Khunatorn -

How Artificial Intelligence EHR Integration Improved Patient Safety

The health IT solution with artificial intelligence (AI) capabilities helped a PA health system improve patient safety through more complete medication history records.

A Pennsylvania health system improved patient safety and cut costs through the integration of a health IT solution with medication transcription and translation artificial intelligence (AI) capabilities.

Prior to the EHR integration, patient records from external medication history sources were often incomplete, inconsistent, or missing at WellSpan Health, Robert Lackey, MD, FAAFP, CMIO told in an interview.

This required a convoluted medication transcription process; employees had to reach out to pharmacists and other providers to gather medication data and then manually enter it into the EHR, presenting the opportunity for transcription errors and potential adverse drug events (ADEs), he explained.

At the end of 2020, WellSpan integrated MedHx, a health IT tool developed by DrFirst that provides a comprehensive medication database comprised of local and national medication history sources, including HIEs and EHR partners, directly in the native Epic workflow.

The technology also identifies and connects local pharmacies and healthcare organizations that share mutual patients, making the dispensed fills available within the EHR and allowing for more comprehensive patient medication records.

While this solution aided in the medication transcription process, WellSpan still experienced the administrative burden of EHR translation.

Traditional patient data exchange is fraught with what Lackey referred to as the “yellow triangle syndrome.”

When patient data is sent from an EHR, it often comes in a different language than the receiving organization’s EHR, spurring yellow triangle notifications that indicate a manual data translation is needed.

For instance, one EHR may use the terminology “take by mouth” in a medication sig while a different EHR may say, “The route is oral, not by mouth,” he noted.

“It's those subtle differences that there needs to be a translation for,” Lackey explained. “It can take a ton of clicks and scrolling and selecting dropdown menus for a human to do that translation manually. If you're expected to do a bunch of work, it decreases your efficiency, creates opportunities for errors, and that’s where your patient safety issues can show up.”

To automate the translation process, WellSpan leveraged the vendor’s AI solution SmartSig, which takes prescription transaction data and translates it into the native EHR nomenclature.

“Knowing that Epic doesn't use the word ‘by mouth,’ it uses the word ‘oral,’ it changes the word for us back into discrete data that is in the Epic terminology,” Lackey said.

The automated system saved clinical time and cognitive effort, allowing employees to focus on providing quality care, Lackey explained.  

WellSpan also put the AI solution to work in its third wave of go-live for transferring medication history data from three separate EHRs into one EHR, Lackey said.

During the health system’s first two waves of go-lives, executives had to make significant staffing investments to manually translate data from the old system into the new system.

“We had to pay consultants to help fill in some of the staffing needs, and of course consultants are expensive,” Lackey noted. “We would get our staff gathered into a central location so we could use their proximity to each other as shared knowledge to say, ‘Hey, I'm working on this patient from EHR X, but I don't have experience with EHR X. Can you help me understand what I'm looking at here?’

“We learned how much that costs, and it's very expensive,” Lackey continued.

With the new constraints of COVID-19, WellSpan was not able to gather all those people into a single room to collaborate on the translations during the third wave of go-lives.

“What we were able to do, especially for the medication history portion of the data, was take the continuity of care documents, or CCD extracts, out of each of the legacy systems and present that as the source of data to the staff,” Lackey said. “That helps to eliminate the need for employees to know how to use the particular EHR; they could work primarily from the CCD extracts.”

Lackey said that WellSpan staff still had translation issues as some of the data was in disparate nomenclatures, making the task of bringing the old data into the new system burdensome.

Then, they came up with an idea.

“A light bulb went off and we said, ‘Wait a minute, we have this vendor that has this artificial intelligence engine that can convert all these medication histories into the nomenclature we need it to be in; what if we ask them to run that artificial intelligence engine on our legacy data?’" Lackey explained.

The vendor and WellSpan worked together to simplify the EHR medication translation process. The health system sent the medication portions of the CCD extracts to the vendor. Then, the vendor ran the CCDs through the SmartProcessor to normalize medication data across three legacy systems as it transitioned multiple facilities to one EHR.

The AI modifies itself to become more accurate, Lackey explained. For example, the system learns over time that the only way atorvastatin pills are ever taken is by mouth, so it's safe to fill in that gap and change it from one daily to one by mouth daily, he said. 

At first, translating each patient record into the new EHR took an average of 20 minutes, Lackey said. The health IT integration shaved off five to seven minutes.

Lackey noted that the time WellSpan saved turned into money.

“I think we spent close to 40 percent of our budget,” he said. “That budgetary savings wasn't entirely due to this, but partly it was.”

WellSpan also saved workforce costs because staff worked remotely and the health system did not need to hire expensive consultants.

Lackey said that the health system plans to continue improving patient safety by applying the same type of logic to medication refill requests to make the renewal process less labor-intensive and more efficient.

Next Steps

Dig Deeper on Interoperability in healthcare

Cloud Computing
Mobile Computing