Thanks to a collaboration between Epic Systems and Microsoft, healthcare providers can now receive a time-saving hand from generative AI technology that drafts responses to patient messages and inquiries.
For years, Microsoft has allowed Epic to use the cloud hosting platform Azure as part of their longstanding collaboration, Garrett Adams, product lead for Epic’s Ambulatory Research and Development Division and member of the Cognitive Computing Platform team, said.
The recent expansion involves the use of Azure OpenAI, which Adams described as a large language model that is “very good at predicting the next word.” This specific technology has been trained over a large collection of information to draft responses to patient messages for providers.
“It is really a master of language,” Adams said.
Utilizing OpenAI will save an incremental amount of time for the provider to respond to patients, Adams said. This became especially relevant after the pandemic, which saw a large rise in telehealth services and patient messaging.
The asynchronous messaging model happened to prove itself so useful during the pandemic that it stuck around, Adams said. The goal now is to reduce the workload associated with responding to such messages.
“This overall has incredible potential in our industry to reduce the administrative burden, to automate mundane tasks and to give providers as much as that face-to-face time with patients as possible,” he said.
By using the Azure OpenAI service, Epic can leverage patient data and information securely because it is HIPAA compliant, Adams said.
Adams emphasized the AI technology provides draft messages, meaning the messages are not automatically sent. Healthcare providers can review the AI generated message, modify and then send it – or opt to draft a blank message as they would’ve before.
An example of the technology is when patients reach out to providers for information about medication refills. The AI brings in additional information from the chart in a secure, private way and combines that data with the patient request – along with some additional cues for sentence structure and tone to avoid sounding robotic, Adams said.
Generative AI is also being used in Epic’s reporting tool SlicerDicer to enhance conversational support, Adams said. Through SlicerDicer, providers can customize searches and find data on patient populations.
Prior to this technology, providers had to filter through information and hunt through directories but can now have a time-saving conversation with the software for information via chat or voice, Adams said.
“I think that as much as we can do to improve the happiness and reduce the burden on users – that’s just gonna pay dividends on the care that they’re able to provide for patients,” he said.
Three organizations are currently live with the Azure OpenAI message drafting software: UC San Diego Health, UW Health and Stanford Health Care. These organizations are experimenting with a subset of users that Epic has helped set up for success, Adams said.
“We’re taking a very thoughtful and incremental approach to how we roll this out,” he said.