Hospital pharmacists tap AI to improve drug information searches
By Watson Health | 4 minute read | October 18, 2018
Roughly 73 percent of hospital outpatient visits involve drug therapy. In the emergency department (ED), that number jumps to 79 percent, which results in a combined total of 732 million drugs dispensed in hospital outpatient and ED settings alone each year1. In each situation, hospital pharmacists must weigh dosing guidelines, administration preferences, possible contraindications, and a host of other factors before dispensing a single dose.
Add the threat of adverse drug events and you begin to understand the pressure on hospital pharmacists to make the right decisions. Plus, pharmacists must consider the budget implications of escalating drug costs, which have been increasing at a rate of 23 percent per year2.
For Dennis Killian, Director of Pharmacy at Peninsula Regional Medical Center in Salisbury, MD, that ever-present need to process huge amounts of information and deliver fast, accurate drug information to frontline clinicians has given him strong drug database research skills.
“I’ve been doing this for many years, so I have built up a lot of muscle memory for accessing drug information very quickly,” Killian explained. “As pharmacists, we get called upon for any type of drug information and time is always a factor. Frontline clinicians are looking for things like dosing information for different medications, IV compatibility, potential drug interactions, and my team needs to be able to deliver answers quickly, but we also need to be certain that they are accurate for each patient.”
Improving the pharmacist workflow
A typical example of Killian’s drug search process starts when a clinician calls to ask about a dosage adjustment for a particular medicine based on that patient’s kidney function. He’ll access his drug database, enter the specific drug, then toggle through options to identify dosage guidelines based on patient age, weight and other factors to provide evidence-based recommendations.
Earlier this year, that workflow got a little easier. Killian and his pharmacy team at Peninsula Regional were part of a limited availability roll-out of a new, first-of-its-kind solution that incorporates artificial intelligence (AI) into their drug information database. This new layer of AI-powered search technology allowed Killian and the team to bypass the old keyword-based research process in favor of a new search box that accepts natural language queries.
Killian offered some examples: “With the new technology, we could just type in questions just the way a clinician would ask them, like: What’s the adult dose of rivaroxaban for DVT? Are there renal dosing adjustments? Can I give with food? What should I monitor? The answers appear instantly, without our team needing to navigate and toggle through menus.”
This new way of accessing information has made it easier for Killian’s team to juggle the rapid-fire requests that often come into the pharmacy. Rachel Cordrey, Pharmacy Supervisor at Peninsula Regional explained: “When you’re searching keywords, you’re thinking in a different mind frame. You need to break your train of thought and switch your brain into research mode. Now, we’re able to stay on task and just ask a question as we would when speaking to fellow clinicians. The technology fills the gaps and we can continue on our day.”
“The system gets us to the answers we need faster,” Killian added. “We really see that advantage in people who are less experienced with legacy drug database searches and may not have that muscle memory that lets them toggle through the system with their eyes closed. It makes the learning curve much shorter.”
While the efficiency gains resulting from the new technology are hard to ignore, the other critical variable is accuracy. Pharmacists need to be able to trust the results.
Among the dozens of hospital pharmacists who have been testing this new technology in a limited availability roll-out, all have made it clear that their number one priority is to make sure the information they are delivering to their colleagues and friends is rock solid.
“I don’t want to put anyone in harm’s way and I don’t want to let anyone down,” said one pharmacist, echoing the sentiments of many of his colleagues. “So, while it’s important that we get on board with technology and keep innovating to improve the care we can provide, I also need to be able to know that the information I’m getting is legit.”
First Steps to the Future
These first steps toward integrating AI technology into hospital pharmacist tool kits are just the beginning. Peninsula’s Killian pointed to the huge potential that exists once the technology can be integrated into the hospital’s EHR platform, ultimately linking patient-specific data to pharmacy information. He also highlighted the system’s ability to learn.
“The speed at which this technology has advanced is amazing. Over the course of our first 15 weeks testing the technology, we saw its level of understanding advance considerably. Once that capacity to learn and grow is applied across other clinical disciplines and more and more people get comfortable using the technology, the potential is virtually limitless.”
The new AI solution has been tested by more than 50 clinicians in six countries over the last several months. General availability for clinicians around the world began on August 29.