Artificial intelligence has been an integral part of pharmacy for decades. It is called the pharmacy management system: the collection of digital tools that houses patient utilization and drug data in virtually all pharmacy settings.
Wayne Salverda, RPh, senior director of clinical services for National CooperativeRx, remembers visiting Eli Lilly and Company while in pharmacy school. “They had a crazy supercomputer working on drug design and that was 30 years ago,” he said. “About 1 in 10 drugs in development makes it to market, but companies like Vertex or Regeneron that have invested in and leveraged AI see 30% to 40% of their new molecular entities make it to market. It’s not a coincidence; it’s putting smart science and AI into drug development.”
AI can be just as useful in medication management. Pharmacy benefit manager Prime Therapeutics worked with AI provider SAS to save clients $355 million over 18 months by consolidating and analyzing data from pharmacy, drug claims, and medical services to target fraud.
“We have gotten some good recoveries and great cost avoidances with AI,” Salverda said. “We will be able to leverage AI even more in the future to help make sure that individual patients are getting the most appropriate medications at the most appropriate dose by factoring in their specific genetics and health conditions. We’re going to see a lot of advances in pharmacogenomics with AI.”
Beyond the Human Brain
AI is a concept, not a gadget like the robot Maria in the 1927 film Metropolis or R2-D2 in Star Wars 50 years later. “AI is nothing more than a means to an end, improving processes and improving outcomes,” said Codrin Arsene, chief marketing officer for Digital Authority Partners. “It’s just an automation of data, rules, and statistical processes that help people do their jobs more efficiently.”
When the American Society of Health-System Pharmacists (ASHP) and the American Medical Association discuss AI, they mean augmented intelligence: leveraging the strengths of computers and clinicians to improve outcomes for patients. “AI today is everywhere in pharmacy: drug discovery, drug design, drug development, trial design, participant selection, clinical decision support, drug response and risk prediction, personalized therapy, everywhere,” said Wei-Hsuan Lo-Ciganic, PhD, MS, MSPharm, associate professor in the Department of Pharmaceutical Outcomes and Policy at the University of Florida College of Pharmacy.
AI has come to pharmacy in 3 waves, noted Chenglong Li, PhD, a professor in the Medicinal Chemistry, Biochemistry, and Biophysics Department at the University of Florida College of Pharmacy (see Sidebar). Pharmacy management systems, IVR, and DUR screens are first-wave products. Second-wave AI powers some of the more advanced decision-support tools and systems that are starting to track fraud. Third-wave applications will change pharmacy in ways that are difficult to imagine, Li explained.
“AI is going to have the kind of impact we see with digital cameras,” Lo-Ciganic said. “When I was in school, people laughed at digital cameras. But when the iPhone came out, with its smart camera, the world changed. We are going to see the same kind of change in the world of pharmacy.” Lo-Ciganic is using AI to improve risk prediction for opioid misuse by linking Medicaid, human services, and criminal justice data. AI helped her identify 30 key predictors of opioid overdose.
“The methods used today can predict about 30% of opioid overdoses,” she said. “When we use our machine learning rule, we can improve to 70% to 90% accuracy for individuals who have an opioid overdose or misuse. That helps focus time and resources on the people at highest risk.”
Li’s lab modeled an enzyme that represses tumor suppressor genes and a native compound that blocks the enzyme, which allows suppressor genes to be expressed and kill the tumor. “We have a few collections of really potent compounds, maybe 10,000-fold more potent than the native compounds that work in cell culture,” Li said. “We are trying them in animal models to see if they are active in vivo.”
More changes in drug discovery are coming. In July 2021, DeepMind announced the latest version of its AlphaFold platform has identified the 3D structure of 98% of the entire human protein complement. More than 350,000 protein structures are available in a public database hosted by the European Molecular Biology Laboratory’s European Bioinformatics Institute.
AI isn’t making any of these advances on its own. AI needs human intervention as much as humans need AI.
“AI has the potential to help pharmacists improve their workflows.,” said Sophia Chhay, PharmD, assistant director of the ASHP Innovation Center. “Any pharmacist will tell you that ‘anything that gets me away from the computer and spending more time with patients is going to be an answer.”
Freeing up pharmacist time is key, said Ken Perez, vice president of healthcare policy and government a¡airs at Omnicell. The typical pharmacist spends 75% of working hours on nonclinical tasks.
“If you think about every medication as a node on a network, every drug has a journey, and where every dose goes is important,” he said. “Pharmacists spend too much time as de facto supply chain specialists, figuring out where drugs are [and] how to get them to the right patient at the right time. That time would be better spent with patients, not products. And that’s where AI can make a huge difference. The real prize of AI is helping pharmacists [to] practice at the top of their licenses, spending more time with the patients who need them the most.”
AI may shift pharmacist roles by giving them more hands-on time with patients, but don’t expect massive layoffs. “I don’t think any state board is at all comfortable with turning over critical decision-making to automation or AI,” said Lisa Schwartz, PharmD, senior director for professional a¡airs at the NCPA. “AI might make a recommendation, maybe a medication adjustment to avoid an interaction or adjusting the time of day a patient takes something, but it’s up to the pharmacist to decide what needs to be done.”
US clinicians, including pharmacists, are slow to adopt AI, Arsene noted. About 25% of European Union physicians are using voice assistants in daily practice. “For all the research that goes on, this is an industry that has a very low level of innovation and too much bureaucracy to change quickly,” he said. Arsene likened pharmacy to the financial services industry. It wasn’t until Credit Karma, Mint.intuit.com, and other new entrants began to affect customer preferences and banking profits that traditional providers moved to AI.
New entrants in pharmacy are presenting similar challenges. Kit Check uses AI to audit 100% of hospital-controlled substance use in less time than a conventional 5% random audit.
“Nobody can track controlled substances manually; there is just too much information,” said Doug Zurawski, PharmD, Kit Check’s senior vice president of clinical strategy. “We and our competitors can give you a dynamic picture of everything and everyone in your system relative to controlled substances.”
AI systems developed for the relatively closed universe of health systems can be expanded to ambulatory care and retail pharmacy, he added. On the dispensing side, InterLink-AI is expanding Rx verification by identifying every pill as it goes into a patient vial. Different versions of image-based verification can be adapted for a single pharmacy, telepharmacy, load sharing across multiple pharmacies.
“You have a clean record of the chain of custody from the raw materials used in a batch of medication through distribution of the stock bottle,” said Ram Subramanian, PhD, InterLink’s chief business officer. “But from the stock bottle on, there are questions. Verifying 100% of the pills that go into every vial optimizes the pharmacy workflow and frees up time for technicians and pharmacists. Improving the workflow puts the pharmacist in the driver’s seat regarding patient care.”