3 takeaways from Veradigm, ScienceIO CEOs on the potential for AI to transform EHRs

Veradigm acquired ScienceIO earlier this year to build out healthcare AI solutions.

Interim CEO at Veradigm Shin-Yin Ho and Will Manidis, CEO and founder of ScienceIO, spoke about the potential of artificial intelligence on a panel at the American Telemedicine Association’s Nexus conference in Phoenix. Here are three takeaways from their conversation.

  • Electronic health records have acted as filing cabinets with little to no regard to how the information was structured inside that cabinet. AI will change that.

Ho: “In the very early years, it was all about trying to get people to adopt it, about putting information in; we didn't terribly care how the information went in as long as it went in. And so there wasn't a thought process around how to structure data, it was only about can we just capture as much information as we could. And then it would sort of sort itself out later."

Manidis: “If you go and read clinical notes, these are massive unstructured data, right? And if you want to ask and answer questions, it typically requires having a human read that data and make sense of it. Think of generative AI as the world's best paid intern to make sense out of that.”

  • Large quantities of structured data could make a big difference in healthcare for macro and micro trends.

Ho: “Early in COVID, people came to primary care and started saying that they were losing their sense of smell. And if you think about it today, we all associate that with COVID. But at the time no one was thinking about it … sense of smell would never have been captured in a structured space or a field inside of an electronic health record, it would have been captured inside of the unstructured clinical note … And only if you could have structured it could you have pulled it out and then counted it. 

“The interesting thing about the moment you structure information, something like a symptom like a loss of smell, is that now you can count it. And if you can count it, you can quantify it. And if you can quantify it, you can map out its trend.”

Ho continued, “If you have more time to think about your patient, and you have more time to think about the information now that it is structured for you, you can start looking for patterns, you can start wondering whether or not there were better ways to maybe pivot and change on a care plan."

Manidis: “I think [AI] is going to look like quality of life improvements for providers, things like automatically generating notes, automatically filling out prior authorization forms, automatically doing intake and outtake forms. But over time, as we scale quantity of care, I think we will also be able to scale quality of care, right, because it's not only about freeing up that time, but the incredible things you can do when half your encounter is not just looking at a screen typing in notes."

Ho said that in a few years, AI tools will be able to be leveraged to power clinical research because of how they are structuring and organizing clinical data to be accessible to researchers.

He added, “So we look out two, three years out, you are really powering the research industry because what you're doing is you're already capturing information at a higher quality level. It's already structured, and you can count it and quantify it. That means going forward, you can start saying, 'Listen, I would like to ask certain questions. Are there certain subpopulations that actually have comparatively better outcomes, and others are comparatively worse outcomes than others? Is there something here in this constellation of symptoms?'"

  • EHRs will likely have to significantly change to accommodate AI in the future. 

"Over time, you'll start focusing on the data itself, the data layer, is it really going to be a record that we're looking at? Or is it going to just be inputs that are coming in, that you can pull together and populate any form that you want going forward? ... Over time, it doesn't matter what the tool is that captures it. As long as the information that is being pulled together is able to be pulled together to support care, or to support research, it becomes more of a meta question around data," Ho said.

Manidis added, “It's going to take certainly a reengineering of what the EMR looks like and what our relationship with the environment looks like.”