Q&A: How AI is Reshaping Pharmacy and Personalized Medicine | APhA 2025
Technological evolution promises to enhance patient care, reduce errors, and provide intelligent support that augments human expertise rather than replacing it.
Artificial intelligence (AI) is rapidly transforming pharmacy practice, offering unprecedented capabilities in predictive modeling, clinical documentation, and personalized medicine. By leveraging advanced technologies like machine learning, natural language processing, and deep learning, pharmacists can now anticipate medication outcomes, generate clinical notes, and reconcile complex patient data with genomic knowledge more efficiently than ever before. This technological evolution promises to enhance patient care, reduce errors, and provide intelligent support that augments human expertise rather than replacing it, according to Ravi Patel, PharmD, MBA, MS, lead innovation advisor in the School of Pharmacy at the University of Pittsburgh.
However, as AI becomes increasingly integrated into medical practice, pharmacists are emerging as crucial navigators of this technological frontier, uniquely positioned to address the ethical, practical, and human-centered considerations that accompany digital innovation. Their role extends beyond technical implementation, encompassing critical responsibilities such as ensuring data integrity, mitigating potential biases, protecting patient privacy, and maintaining the fundamental human elements of care that no algorithm can fully replicate. By asking probing questions about data ownership, patient consent, and technological limitations, pharmacists can help shape AI's development into a collaborative tool that enhances rather than replaces human expertise.
Technological evolution promises to enhance patient care, reduce errors, and provide intelligent support that augments human expertise rather than replacing it. | Image Credit: Tom | stock.adobe.com
Drug Topics: What are the key differences between machine learning, deep learning, and natural language processing in the context of pharmacy practice?
Ravi Patel, PharmD, MBA, MS: It's interesting now because when we think of artificial intelligence, we think a lot about generative artificial intelligence, and generative artificial intelligence is actually made up of a lot of work, over 50 or more years of computer science and evolution. So, when we break down some of the components, concepts like machine learning are really important, but really widespread. When, as pharmacists, we think about concepts like logistic regression, we're doing forms of machine learning, but when we start layering on some interesting advanced computing, we have the chance to use concepts like natural language processing to take text or spoken word and break down sentiment and tension of definitions or to even think about whether or not there's a concept being addressed or framed. When we add deep learning, being able to use neural networks, which, again, as a concept, have been around since the early 1900s, being able to frame what we might do with generative learning is possible because of those advanced neural networks in deep learning.
Drug Topics: How can these be used in pharmacy, and what are the benefits?
Patel: In pharmacy, when we break down different tasks, being able to apply concepts like logistic regression when we're trying to make a predictive model of categorical variables. As an example, if I know a class of medication, we can use concepts like machine learning to make a prediction. If I know yes or no, which medications am I on, we can build that into creating a model that says yes or no; am I going to develop a disease state or a chronic condition? As we think about more advanced applications, natural language processing, when I am asking a question of my voice assistant on my cell phone or my smart speaker at home, we're using natural language processing. As a project, we once tried to compare what can I ask my smart speaker to ask counseling questions, "when should I take this medication?" against what a human might say to some really interesting results that have actually changed over time. When we use deep learning, we can think of creating images like we've seen with generative AI or being able to write an example SOAP note when we look to create clinical documentation with, of course, some clinical check and verification.
Drug Topics: How is AI being used to analyze patient medical records and improve medication reconciliation?
Patel: Artificial intelligence might already exist in ways that we engage with but don't know about when we think about our electronic health care records. The systems that allow us to document and store this data already might be running artificial intelligence. One way that we experience it as consumers is to use ways of auto-complete for searching on engines like Google. When we can think about how we use that in clinical documentation, being able to start something like a SOAP note and automatically pull information from a chart are examples that generative AI are already being used.
Drug Topics: Can AI be used to optimize medication adherence and personalize dosing regimens? If so, how?
Patel: It's really exciting to see the advancement of computing, especially when it comes to personalized medicine. One way that we've been using these advanced methods for personalized medication has come in mapping the human genome. When we think about pharmacy practice today, we can think about being able to take already existing knowledge about genomics and comparing a dynamic patient chart, one that changes over time, to our dynamic knowledge of personalized medicine and genomics. So, being able to reconcile information in ways that a single human mind might take a long period of time or might not even be able to recognize is one way that artificial intelligence can factor into the application of personalized medication.
Drug Topics: Can AI be used in compounding to ensure accuracy? If so, how?
Patel: When we look at some of our basics in compounding, there's an immense amount of variation, both in task, identifying quality, and thinking about if we've completed each of these steps one way that we've experimented in applications of artificial intelligence, having been including computer vision mapping what a hand or emotion for a pharmacist might have used in order to get a specific ingredient or in applying the right medication in the right order. Granted, we do this automatically with our paper and pencil protocols, but with layering on artificial intelligence, we're able to think about using a computer as a double check, a way to say, "Can I remember if I did this with the help of a computer," rather than by the replacement of a computer?
Drug Topics: What are the potential challenges of integrating AI into pharmacy workflows?
Patel: Any technology can often be challenged by the thing that makes us the true clinicians we are: being human. Artificial intelligence can face challenges in implementation and workflow from a technical point. Do we have access to the data to a human-based point? Do I trust the data that's being generated? When we think about how a pharmacist can be a part of addressing these barriers in applying artificial intelligence in workflow by working with the end users, whether that's a pharmacist, a patient, a caregiver, to ask questions like, "What problems are we trying to solve, and how might we use technology to solve those human problems?" We can think about barriers of workflow trust or even access to data.
Drug Topics: What are some of the ethical considerations related to the use of AI in pharmacy practice?
Patel: Using data often brings ethical considerations. With artificial intelligence, a common question of any use of data is, who owns this data? For example, in current research, medication records or health records already exist in a commercial format. So when I get something filled or if I visit a hospital, my records become part of research data sets already; we face ethical questions of what happens, of my ability to say no thank you or opting out of that process. We face challenges, even outside of artificial intelligence and access to data. When we think about applying these unique methods to generate new data, we might have to ask the question, "Who owns the original data, my records, my medication information, and who owns the new data, what was generated as a result of my information?" So when we think about new opportunities or new roles for pharmacists, being able to think ahead to what those questions or challenges and ethics might be can be a really great role for pharmacists to be a proactive part of implementing artificial intelligence.
Drug Topics: How can we ensure that AI tools are used equitably and do not exacerbate existing health disparities?
Patel: Any technology can carry a risk, increasing a digital divide. How do we think about assessing both clinician and patient understanding of the technologies that inform how they practice or the care they receive? One challenge, especially with artificial intelligence, is the risk of biased data. If I only use data from a specific population and try to apply that to another population that might not be represented, this might exacerbate some of those limitations or biases. One of the interesting components of artificial intelligence implementation, or of any technology, is it might ask us, "How can we as humans be better at recognizing our differences?" So while we might face challenges of identifying compounding effects of bias or in increasing that digital divide, we can use this opportunity with the rapid iteration of artificial intelligence to ask, "What did our last version miss, or how, when we make a new technology implementation plan, can we better address the risk of bias?"
Drug Topics: Is there anything else you would like to add?
Patel: With technology, we often talk about what changes as being an important part of what we look at. When we think about technology, I think we can look at what stays the same when we think about those core competencies, the key things that we work on, or the same data that we use. We have an opportunity to make a new technology more than just a novelty or a scary idea, and make it a true collaborative approach between the humans and the technology that can improve their work.
