A patient-safety argument for EHR data mining

Did you know that combining a drug commonly prescribed to treat depression (Paxil) with another popular medication to lower cholesterol (Pravachol) could create a dangerous reaction by raising blood glucose to high or potentially harmful levels?

This startling news emerged several days ago in a study that appeared in the journal, Clinical Pharmacology and Therapeutics. The study, though, also provides insights of how electronic health records (EHRs) can be mined to detect crucial findings about medical treatments that may not even be on the healthcare community's radar screen.

Researchers of the study--which was conducted at Stanford University Hospital, Vanderbilt University Medical Center, and the hospitals of Boston-based Partners HealthCare--were surprised at the finding because neither of the drugs had a similar effect on blood glucose by themselves.

The study used the U.S. Food and Drug Administration's (FDA) adverse-event reporting database and the three participating institutions' EHRs. Data-mining techniques were used to detect patterns among large populations of patients that would not immediately be noticed by physicians treating individual patients.

The researchers estimated that as many as 715,000 people may be taking the two medications together.

These types of drug interactions may be occurring all of the time, but because they are not part of the approval process by the FDA, the healthcare community only learns about them after the drugs are on the market, Russ Altman, MD, PhD, a professor of bioengineering, genetics and medicine at Stanford and one of the study's senior authors, wrote.

"It's very exciting because we were led to this conclusion by mining data that already exists--but of which many people were skeptical," Altman said. "Physicians tend to think of [EHRs] as ways to better track data about single patients, but there's another really important component to them--their utility in looking at population effects."

The EHRs could make it possible to study a whole population of patients in real time, "in the wild, so to speak, as we are cared for, because that's the most realistic study sample," Isaac Kohane, MD, another study author and co-director of the Harvard Medical School Center for Biomedical Informatics, told the Boston Globe.

"You can come in with a question and literally, in weeks rather than years, answer important epidemiological questions," he added.

So is this the beginning of a new research trend--of creating answers for questions that have not yet been asked? Maybe. It could add a whole new dimension to patient safety and patient treatment issues. - Janice