Artificial Intelligence Can Improve Medication History
By Marcus A. Banks | September 25, 2024 | Article Link
At Carle Health in Illinois, pharmacists and pharmacy technicians spent a lot of time calling retail pharmacies to learn whether a patient had picked up a medication to compile an accurate medication list upon admission. The electronic health record (EHR) contained incomplete medication history data, with some prescriptions missing and vital instructions omitted about how the patient should take the medication.
“This time spent on the phone pulled staff from completing as many medication histories as possible and added time and work for the retail pharmacy staff, who were often very short-staffed during the pandemic,” said Andrea Schweska, PharmD, MBA, the inpatient pharmacy manager at Carle Health, in Champaign, Ill.
Carle Health decided to partner with healthcare technology vendor DrFirst, which imports more complete prescription data—including whether the patient has obtained the medication—into the EHR for improved efficiency and patient safety.
“Prior to partnering with DrFirst, patients often told our staff inquiring about their medication history, ‘If I have it, I’m taking it.’ With DrFirst integrated directly into our EHR, our staff can easily see if a patient has or has not picked up a medication from a retail pharmacy,” Dr. Schweska said.
A More Complete View
Colin Banas, MD, the chief medical officer at DrFirst, in Rockville, Md., noted that inefficient medication histories are not uncommon. “When medication history data is imported from the standard industry feed, the information often lacks detailed prescription instructions (known as sigs) and fails to capture all prescription fill data, especially from local or independent pharmacies,” Dr. Banas said. “We import more data and then use clinical-grade AI to pre-populate that data in the [EHR] system.”
According to internal DrFirst statistics, this has enabled Carle Health to import patient medication history for 93% of patients—efforts that led DrFirst to honor the health system with a 2024 Healthiverse Heroes Award, which the company sponsors.
Besides offering more comprehensive medication histories, DrFirst uses artificial intelligence to translate free text and fill in missing sig data, such as dose or frequency. When the AI is unable to accurately translate the data, it leaves the field blank and flags it for attention, such as in the case of a complex warfarin dosage regimen.
“You’re never locked into what the AI has suggested, and no prescription details are committed to the EHR until a pharmacist or other provider has reviewed them and signed off,” Dr. Banas said. “This is AI that keeps a human in the loop.”
Making the Business Case
“Before you invest in AI, ask yourself, ‘Why am I doing this?’ Identify a gap that the technology can solve,” said Ghalib Abbasi, PharmD, MS, MBA, the systems director of pharmacy informatics at the Houston Methodist Hospital system.
For example, Dr. Abbasi said, an AI tool such as DrFirst might improve the accuracy and speed of gathering patient medication histories. Faster availability of patient medication records should in turn improve patient care outcomes. “But if you’re already satisfied with your medication data acquisition practices, there’s likely less value in … [adopting] a new AI tool,” he noted.
Dr. Abbasi, who spoke at the 2024 ASHP Pharmacy Futures summit about innovative use of AI in pharmacy, said the ethical use of AI in healthcare is to free up clinicians to do higher-level work; the point is not for AI to replace employees.
“Never tell a senior leader that the goal is to replace someone,” Dr. Abbasi stressed, adding that once a position has been eliminated it may be hard to get it back.
Dr. Abbasi advised starting small with any new AI tool—for example, rolling it out to a selected number of pharmacists or units, rather than throughout an entire health system. “If it turns out that the human workflow required for an AI tool takes more time than it saves, or the expected AI value not realized, a health system can abandon it early on,” he said. “But if pilot users grow to like and value the tool and it demonstrates return on investment, it can then be deployed more widely.”