Using big data to prevent drug errors
Dr. Gidi Stein recalls how disturbed he was when he heard about a malpractice case involving the death of a pediatric patient that was the result of a medication error. The child’s doctor had accidentally selected the wrong medication from hundreds of drugs listed on the drop-down menu of a computerized order-entry form.
Stein, a professor of medicine and molecular imaging at Tel Aviv University in Israel, was dismayed that the tragic incident was described as an “unavoidable error.” With his background in software engineering and computational biology, he rejected that thinking as wrong-headed and dangerous. “Even if it’s a one-in-a-million chance, hundreds could die like that,” he said. “It’s completely unacceptable.”
In 2012, Stein co-founded an Israel-based company called MedAware. The firm offers a big-data software platform that integrates with a hospital’s electronic health-record system to detect prescription errors before they happen. It draws from patterns in millions of patient records to flag medication-order outliers.
If a physician chooses a drug that doesn’t match any condition in the patient’s record or diverges from how other patients with similar histories have been treated, the discrepancy is flagged. The system blocks the drug order until the doctor confirms its accuracy or cancels and re-enters the order.
MedAware is catching the attention of safety leaders working to reduce medication errors, who say they aren’t aware of other products taking this approach. “It’s a nice added feature to what’s out there,” said Matthew Grissinger, director of error reporting programs for the Institute for Safe Medication Practices, a Philadelphia-based not-for-profit focused on drug errors.
Medication errors are gaining more attention with the broader dissemination of EHRs and computerized prescription orders. A recent study in the journal BMJ Quality and Safety found that of more than 1 million medication errors reported to the U.S. Pharmacopeia MEDMARX reporting system between 2003 and 2010, more than 63,000 were related to computerized provider-order entry.
Many errors arose from user issues, including typing and pull-down menu errors, as well as ignoring or overriding alerts.
Typically, EHR alert systems are programmed to spot dangerous drug interactions, higher-than-normal doses and duplicate prescriptions. But these systems have not used large volumes of aggregated data to determine in real time the likelihood that the wrong drug was selected for a particular patient.
“This has a huge benefit,” Grissinger said. He noted that pharmacists as well as doctors can make mistakes. As the first few letters of a drug name are typed into an ordering system, a list of drugs that begin with the same letters appears. Selecting the wrong one happens frequently.
MedAware uses the same outlier approach that credit card companies use to detect fraud. Consumers typically exhibit predictable patterns such as shopping at particular stores and spending within a certain range. If a larger-than-usual charge suddenly appears in a different state or country, “it would be a complete outlier to your specific behavior,” Stein explained. “Credit card companies will call you up.”
Drawing from millions of patient records, MedAware uses a mathematical model to predict the likelihood of particular types of patients being prescribed specific drugs. If the software detects an outlier when a clinician enters a prescription order—such as prescribing a drug for an infant that generally is used for senior citizens, for example—an alert message immediately appears on the order screen.
Although the software is not currently used in hospitals, preliminary findings on MedAware’s effectiveness were presented at the Healthcare Information and Management Systems Society’s annual meeting in April. Retrospective analyses of more than 44 million filled prescriptions from two large hospitals and one HMO in Israel generated alerts for more than 7,000 patients, with a low rate of false alarms.
Brigham and Women’s Hospital in Boston is testing MedAware’s effectiveness in flagging errors through a retrospective look at filled prescriptions in more than 748,000 patient records, representing more than 9 million prescriptions. Findings are expected by the end of the summer.
Dr. David Bates, the hospital’s chief innovation officer, said he’s excited about MedAware’s potential. “It’s the kind of approach that can get you to a higher level of safety,” he said.
Grissinger said MedAware holds great promise when used in conjunction with established drug-interaction, dosing and duplication alerts. But providers still need to have other safety precautions in place, such as evidencebased clinical protocols and staff training in fully and accurately entering patient information into EHRs, he added.
“You can have all the data mining in the world,” Bates said. “But if the needed fields in the EHR are empty, you’re going to get nothing out of it.”