Pharmacists’ Skills with AI Tools
Developing an accurate and standardized master formulation record (MFR) for compounded medications is a critical skill for all PharmD students to master. But Elizabeth Hageman, PharmD, was frustrated with limitations on her ability to provide students with the opportunity to practice creating MFRs multiple times, receive immediate feedback and go back to build on what they’ve learned. Pharmacists’ Skills with AI Tools.
“To do that with a faculty member would require hours and hours of time that we don’t have,” said Dr. Hageman, a clinical assistant professor of pharmacy practice and an assistant director of skills education at Binghamton University School of Pharmacy and Pharmaceutical Sciences, in New York. “I wondered if it would be possible for an AI program to be trained to grade and review MFRs.”
It was the perfect project for Dayanjan (Shanaka) Wijesinghe, PhD, an associate professor in the Department of Pharmacotherapy and Outcomes Sciences at Virginia Commonwealth University School of Pharmacy, in Richmond, and the students in his five-week digital health Advanced Pharmacy Practice Elective (APPE) rotation. An expert in precision approaches to health with a focus on digital technologies, Dr. Wijesinghe trains pharmacy students in both the development and real-world implementation of digital health applications.
Under Dr. Wijesinghe’s guidance, pharmacy students Dorjan Leka and Talal AlMuzaini collaborated with Dr. Hageman to use OpenAI’s GPT-4 foundational model in building a customizable artificial intelligence tool. This tool assesses students’ ability to create MFRs according to USP General Chapter <797>. It comprises a 10-section grading rubric, including sections such as ingredients and calculations, compounding steps, and beyond-use dates and storage requirements, with the AI “teaching assistant” programmed with detailed instructions as to how to add and remove points from each section.
The tool was completed in April, and Dr. Hageman plans to roll it out to one of her student cohorts in the coming academic year. It’s just one of the latest pharmacy education AI tools that Dr. Wijesinghe has created with his digital health APPE students over the past few years. “We observed that pharmacy students were not getting enough experiential education opportunities before moving on to their P4 rotations,” he explained. “Our efforts to address this gap led us to develop a series of AI tools to help students hone their clinical skills.”
Honing Pharmacy Counseling Skills
One of the first applications they developed, released in June 2023, was a conversational AI that allows students to practice their pharmacy counseling skills. Dr. Wijesinghe said the AI emulates a patient with a distinct and detailed persona and health issues so convincingly that, without prior knowledge of it being a chatbot, “students would think it was a real person.” These interactions “allow the student to start pulling information from all these different buckets of knowledge they have learned and integrate it into patient counseling. That’s a skill set that requires practice.”
The AI has been taught to exhibit characteristics of real patients, such as confusion if the student is unclear or complex in their explanations, or misconceptions about the medication they’ve been taking. After the conversation ends, the AI becomes the evaluator and tells the student how well they did on various skills, such as empathy, quality of the questions they asked and the responses they gave, and use of jargon. After getting that feedback, the student also can ask the AI to offer more detailed information about how they could have performed better.
“We don’t have time to provide students unlimited opportunities to practice these skills in a classroom setting,” Dr. Wijesinghe said. “Something like this can be done independently of classroom learning and allow students to practice on their own time as much as they want in order to improve.”
Dr. Wijesinghe and his students have developed and deployed a range of AI tools for pharmacy education, including applications for SOAP (subjective, objective, assessment and plan) note practice, pharmacogenetic counseling, vaccine recommendations, medication reconciliation and geriatrics pharmacy counseling. They also have created an innovative Socratic tutoring tool for case-based learning, which presents students with patient scenarios and prompts them to propose diagnoses and appropriate tests.
“The AI doesn’t provide answers outright; instead, it encourages students to think critically and formulate their own responses based on given information and test results,” Dr. Wijesinghe explained. “Each conversation is unique, ensuring varied learning experiences.”
Like many of their projects, it was developed collaboratively, in partnership with Ashim Malhotra, BPharm, PhD, a professor of pharmacology and the assistant dean of academic affairs and accreditation at California Northstate University College of Pharmacy, in Elk Grove. The success of these initial tools laid the groundwork for their latest project, the MFR assessment tool developed with Dr. Hageman.
AI Tutors Graded on a Scale
Dr. Hageman cautioned that there are definitely limitations to AI pharmacy tutors. In pharmacy compounding, for example, “there are a lot of nuances” that AI may miss, she said. When applying Chapter <797> in practice, “there are some gray areas that exist in terms of the data that would be acceptable, such as adding additional ‘miscellaneous’ information to … ensuring [a compound] is dispensed with a filter. Other parts, however, are easy and straightforward, such as the name of a compound, finding the concentration of the base drugs and the actual steps for the compound.”
Dr. Hageman also noted that students and faculty should not become overreliant on AI. “Drug information is changing all the time, and you might not know the lapse in time from when new information is released or data are published to when the AI is able to incorporate it into the algorithm,” she said. “It’s an incredible tool, but we need to understand that it doesn’t eliminate clinical decision making or the need for faculty to teach. It’s our responsibility to use these tools in a way that complements our teaching.”
‘Future-Proof’ Design
The AI tools developed by Dr. Wijesinghe and his team are designed to be adaptable and “future-proof,” meaning that they can operate on various platforms and incorporate different AI models, including open-source options, ensuring flexibility as the field rapidly evolves. This approach allows educational institutions to easily update or switch AI models without overhauling their entire system.
“The worst thing we could do with these AI educational tools is lock into a single vendor, because there’s no clear winner here yet, so our programs can easily switch between different AI models on the back end,” Dr. Wijesinghe explained. This flexibility not only keeps the tools current but also reduces costs and improves performance over time as more advanced AI models become available, he noted.
The AI tools are available to other pharmacy schools and students, free, on a web-based platform called PharmtutorAI.com, with more to be released soon. They also can operate in multiple languages, including Spanish, Arabic and Mandarin. “Anyone anywhere in the world can click on one of these and get access to these applications and practice their skills,” he said. “Because we are not making money on this, we are doing it as a service to improve pharmacy education globally.”
Source: Pharmacists’ Skills with AI Tools
https://www.techedmagazine.com/category/news-by-industry/health-sciences/
https://www.pharmacypracticenews.com/Pharmacy-Technology-Report/Article/09-24/AI-Tools-Pharmacy-Education/74775