A pharmacy clinical-decision support tool that employs artificial intelligence, integrated into an electronic health records system, is providing tangible benefits for clinicians and pharmacists at one Maryland hospital system, its presenters explained in a recent HIMSS20 Digital presentation.

The system is saving clinicians time in finding answers to frequent common questions about medications and their administration that clinicians traditionally have answered either by using background knowledge to make educated guesses or by contacting the pharmacy department for answers.

Now, with the AI-based system, they're finding most answers quickly, using natural language to form their queries.

The adoption of the system at Peninsula Regional Medical Center in Salisbury, Maryland, is freeing up time for clinician-pharmacy department dialogues on more complex questions or patient-specific queries, said Rachel Cordrey, pharmacy supervisor for the organization.

Cordrey, along with Dr. Mark Weisman, the organization's Chief Medical Information Officer, provided insights during their HIMSS20 Digital presentation, Pharmacy CDS with EHR-Integrated AI Natural Language Search.

Peninsula Regional, a 289-bed, Level 2 trauma center, integrated the clinical-decision support tool, powered by Watson AI from IBM, into its Epic EHR. It is one of the first organizations to achieve that integration.

The move has helped it improve efficiency at the bedside and also increased access to evidence-based information on medications and how they can be safely administered, said Cordrey and Weisman.

This type of information is crucial for providers, who need nuanced answers when ordering prescriptions for patients, especially children or women who are pregnant. "We want to have good information, especially if we're counseling patients on adverse effects, and especially if they're taking a drug for the first time," Cordrey said.

Easier to use, more likely to be used

But prescribing clinicians are under immense time pressure, and they often are uncertain about precise prescribing information when they're making decisions, Weisman said.

"The questions that a clinician would like to look up tend to go unanswered … if a decision is not life-threatening, they may have to deter the research on that until after-hours or not at all. That led us to explore getting clinical-decision support into the workflow."

Cordrey said both doctors and nurses often have straightforward questions on IV compatibility, dosing and side effects that at least half of clinicians were calling pharmacists about several times a shift. An internal survey found that this manual process was taking one to three minutes per session, "and that takes time away from direct patient care."

Drug information databases provide an alternative, although they often lie outside the EHR, and they can take an average of 1.66 to 3.5 clicks to find information, according to recent studies.

"We're all about streamlining processes here – the ability to ask questions in plain language," Cordrey said. "The easier you make a tool, the more likely people are actually going to use it."

The technical work to integrate the Watson AI tool into Epic was easy, Weisman said. It was first tested in Peninsula Regional's pharmacy department in February 2019, before it was pushed out to other hospital departments later that spring.

Clinicians can access it via a MedHelp tab along the top of the Epic screen. It's contextualized to the questions and roles that nurses, doctors and pharmacists might ask.

A dialogue box appears on the right side of the screen, enabling questions to be asked in natural language. The CDS tool provides a series of answers related to the question. The clinician can then ask follow-up questions, and the tool understands those questions to be related to the previous query, which offers a natural dialogue on a topic.

In May 2019, the first month after introduction, the AI tool was used 489 times, and users who were surveyed about it overwhelmingly said it saved time, was easy to use, made their job more efficient and supplied trustworthy information.

Also, 92% of respondents said that integration into the EHR increased the likelihood that they would use it.

Time studies with clinicians and pharmacists also showed that it decreased the time needed to find information – in many cases to less than a minute.

The tool has not eliminated clinician-pharmacist interactions, but increasingly, those discussions are on more complex topics, such as medication impacts on specific patients or on administering complex order sets for critically ill patients, Cordrey said.

The organization has struggled with gaining overall physician acceptance of the EHR system because of limited training opportunities, Weisman said, noting that the organization is trying to demonstrate how this tool can save them time.

"Trying to add new functionality can be met with resistance. One strategy is having an optimization program in place, which will help overcome resistance to new tools."

Future opportunities for expanding and improving the CDS include making it work on mobile devices, improving the AI to pinpoint better responses for users and pushing out appropriate medication information to clinicians in their workflows, instead of requiring them to pose questions to seek answers.

Fred Bazzoli is a contributing writer to Healthcare IT News.
Twitter: @fbazzoli
Healthcare IT News is a HIMSS Media publication.