Completing the Medication History Picture With Artificial Intelligence
As the COVID-19 pandemic spread across the U.S., including into Michigan, it flooded Detroit-area hospitals with critically ill patients in the Great Lakes Bay Region. At the Michigan-based Covenant HealthCare, organizational leaders took precautions—aided by artificial intelligence (AI)-driven technology—designed to protect staff and provide quality care for patients in its emergency care unit.
At Covenant Medical Center in Saginaw, Mich. — part of the broader Covenant HealthCare system — one of the new protocols, decided upon a few months ago, was to completely automate the medication reconciliation process for ER patients. Medication reconciliation is recommended every time a patient is admitted, transferred, or discharged from a healthcare facility, and accurate medication reconciliation is a national patient safety goal of The Joint Commission.
The process at Covenant has traditionally involved a pharmacy technician interviewing patients face-to-face for information about their medication history, as knowing this information helps prevent treatment errors. One pharmacy technician may interact with tens of patients, depending on how busy the ER is.
Of course, once COVID-19 hit, without knowing which patients presenting to the ER were positive, clinicians ultimately deemed the risk to pharmacy technicians too high to continue the face-to-face practice, especially since they weren’t necessarily familiar with the personal protective equipment (PPE) critical to protecting them and other frontline healthcare workers from exposure, says Becky Sulfridge, Pharm.D., clinical pharmacist specialist, emergency medicine, at Covenant HealthCare. She notes that technicians having to change their PPE to see each of the 65 to 70 patients who were coming into the ER per day was never a realistic option.
Instead, ER staff at the patient care organization more heavily relied upon a technology solution from DrFirst, called MedHx, which is integrated into its Epic electronic health record (EHR) platform to gather medication history data for ER patients—many of whom have been coming in sicker than usual, making it difficult for staff to gather information such as medication history.
Sulfridge says that when the pandemic first hit the region, Covenant’s clinical and pharmacy leaders were faced with a dilemma: they could either stop offering the medication review service or they could modify their workflow. “The first option was not viable for us, so we sat down and discussed how we could try to keep people as safe as possible by continuing to offer that service on the same level and at the same standard our co-workers were used to,” she recounts. The decision was then made to pull the pharmacy technicians out of the room and eliminate the face-to-face component of the workflow, instead relying on the DrFirst technology.
“In the past, we’d interview the patient [about his or her medication history], and then use DrFirst to fill in the gap. All of a sudden now, we’re using DrFirst as our main source of information and then using phone calls to fill in the gaps from there,” Sulfridge explains.
AI completes the picture
Within the MedHx platform is technology called SmartSig technology, which uses patented AI designed to improve the quality of patient medication history when it is imported into hospitals’ EHR systems. According to research estimates, some 66 percent of data is missing essential prescription instructions, or sigs, increasing the risk of adverse drug events that compromise patient safety. Sulfridge further notes that the SmartSig technology helps to fill in information gaps about how the drugs should be taken, providing a more complete medication history. Indeed, as pointed out by DrFirst clinical leaders, when medications are imported into EHRs, sigs associated with them typically arrive as unstructured free text, often with missing pieces of information and using a variety of terms for the same instructions (e.g., “by mouth” vs. “orally”), making the process of entering the medications labor-intensive while increasing the risk of adverse drug events.
“It’s one thing to have the person’s medication list, but there is a lot more to [the medication picture], such as how many times per day the [drug] is being taken and what the exact dosage is,” says Sulfridge. “Is the patient splitting the dosage in half or take two at once? Sometimes patients will know what they took, but we still have to call the pharmacy to get [complete] information on what the dose is and how often they took it.”
SmartSig AI aims to address these challenges by producing accurate, structured, real-time translations. The technology converts free text elements of medication sigs into a health system’s standard terminology and processes the data into appropriate fields so that it becomes functional within the EHR, according to company officials, who note that SmartSig 2.0 accurately translates nearly 93 percent of incoming prescription information, helping avoid medication errors and saving up to 30 seconds of work for each drug entered during the medication reconciliation process.
“We were early beta testers of SmartSig 2.0 and we’re now live with it. With this tool, we saw a 14 percent gain in our throughput, and our sig accuracy is at 93 percent,” says Sulfridge. I have been doing this [work] going on six years, and if you would have told me six years ago we would have 93 percent accuracy from a database for home medications, I would have probably looked at you like you were a little bit nuts,” she admits.
What’s more, with the previous medication reconciliation approach, Sulfridge recalls, technicians were spending between 25 to 30 face-to-face minutes per patient, which was necessary at that time in order to get all the pertinent medication history details. Since switching to DrFirst for automated medication reconciliation, however, the staff at Covenant has seen “significant time savings.” Sulfridge reports that “We were able to increase the number of patients we saw in a day by decreasing the amount of time spent with each [patient]. That was a big benefit for us,” she says, adding that “Instead of now relying on the face-to-face interview with technology to fill in the gaps, we will rely on technology to get our [medication] lists and then the interview to fill in the gaps.”
