Dive Brief:
- Amazon has become the latest tech giant to announce a clinical documentation service that allows providers to automatically create medical notes using generative AI.
- The Amazon Web Services tool announced Wednesday, called HealthScribe, allows providers to build clinical applications that use speech recognition and generative AI to create transcripts of patient visits, identify key details and create summaries that can be entered into an electronic health record.
- HealthScribe is being previewed for two specialties: general medicine and orthopedics. An Amazon spokesperson said AWS could expand to additional specialties based on client feedback. HealthScribe costs users a set amount per second of audio processed each month.
Dive Insight:
AI-based clinical documentation is an increasingly crowded market, including Microsoft-owned Nuance and Google partner Suki — both of which integrated generative AI into their notetaking products earlier this year. Proponents hail the algorithms, which can create original content, as a game-changer for healthcare delivery and operations.
The goal of medical scribes is to pare down documentation burden on clinicians. Physicians can spend up to six hours per day logging notes into an EHR, cutting into time with patients and contributing to burnout.
However, documentation companies like Nuance and Suki have yet to publish accuracy measures for their AI-backed transcription tools, sparking concerns about quality.
Amazon did not respond to questions about accuracy rates by time of publication.
During a press briefing Wednesday at the AWS Summit in New York, AWS VP of Product Matt Wood noted HealthScribe cites the source of generated text in the summary from the original transcript, making it easier for doctors to review clinical notes.
It’s theoretically possible that a generative AI system could become advanced enough to remove humans entirely from the process, but it’s unclear if that’s something provider clients would want, according to Wood.
”I don’t know what level of automation physicians will be comfortable with,” Wood said.
Clients that plan to integrate the service include 3M Health Information Systems, which says it manages software used by more than 300,000 clinicians; population health company Babylon, which serves roughly 1,000 providers globally; and virtual medical scribe provider ScribeEMR, which says its clients span “hundreds” of practices.
AWS declined to share exactly how many providers are using HealthScribe.
HealthScribe is not Amazon’s first foray into the medical transcription space. In 2019, AWS launched Transcribe Medical, a voice-to-text technology meant for clinical visits. Unlike HealthScribe, Transcribe Medical does not autopopulate relevant information into the EHR.
Generative AI is a significant source of interest for healthcare companies looking to curb costs while improving care quality and operational efficiency.
Startups with generative AI solutions in healthcare delivery and life sciences have collectively earned more than $20 billion in funding, according to a recent market analysis by health tech-focused VC firms.
Over the past few months, tech giants have announced rapid-fire partnerships with EHR vendors and hospitals to develop and test generative AI tools. Use cases run the gamut from automatically drafting replies to patient messages, to more easily querying databases.
Microsoft recently announced integrations of its generative AI software, which includes Nuance’s notetaking products, with EHR vendor Epic and telehealth provider Teladoc.
Meanwhile, Google has made its own large language model, called Med-PaLM 2, available to a select group of customers, and announced a partnership with Mayo Clinic in June to explore generative AI’s applications in the hospital.
For-profit hospital giant HCA is also piloting ambient documentation software in the emergency rooms of two hospitals through a partnership with Augmedix.
Amid the hubbub of activity, concerns about accuracy, bias and oversight have given rise to some safety-focused organizations advocating a slow-down on implementation, including a new transatlantic Responsible AI in Healthcare consortium that advocates for safe AI in hospitals.
The American Medical Association also decided to develop recommendations regarding the use of AI at its annual meeting earlier this year.
