trust in the notes. The summaries, furthermore, can be added to the physician’s electronic records.

Earlier this year, Microsoft and its subsidiary Nuance announced a similar product called DAX Express, which also makes use of generative AI to monitor the ambient sound in the doctor’s office to produce reports.

Syed said HealthScribe differs from DAX Express in that it’s an API to be used by partner companies to produce solutions, while DAX Express is a product for the end-users – namely, doctors.

3M and its subsidiary M*Modal is one of the first companies to use AWS’s HealthScribe; it will begin marketing a product in the Fall that can be deployed in doctors’ offices to help produce transcripts, summaries and reports.

Nuance and 3M/M*Modal are the giants in the dictation/transcription business, so it’s a natural progression for them to have created products that make these tasks even easier for doctors.

For its part, 3M claims about 300,000 doctors worldwide are using its dictation technologies, while Nuance says it has approximately 550,000 doctors who are customers.

Syed said that many applications will be developed using HealthScribe and AWS’s other generative AI solutions, such as Bedrock, a foundational system. Using these building blocks, the partner can more quickly design, test and market a solution.

“The partner doesn’t have to do all the work on the underlying Large Language Model,” said Syed. “They can go to market faster, and they can rely on the underlying security in the solutions.”

He commented that generative AI, the latest iteration of artificial intelligence, is different than previous versions in that users don’t have to be data scientists or I.T. experts. With earlier AI systems, a lot of labeling or tagging had to be done, something that was needed for the sake of accuracy but was time consuming.

Now, with gen AI, the systems understand natural language and can be used by the layman. “You don’t have to do all that labeling,” said Syed.

Indeed, the building blocks of gen AI systems have already been trained to understand human languages, whether it’s English, French, Urdu or any number of others. Companies can take these foundational components and create solutions for end-users much more quickly.