AWS rolls out generative AI service for healthcare documentation software

Amazon Web Services announced Wednesday a new AI-powered service for healthcare software providers that will help clinicians with paperwork.

AWS HealthScribe uses generative AI and speech recognition to help doctors transcribe and analyze their conversations with patients and drafts clinical notes, the company announced Wednesday at its AWS Summit New York.

With the new service, healthcare software providers can use a single API to build clinical applications that automatically generate preliminary clinical notes by analyzing patient-clinician conversations. The service can create transcripts, extract key details, such as medical terms and medications, and create summaries from doctor-patient discussions that can then be entered into an electronic health record (EHR) system, the company said in a press release.

HealthScribe is powered by Amazon Bedrock and will make it faster and easier for healthcare software providers to integrate generative AI capabilities into their application starting with two popular specialties, general medicine and orthopedics, without needing to manage the underlying machine learning infrastructure or train their own healthcare-specific large language models (LLMs), according to AWS executives.

Executives said AWS HealthScribe enables responsible deployment of AI systems by citing the source of every line of generated text from within the original conversation transcript, making it easier for physicians to review clinical notes before entering them into the EHR.

Amazon's cloud division has been innovating with generative AI to keep up with the AI race. In June, the company launched a new innovation center focused on generative AI backed by a $100 million investment.

Three healthcare customers, 3M Health Information Systems, Babylon Health and ScribeEMR, are already working with HealthScribe, according to AWS.

'Our healthcare customers and partners tell us they want to spend more time creating innovative clinical care and research solutions for their patients while spending less time building, maintaining, and operating foundational health data capabilities,” said Bratin Saha, vice president of Machine Learning and Artificial Intelligence Services at AWS in a statement.

"Documentation is a particularly time-consuming effort for healthcare professionals, which is why we are excited to leverage the power of generative AI in AWS HealthScribe and reduce that burden," Saha said.

In an AWS blog post, Garri Garrison, president at 3M Health Information Systems said the company is collaborating with AWS to bring conversational and generative AI directly into clinical documentation workflows."AWS HealthScribe will be a core component of our clinician applications to help expedite, refine and scale the delivery of 3M’s ambient clinical documentation and virtual assistant solutions," he said.

At the Healthcare Information Management and Systems Society Global Conference back in April, 3M announced a partnership with Amazon Web Services to use machine learning and generative AI technology to advance automated medical notetaking and virtual assistant solutions for doctors.

ScribeEMR is a provider of virtual medical scribing, virtual medical coding and virtual medical office services for hundreds of medical practices, hospitals, and health systems.

"By harnessing the power of AWS HealthScribe, we can transform the process of healthcare documentation using generative AI. With AWS HealthScribe, our advanced processes can now capture and interpret patient visits more effectively and optimize EMR workflows, coding, and reimbursement processes," said Daya Shankar, co-founder and general manager at ScribeEMR, in the blog post.

AWS executives also said the company prioritized the security and privacy of healthcare data. The HealthScribe service gives customers control over where their data is stored, encrypts data in transit and at rest, and does not use inputs or outputs generated through the service to train its models, according to the company.

"Users have full control over their data and determine where they prefer to store transcriptions and preliminary clinical notes," AWS executives Jason Mark, Sarthak Handa and Tehsin Syed wrote in a blog post on Wednesday.

"AWS HealthScribe is designed to be used in assistive role with the goal of making documentation easier for medical practitioners. Each AI-generated summary sentence is linked back to the consultation transcript, allowing users to easily verify accuracy by cross-referencing the source and understand the context behind the AI-generated note," Mark, Handa and Syed wrote.

As interest in generative AI continues to grow, healthcare software vendors are looking to leverage this technology in their clinical applications to solve common pain points for clinicians. But working with generative AI is complex and requires significant engineering resources, making it challenging for healthcare software providers to bring AI-powered solutions to market quickly, AWS executives said.

To build generative AI capabilities, providers have to train or fine-tune their own LLM to generate accurate clinical documentation, which requires access to in-demand AI experts, massive amounts of carefully annotated healthcare data, and significant compute capacity. An LLM for healthcare also needs to be specially trained to understand complex medical terminology across different specialties.

Software providers must also build with responsible AI in mind, including designing the solution so that clinicians can trace the origin of any generated text to mitigate the risk of errors or hallucinations. 

AWS says it is taking on this heavy lifting with its HealthScribe service that consolidates many complex capabilities. The tool reduces the need for training, optimizing, and integrating separate AI services and building custom models, allowing for faster implementation, Syed said.

Alongside HealthScribe, AWS also announced the general availability of AWS HealthImaging, a service that makes it easier to store, transform and analyze medical imaging data "at a petabyte scale."