Study shows reduction in medication errors using health IT startup’s software
A study conducted by Harvard researchers showed an Israeli startup’s software showed promise in preventing incorrect prescriptions. The study, published in the Joint Commission Journal on Quality and Patient Safety, estimated 68.2 percent of errors generated by MedAware’s platform would not have been detected by existing systems.
The retrospective study looked at records from 747,985 patient encounters at Brigham and Women’s Hospitals’ outpatient clinics between 2012 and 2013. From that, MedAware’s system flagged 10,668 potential errors, 300 of which researchers sampled for further review. According to the study, 92 percent of warnings generated by MedAware’s system were accurate based on the available data, and 79.7 percent were clinically valuable. The study estimated Brigham and Women’s would have saved $1.3 million in healthcare-related costs, extrapolating on the results from this sample.
Dr. Gidi Stein, MedAware’s CEO and co-founder, said the study not only further validated MedAware’s system for patient safety, but “also proves our considerable benefit to a hospital’s bottom line as well.”
Most healthcare systems currently use rule-based decision-support tools to catch medication errors. They’re helpful, but only in catching problems programmed into their alerting logic.
MedAware’s platform, on the other hand, uses machine learning to detect outliers, flagging them as potential errors. The company’s system says it can detect instances of the wrong drug being prescribed, or a prescription being assigned to the wrong patient. It can also raise a warning for time-dependent events, such as a change in a patient’s vitals or lab work that could make a drug riskier.
The study was led by Ronen Rozenblum, an assistant professor at Harvard Medical School and director of business development for patient safety research and practice at Brigham and Women’s Hospital. MedAware was not involved in the study’s design or funding.
“Everyone is using the buzzword of machine learning. I was curious if MedAware is doing better,” Rozenblum said.
He led a previous study of MedAware’s software in 2017 that showed 75 percent of the company’s alerts were clinically valuable. Since then, Rozenblum said, the algorithm improved. Most of its alerts were classified as “high value,” meaning a physician might want to use them to make a change or decision about the patient’s treatment.
“With high value (alerts), we see basically the patient was given the wrong dosage or wrong medication. It sounds terrible but it still happens,” Rozenblum said. “These systems might save lives.”
Of course, a retrospective study has its limits. Since these drugs were prescribed in the past, researchers weren’t able to interview physicians, patients and other stakeholders.
In the future, Rozenblum said he’d like to do a prospective study, though that process is admittedly much more complex. He also hopes to test MedAware’s system on inpatient data.
“I believe that the findings we found were very significant, but conservative,” he said. “We know the value of the system is even higher when the patient is hospitalized. This is when the implications of medication error are even higher.”
Photo credit: Stas_V, Getty Images