CHICAGO – Two generative artificial intelligence announcements from electronic health record vendors – Epic and eClinicalWorks – using Microsoft’s Azure OpenAI Service, launched in March – coincided with HIMSS23.
EHR and customer relationship management platforms – like Salesforce, which announced its Einstein GPT for CRM in early March – are integrating generative AI into their software to allow provider organizations to search and summarize contextual information with natural language searches powered by cloud computing.
With Azure, EHR developers can integrate custom large language model experiences directly into their own applications, according to Microsoft. They can imbibe existing bots and other tools with the power to use conversational language “to make life simple for end users,” as eClinicalWorks described it Tuesday.
eClinicalWorks’ GPT integrations
In a live EHR demo Tuesday, eClinicalWorks walked Healthcare IT News through operational queries, actions on patient telephone encounters and progress note creation with a new AI-powered virtual assistant that’s in development.
Moving to public cloud APIs was a milestone for the company about a year and a half ago, and it was the basis for “us to do other evolutionary things,” said according to Girish Navani, the company’s CEO.
“It does magic that helps our customers truly benefit from public cloud computing.”
The GPT API in the Azure cloud was the next logical step, Navani explained after the team’s demo.
It “gives us the ability to start looking at some of the applications like a summarization of health records, autoresponders, allowing us to query information with natural language.”
For example, when a telephone encounter is created, the assigned nurse will be able to use the AI tool to create a response based on the patient’s message.
The tool provides recommendations based on what the patient is calling for. The nurse reviews the choices, and then selects the response and adds it to the record with a single click.
Support staff can then call the patient back with the nurse’s instructions, according to the demo.
Doctors too will be able to use eClinicalWorks GPT to create progress note summaries and publish them to the patient record. They can also print results, such as exercise instructions, for the patient.
Integrated AI can help providers with revenue cycle management by searching quickly in the EHR for reports and then exporting the data.
The tool cuts a number of steps when trying to compile and sort patient data – there will no longer be a need to create subsets of data, according to the demo.
For example, a search for the type of visits and condition – such as “dental visits for patients with asthma” – pulled up the result instantly. Longer queries, such as, “Show me a list of all male patients between the ages of 20 and 60 who have diabetes and have not been seen in the last six months” appeared on the screen with one click.
Currently, that same result requires a number of searches to drill down.
Efficacy will depend on the user’s data quality and documentation is the key, the company said, noting that security settings will allow providers to restrict who has access to the AI in their workflows. Administrative logs also indicate when any user triggers an action.
“You will see eClinicalWorks put a lot of guardrails around the release of this technology,” Navani said.
“We don’t believe that technology that can start hallucinating and giving incorrect responses will be acceptable to the community at large.” There will be clear footnotes to providers, and they will have to read and accept GPT outputs at every step. “It should not be the pilot.”
The company said the LLM is still being trained, and the GPT feature is scheduled to be pushed out and available next quarter.
Navani noted eClinicalWorks is also taking other tools like Scribe further with natural language processing, and converting PDFs and other documents “into intelligent data objects to route them to the right person,” he said.
Epic’s work with Microsoft and Nuance
Epic also had a demo ready to show off how new GPT integrations will improve clinician efficiency at HIMSS23.
The EHR vendor is developing a number of automations using LLMs, neural nets and predictive models on its Nebula cloud.
Because there are a lot of unknowns with LLMs, the company says it is focused on accuracy, user experience and ethics. The approach the vendor is taking is to ensure that users understand what they are working with, hone in on the right use cases and learn how use affects the patient and clinical care.
In production now is a new two tool using Azure’s HIPAA-compliant GPT3 that supports clinical efficiency and organizational effectiveness, according to the demo on the HIMSS23 show floor.
The “in-basket” integration to support answering medical questions from patients is being tested with some customers, including UC San Diego Health, UW Health in Madison, Wisconsin and Stanford Health Care.
If a patient asks for an update on their thyroid medication, the automation can decipher which of the patient’s prescribed medications is for the thyroid and how many refills are remaining, and provide options regarding what the patient can do, according to the demo.
It saves time in researching the chart, but the clinician cannot just hit send. They have to choose “Start with draft.”
Clinicians can also prompt the API for instructions, such as for colonoscopy prep. Epic’s integration allows the user to choose to apply results, resubmit the query to get a different result, summarize the result, adjust the language’s grade level and translate it into another language.
To make its Slicer Dicer data exploration visualization tool more accessible, the new GPT integration in development allows for more conversational questions to match concepts and model data, create summarizations and instantly generate line-level data.
Epic is working on other GPT integrations including in its Cosmos database.
“We are developing additional ways to incorporate generative AI across our applications, from ambulatory to inpatient to CRM to revenue cycle,” said Seth Hain, senior vice president of research and development for Epic. He mentioned tools that would generate easy-to-understand instructions for patients at different health literacy levels and provide summaries of opportunities to improve healthcare system operations based on analytics of their data.
Salesforce with GPT
With Einstein GPT, Salesforce customers can choose to run on their own data or their own external LLMs to generate information using natural language prompts directly within their Salesforce CRM.
Healthcare providers can pull data from the EHR into the CRM, and in Slack, care teams can discuss clinical and nonclinical data in a HIPAA-compliant environment. With the Einstein GPT integration, they can summarize chat information between multiple members of the care team and complete Salesforce CRM tasks.
The benefit of generative AI is answering simple questions that often take multiple steps, such as, “I have neck pain, which doctor should I go to?”
But to scale GPT in healthcare, a few things need to be worked out, said Amit Khanna, senior vice president and general manager of health and life sciences at Salesforce.
“We should be able to rely on and trust the source of information.”
On the road map, customers will able to use their own data or bring their own LLM in, so they can trust the AI’s source, but there is a lot to work out for healthcare, he said.
Andrea Fox is senior editor of Healthcare IT News.
Healthcare IT News is a HIMSS Media publication.