But the University of Pennsylvania sociologist specializes in analyzing interactions between medical computer systems and the people who use them, and he's found enough problems to turn him into an industry gadfly on the potential dangers of EHRs.
Just one example: clinical alerts, a feature that's supposed to improve care by automatically warning physicians about potential drug interactions, overdosing, allergies, or other hazards resulting from the orders they enter. "Eighty to 90 percent of alerts are overridden," says Koppel, who's also on the faculty of Penn's medical school. Experienced physicians often find alerts annoying and intrusive. Koppel works with residents, who commonly rotate among several hospitals and EHRs. Each system's alerts may be based on different criteria, or turned off entirely.
"A resident will get an alert at 50 [milligrams of a certain drug] at one hospital, 60 at a second hospital, and no alert at a third hospital because they turned it off," Koppel says. "So he thinks the 70 milligrams he's ordered there are safe. The residents don't know whether the alerts are on or off. They're not familiar with many medications and they start a new rotation every thirty days. They use these alerts as safety bumpers and that's not safe."
Koppel has found plenty of other glitches, from outright programming errors to user interfaces that make life difficult for clinicians. Numerical values appear in an order that makes sense to the computer but looks random to a human; positive test results aren't always flagged for review; weights aren't consistently labeled as pounds or kilograms (which can lead to babies, for example, being given twice, or half, the medication they need).
"Everyone focuses on why physicians are resistant to computers, but I would rather focus on how difficult the systems are to use," Koppel says. "Most physicians are the smartest guys in the room. Their resistance to technology as such is zero, but they resist software that has a clunky structure."
In the top 10
Every year the ECRI Institute, Plymouth Meeting, Pa., a not-for-profit organization that evaluates health technology, issues a top 10 list of technology hazards in medical care. "Problems with computerized equipment and systems" ranked seventh this year, right behind "needlesticks and other sharps injuries" and ahead of "surgical stapler hazards." Most of the incidents reported to ECRI by its 5,000 members (hospitals, health systems, payers, and other interested parties) were due to convergence of computers and medical devices in areas like medication management and the routing of device alarms to clinicians' cell phones and pagers. (ECRI Institute points out that such problems are "most certainly underreported.")
The current rush to get the entire medical community hooked into EHRs is based on the assumption that they'll make care better. Of course they can, by eliminating unreadable physician handwriting, speeding up communications, aggregating all patient data in one place, and providing reminders and alerts (preferably in a helpful, non-annoying way).
But EHRs can easily cause errors, too. Plenty of experts believe that too many systems are being installed too fast into environments too complex to be easily computerized. In the frenzy to be eligible for federal EHR meaningful use incentive payments, and avoid reimbursement penalties starting in 2015, institutions may be setting themselves up for disastrous computer-induced medical errors.
The U.S. Food and Drug Administration collects voluntary reports on safety incidents involving medical devices. In testimony prepared in February for a work group of the Office of the National Coordinator for Health Information Technology, FDA official Jeffrey Shuren, M.D., identified 260 problems over the past two years that were directly or indirectly related to EHRs. Of those problems, 44 caused injury to patients and six caused death. "They may represent only the tip of the iceberg in terms of the HIT-related problems that exist," said Shuren, who heads the FDA's Center for Devices and Radiological Health. And those incidents happened before the current EHR boom.
"I'm one of the biggest believers [in EHRs], but there's tremendous pressure to implement these systems so fast," says medical informaticist Dean Sittig, associate professor at the University of Texas Health Science Center at Houston and a leading researcher on successes and failures of EHR implementations. "It worries me that people won't have adequate time to come to grips with what they're doing and test their systems properly."
The medical environment is more complex than other fields like aircraft navigation, which is already hard enough to computerize, notes Nancy Leveson, professor of aeronautics and astronautics at the Massachusetts Institute of Technology. She's a pioneer in software safety and brings cross-industry experience to her role on the advisory board of the EHR Safety Institute, an initiative of Geisinger Health System, Danville, Pa.
"We're talking about a professional environment of doctors, and changing the way they do business," Leveson says. "Most other kinds of automation aren't doing that. Because software engineers aren't taught about usability and the impact of their systems on the world, they think they'll just automate it the way they want and make people do it their way. There's a lot of stuff out there [in health care] that's very difficult to use. The industry is naive about introducing software and the change it requires and the potential hazards it introduces, and they think it's going to be all right."
Types of errors
EHR-induced medical errors can occur for a variety of reasons: programming problems, incomplete interfaces between systems, preprogrammed order sets that haven't been thoroughly vetted, bad user interface design, wrong information incorporated into the software, or just the ease of entering data in the wrong spot.
One of Sittig's big worries is "wrong patient" errors, which can happen if physicians have more than one record open on the same screen. "We've solved the illegibility problem, but I can get a perfectly readable prescription that's not for a problem I have," because the physician has entered the data in the wrong window. Sittig advocates remedies like putting the patient's picture in the record, and allowing computers in patient rooms to show only the record associated with the patient registered to that room.