Andy Jassy, Amazon AWS
Amazon‘s next big step in health care is with voice transcription technology that’s designed to allow doctors to spend more time with patients and less time at the computer.
At Amazon Web Services’ re:Invent conference on Tuesday, the company is launching a service called Amazon Transcribe Medical, which transcribes doctor-patient interactions and plugs the text straight into the medical record.
“Our overarching goal is to free up the doctor, so they have more attention going to where it should be directed,” said Matt Wood, vice president of artificial intelligence at AWS. “And that’s to the patient.”
At last year’s re:Invent, AWS introduced a related service called Amazon Comprehend Medical, which “allows developers to process unstructured medical text and identify information such as patient diagnosis, treatments, dosages, symptoms and signs, and more,” according to a blog post.
Wood said the two services are linked and can be used together.
Voice-to-text transcription is one of the many areas where Amazon is battling with cloud rivals Microsoft and Google. All three companies operate speech assistants that can in real time translate spoken words and sentences and offer text translation. Businesses can use the technology in a variety of ways to weave into their applications.
AWS’ software is designed so that it can be embedded into any device or an app via an application programming interface. It’s only available to customers using the AWS cloud, which presents a potential limitation because of the rising competition from Microsoft and Google. Microsoft Azure is working on similar tools with Nuance, and Google is researching the space with Stanford University.
In the medical sector, many doctors today rely on legacy dictation software that still requires them to spend hours on clinical documentation. Others rely on costly human scribes or will dictate notes into a recorder and then submit the voice files to a third-party transcription service, which can take a few days to return a response. Wood said Amazon’s service even has built-in punctuation, so there’s no need for a doctor to say out loud that a comma should be inserted.
The technology was developed with the help of some AWS customers, including electronic health IT company Cerner and Suki, a venture-backed transcription start-up. Wood said the company created the software because there was a “lot of demand for it.”
As Amazon moves deeper into the $3.5 trillion medical sector, it is juggling working with partners to develop tools on its behalf with potential initiatives that might someday prove competitive to the incumbents.
Amazon is taking aim at the pharmacy supply chain with its PillPack team and is looking to improve health-care services for its employees with Haven, a joint venture with J.P. Morgan Chase and Berkshire Hathaway. The company also has a primary care group, Amazon Care.
A big challenge for Amazon, a huge consumer company with tons of customer data, is ensuring that its health-care tools are compliant with privacy rules and regulations under the Health Insurance Portability and Accountability Act and when it comes to transcription, maintaining an extremely high level of accuracy to avoid problematic outcomes or potential liability. Imagine, for instance, if the machine learning system inputs the term “hyper” instead of “hypo,” or if doctors noticed so many inaccuracies that they ended up doing the work manually anyway.
Wood said the service is HIPAA compliant. He said it took a lot of work for the technology to correctly annotate the “domain specific language and abbreviations” that are common in the medical field, and added that the accuracy is very high. Amazon hasn’t published research showing how its accuracy compares with other offerings, but Wood said the company hasn’t ruled it out.