[ad_1]
Amazon has taken its first steps into offering AI to the healthcare sector with a tool that automates note-taking during patient consultations and creates summarised clinical notes.
The new service, launched at Amazon Web Services’ New York summit, is called HealthScribe.
The AI was able to separate out a patient’s chief complaints, as well as the doctor’s assessment, and details of a treatment plan.
It was also able to identify the roles of the speakers, including if there was a caregiver in the room.
Healthcare software vendors including 3M, ScribeEMR, and Babylon were already using the new AI.
The system is intended to reduce the amount of time doctors spend recording details of consultations.
Clinicians had the opportunity to read the notes prior to filing them.
The system does not currently have the ability to assess the clinician’s diagnosis for any symptoms that might have been missed, or to suggest other diagnoses that might not have been considered.
Tehsin Syed, who is general manager of AWS Health AI, said an AI-powered diagnostic tool was not currently being considered, but the ability to create such a tool was possible.
He said compliance requirements were a driver of burnout among doctors.
“For every hour they spend with a patient they spend two hours with documentation,” he said.
The data that HealthScribe accumulated from doctor-patient interactions would not be used to train the AI. Insights and data would be owned by the companies that used the tool.
AWS already offered some services in the healthcare space, but these did not use generative AI.
Supplied
AWS vice president of data and machine learning Swami Sivasubramanian announces a range of new AI tools at the company’s New York Summit.
Previous healthcare sector-specific programs created by AWS included HealthLake, which allowed companies to securely store and analyse health data.
Information used in the AI-generated clinical note included references to the original consultation transcript, which was intended to make it easier for doctors to check the AI’s summation was accurate.
Charges for HealthScribe were usage-based, with charges based on the number of conversations it summated, rather than by the number of patients or physicians using it.
Third parties could also adapt the program so that images from the consultation or follow-ups were included in case notes.
A number of other tools were unveiled at the summit, including AIs that could create graphs and do data analysis based upon natural language cues, and AIs that could act as virtual agents, corresponding with customers.
[ad_2]