Some people receive constant reminders on their personal smartphones: birthdays, anniversaries, doctor’s appointments, social engagements. At work, their computers prompt them to meet deadlines, attend meetings and have lunch with the boss. Prodding here and pinging there, these pop-up interruptions can turn into noise to be ignored instead of helpful nudges.
Something similar is happening to doctors, nurses and pharmacists. And when they’re hit with too much information, the result can be a health hazard. The electronic patient records that the federal government has been pushing — in an effort to coordinate health care and reduce mistakes — come with a host of bells and whistles that may be doing the opposite in some cases.
What’s the problem? It’s called alert fatigue.
Electronic health records increasingly include automated alert systems pegged to patients’ health information. One alert might signal that a drug being prescribed could interact badly with other medications. Another might advise the pharmacist about a patient’s drug allergy. But they could also simply note each time that a patient is prescribed painkillers — useful to detect addiction but irrelevant if, say, someone had a major surgery and is expected to need such meds. Or they may highlight a potential health consequence relevant to an elderly woman, although the patient at hand is a 20-something man.
The number of these pop-up messages has become unmanageable, doctors and IT experts say, reflecting what many experts call excessive caution, and now they are overwhelming practitioners.
Clinicians ignore safety notifications between 49 percent and 96 percent of the time, said Shobha Phansalkar, an assistant professor of medicine at Harvard Medical School.
“When providers are bombarded with warnings, they will predictably miss important things,” said David Bates, senior vice president at Brigham and Women’s Hospital in Boston.
Now, doctors, health information technologists and software vendors are trying to fix the problem.
Research on this human-computer interaction is starting to explore the degree of risk posed by excessive alerting versus the benefits the alerts produce. The companies selling electronic health records say advances are moving their systems toward more targeted, relevant warnings, instead of broad-brush signaling.
“This is an issue that everyone’s going to have to wrestle with eventually,” said Bill Marella, executive director of patient safety operations and analytics at ECRI Institute, a nonprofit organization that studies health care safety and quality issues. In April, the institute ranked design and implementation of new health IT systems as its top safety concern for 2016.
Some hospitals and health systems are already paving the way.
Take Children’s Hospital of Philadelphia. In 2012, the inpatient facility switched over to a new electronic health record, said Eric Shelov, a physician and the hospital’s associate chief medical information officer. Immediately, he said, practitioners began seeing far more alerts, to the point that doctors were overriding almost all of them. The problem, Shelov said, is that “if you see enough nonsense, you’re going to start ignoring it.”
That has consequences. In one instance at Children’s, doctors ignored relevant information about how a patient might respond to a drug, Shelov said, because it appeared alongside heaps of other superfluous notifications — warnings, for instance, about drugs that posed minimal risk of interfering with each other. Consequently, the patient received medication that induced a potentially lethal reaction.
The hospital caught the mistake in time, but the incident spurred a series of changes. A team of pharmacists, doctors and other clinicians have sorted through what triggered alerts in their system, turning off the ones they decided weren’t actually relevant or necessary. That has helped. But it’s still an ongoing battle, Shelov said. “It’s a little bit of trying to turn off the firehose.”
Systems such as Cleveland-based MetroHealth, the University of Vermont Medical Center and Group Health Cooperative of South Central Wisconsin have undertaken similar projects. Still others, like Brigham and Women’s, are working on it.
But figuring out what merits a computer warning takes time, manpower, expertise and money. Not all hospitals have those resources, Bates said. It’s inherently subjective. Some stakeholder groups have put out recommendations, and hospitals like Children’s have presented on ways to combat alert fatigue. But individual hospital task forces often end up deciding for themselves what’s risky enough to warrant an alert.
Patients, meanwhile, aren’t standing beside their doctors as they scroll through their medical records, noted Helen Haskell, a patient safety advocate. Patients can request access to their records, but that’s a static page they’ll see only after getting care. That means that, while this hyper-alerting poses a danger, there’s no way for consumers to know if, say, they got worse care because the doctor missed a warning.
“It’s very rare that patients are granted that perspective,” she said.
Software vendors say they’re trying to make their systems smarter.
Epic Systems, outside Madison, Wisconsin, for example, has been hearing feedback for years from doctors about redundant or irrelevant alerts, said Bret Shillingstad, a physician who works on Epic’s clinical informatics team. They’ve added in functionality for hospitals to turn those alerts off. They’re working now to develop software that might target alerts based on things like a patient’s health condition or recommend medications that better match someone’s overall profile. Then there are simpler adjustments, like changing a system so that if a patient needs a vaccine, reminders just go to the primary care doctor, not the orthopedist, too.
In the long term, system designers are trying to better consider the nuances of a patient’s medical needs so that they can use fewer warnings and send them only when they matter, said Terry Fairbanks, an emergency physician and director of MedStar Health’s National Center for Human Factors in Healthcare in Washington, D.C. For instance, people with advanced cancer often need doses of morphine that might be unsuitable for other patients. A smarter system would warn doctors about that morphine order for patients who don’t have cancer but would treat it as normal for someone in the disease’s late stages.
Such a change could limit distractions so that physicians act upon pressing reminders — like notifications highlighting if a patient is at risk for sepsis, which can be deadly if it’s not noticed early.
But there’s still debate. Haskell said she would argue doctors should always be warned about certain medications and drug interactions.
“All of these alerts have really reduced medication interactions. It’s a service,” she said. “It just needs to be refined.”
But there’s clearly a cultural shift underfoot, added Phansalkar, who also works as director of informatics and clinical innovation at Wolters Kluwer Health, which supplies drug information to electronic health record systems. Alert fatigue is no longer “just something providers complain about,” she said. In health care, people are trying to devise more effective, nuanced ways for electronic systems to improve care.
“Because it’s so easy to put an alert to address a problem, that’s people’s natural, knee-jerk reaction,” said Douglas Gentile, medical director of clinical information systems at the University of Vermont Medical Center. But “as you add those, it creates additional problems. And you get collateral damage.”