Electronic alerts are intended to improve safety and efficacy for patients. But because of the volume and poor design of most alerts, the result has led to wide-scale fatigue and lack of engagement by physicians, nurses, pharmacists and other care coordinators.
A 2013 study of clinical decision support alert effectiveness by Nanji and colleagues showed that 52.6% of alerts were ignored. A study by Issac and colleagues found a 90.8% override rate for drug-drug interaction alerts among ambulatory care providers.
The problem is caused by volume and the complexity of the communication, explains Alissa Russ, PhD, a VA research scientist. “It’s not just that there are too many alerts or too many false alarms, it’s that some alerts have way have too much text.”
Phil Smith, MD, author of Making Computerized Provider Order Entry Work, has been studying clinical decision support and alert systems for several years. He started his work in an environment where the out-of-the box vendor solution would have triggered an excessive 114 alerts per 100 medication orders.
Smith and colleagues worked to refine the system. After several rounds of enhancements, they came up with a ratio of seven alerts per 100 medication orders. Of the seven alerts, six resulted in doctors choosing to do something different after receiving the alert.
“We went from probably 100 false alarms per 100 orders down to one false alarm per 100 orders,” Smith says.
Now the chief medical information officer at Moffitt Cancer Center in Tampa, Florida, Smith says one problem is that most developers don’t approach alerts as a strategic opportunity, but view them as an operational necessity.
“They end up trying to use alerts and rules to mitigate behavior and structural issues and [are] not really thinking through what’s important to [include in the] alert,” he says. Institutions and companies also often fail to consider who needs to see the alert and who doesn’t, Smith adds.
In practice, alerts aren’t just improperly used, but are creating nuisances, according to Ron Razmi, MD, a cardiologist and chief executive officer of Acupera, a digital health solutions provider in San Francisco. Physicians are not receptive to alert systems or data that interrupt their work flow or add time to their work day.
“Right now, at the point of the discussion, it’s not really practical for doctors to be able to take a piece of data from something that’s outside of their four walls, know what context that data is in and do something with it. [That] would require them to change how they work,” Razmi says. “They are extremely resistant to that because, right now, practices have to see more and more patients. Asking them to also incorporate this into the practice is a lot to ask.”
To reduce alert fatigue and to make them more effective, alert systems need to be timely, focused on the key information physicians need and trustworthy.
“If you give [doctors] bad data once or twice they are never going to trust it again,” says Keith Blankenship, vice president of technical solutions for Lumeris, a company that provides technology and consulting services for risk-based population health management and value-based contract
When it comes to successful alert systems, three key considerations include:
Russ said her research shows that the content of the alert isn’t the only element to consider. The design of an alert can also make an impact on physician behavior.
In one study, Russ was able to test how an alert’s design affected this. In fact, one key finding showed that simple design changes documented several positive outcomes including a reduction in prescribing errors, increased satisfaction and improved efficiency compared with the original alert design.
Physicians typically prefer alert systems incorporated into the EMR, Blankenship adds. If it isn’t possible to incorporate an alert into the EMR, it should be as convenient as possible for the physician-even if it means having someone from the front desk print out the data and attach it to the front of the chart.
One physician group asked to see information upfront on patient utilization. The providers wanted to know how much a patient was actually costing the practice compared to a cost benchmark based on the condition or disease.
“If you see that a patient should be costing you $10,000 a year, and they are costing you $40,000 a year, then you need to get involved,” he adds.
As more providers move away from fee-for-service structures, Blankenship says physicians may have a greater responsibility to view and act on alerts-particularly in financial arrangements where they’ve taken on risk.
While there will always be a need for clinical decision support, Smith says most providers haven’t yet spent the time
necessary to refine their alerts to produce the best results.
“They allow too many people to design alerts and not enough people to really vet them out and make sure that they are effective,” he says.
Jill Sederstrom is a freelance writer based in Kansas City.