Providers aim to take decision support to the next level

You might think of this as clinical decision support 2.0.

Healthcare systems are rethinking how they apply CDS tools to assist—and not hinder—clinicians in daily clinical care.

They are implementing CDS to help providers with decision-making in a variety of areas. When KLAS Research surveyed executives about the “core focus for your organization going forward” for CDS, 82 percent of survey respondents mentioned order sets, followed by surveillance (46 percent), care plans (43 percent), point-of-care disease reference (33 percent), infection control (26 percent) and diagnostic support (21 percent).

Many systems want to get more of their CDS tools from electronic health-records vendors. Most respondents using Epic (61 percent) and Cerner (58 percent) said they expect those vendors to play “very significant” or “significant” roles in their future CDS plans, KLAS found in its 2016 survey research. Far fewer respondents envisioned a “very significant” or “significant” role for other EHR vendors, including Allscripts (40 percent), Meditech (36 percent) and McKesson (33 percent).

Some examples of use cases include information about key safety issues, such as drug-drug interactions or risk of life-threatening medical problems; best practices to improve clinical outcomes or contain costs; and compliance with requirements from payers, most notably the Centers for Medicare and Medicaid Services.

With so many potential use cases for CDS, it’s not surprising that Frost & Sullivan expects the U.S. clinical decision support market to grow to $4.97 billion by 2021.

Overall, the challenge facing IT departments is how to design CDS systems that sift through vast amounts of data quickly, providing key insights just as providers are making decisions about patients’ medical care.

Combating Alert Fatigue
The first step: rethinking how IT departments use pop-up alerts. “We affectionately call this the sledgehammer alert. It is going to stop you,” says Thomas Selva, MD, a pediatrician and chief medical information officer at University of Missouri Health Care. “It interrupts your cognitive thought process,” he says.

UH-Patient-Care-Tower151208 PatientCareTowerSunset-013-2 (6).JPG
University Hospital's Patient Care Tower with sunset

Too many of these pop-ups can lead to alert fatigue, which happens when clinicians become desensitized to the warnings and then fail to respond to them appropriately. The problem could lead to patient safety issues if clinicians do not react to critical alerts, such as those warning of harmful drug-drug interactions.

The Agency for Healthcare Research and Quality, which has written about this topic extensively, recommends four steps to combat alert fatigue:

· Reduce or eliminate inconsequential alerts.
· Integrate alerts into clusters of related physiologic indicators.
· Tier alerts according to their importance.
· Apply the concepts of human factor engineering, which analyzes how well a device or product is likely to work in complex real-world environments.

Using Fewer Intrusive Alerts
The IT staff at Cheyenne Regional Medical Center recently addressed alert fatigue by overhauling its approach to Epic’s best practice advisories.

Lucy Stacy, program director for Epic at the medical center, said physicians and nurses complained that the alerts fired too often or fired for the wrong job roles. “In addition to that, on our side, we were getting multiple requests from multiple places to put these in place,” she says.

The team decided to act. The first step was a comprehensive audit of the medical center’s existing best practice advisories—they found 104 of them. They also compiled information on how often each alert was used or overrode.

The next step was forming a committee, which included representatives from the medical staff and nursing, to analyze the alerts. Based on the committee’s work, Cheyenne Regional Medical Center retired 25 of the alerts and kept 37 of them as is.

The committee also modified 42 other best practice advisories, ensuring that they only fire for the appropriate roles. In some cases, the committee created silent alerts that perform work in the background. For example, if a nurse documents in the EHR that a hospital patient has not had a flu vaccine, the system orders the shot automatically.

To avoid alert creep in the future, the IT staff developed a formal process in which the committee vets all requests for new best practice advisories. “We don’t put a BPA in place anymore unless it has been properly reviewed,” Stacy says.

“From a clinician’s perspective, they feel like they are being heard—that changes are being made,” Stacy says. The alert-reduction project has been so popular with physicians and nurses that the IT staff plans to go through the same analytical process for medication alerts.

Children’s Hospital Colorado also has cut back on alerts, using them in “a much more targeted way and only in the areas of greatest impact,” explains Angela Swanson, director of clinical effectiveness at Children’s Hospital Colorado.

For example, an alert in the emergency department lets physicians know if a patient is at risk for suicide, based on the results of a screening tool. “None of us want to send a suicidal patient out the door because we’ve failed to ask,” says Lalit Bajaj, MD, an emergency medicine physician and medical director of clinical effectiveness at Children’s Hospital Colorado.

A key question to ask when assessing the usefulness of an alert is if it will “immediately change my management of that patient,” Bajaj says. If the answer is yes, the alert probably will help clinicians.

Before new alerts are turned on for providers, the hospital rigorously tests them.

For example, the hospital is building an algorithm using machine learning to predict the risk of emergency department patients developing sepsis. The plan is to embed the tool “in the background, so that we can watch it fire, without notifying the provider, to see what the fatigue and what the firing rates might look like,” Bajaj says.

Other Flavors of CDS
Selva says MU Health Care is designing methods to provide information at the point of care in a less intrusive fashion, such as by placing key insights on the right-hand side of the screen next to the patient chart. “This is where we surface alerts as more of a nudge.”

For example, the academic medical center is using this approach to notify a provider if a child’s blood pressure is high—a calculation involving numerous variables, such as age and gender. “We do the math for the user. There is a slider that pops out that says, ‘You might want to pay attention, the blood pressure looks a little high. Click here to see what it is,’ ” Selva explains.

The functionality, called SmartZone, is part of Cerner’s Millennium PowerChart. MU Health Care was the first pilot-test site for SmartZone, which is now commercially available, according to Cerner.

MU Health Care also incorporates what it calls “related results” into the appropriate order screen. For example, as a physician is ordering a statin medicine, the patient’s most recent cholesterol levels will appear within the order screen without interrupting the workflow.

Cedars-Sinai Medical Center often designs ordering screens to cut down on pop-up reminders. One example is an order for a medication for which a patient takes an initial dose and then a follow-up maintenance dose. The order set will be structured so that physicians order both doses at the same time. “It tees it up for them in a way that makes it easier for them,” explains Yaron Elad, MD, a cardiologist and associate medical director of clinical informatics at Cedars-Sinai.

Despite the strides these institutions have made, reducing the number of intrusive alerts at the point-of-care is difficult because there are so many data points involved in providing medical care for a patient.

A case in point is a new requirement mandated in the Protecting Access to Medicare Act, or PAMA. The law, passed in 2014, includes a program to encourage the appropriate use of expensive imaging modalities, such as computed tomography, positron emission tomography, nuclear medicine and magnetic resonance imaging.

Beginning in January 2021, clinicians will be required to consult a certified CDS tool to ensure they are ordering these tests according to appropriate use guidelines. Medicare will not pay imaging centers for these tests unless they document that physicians used a CDS tool to order the tests.

Testing of the new program—without the negative financial consequences—is scheduled to begin in January.

Cedars-Sinai is in the process of evaluating potential CDS tools, which will be integrated into the EHR workflow, Elad says.

Education and Communication
Elad says the medical center doesn’t just rely on CDS solutions to encourage compliance with new programs, such as government mandates. It also creates educational programs and incorporates marketing tactics, such as creating screen savers or sending email blasts to publicize new mandates.

Children’s Hospital Colorado also uses a variety of techniques—including branding, education or pledges—in combination with CDS to launch new initiatives.

One example is a project to encourage pediatricians to follow care guidelines from the American Academy of Pediatrics for treating bronchiolitis. The illness, which is a viral infection causing mucus plugging and lower airway swelling, primarily occurs in infants and toddlers between the ages of one and 24 months.

The hospital wanted to decrease the utilization of X-rays, viral tests and bronchodilators to treat non-severe cases of bronchiolitis. While pediatricians commonly use these tests and treatments, evidence suggests that they are often unnecessary. And chest X-rays can have unintended consequences because bronchiolitis looks similar to pneumonia in these images, leading physicians to prescribe antibiotics.

To launch this improvement project, a multidisciplinary team developed a care pathway and then revised the order set to lead pediatricians in the new direction. They also produced a dashboard, which tracks providers’ treatment patterns, but they did not implement a pop-up alert.

They also used tactics to publicize the program and its goals, including creating a slogan, “rest is best”; T-shirts; informational handouts and videos for parents; and a pledge for physicians to sign.

Because of the program, the hospital logged utilization decreases of 40 percent in chest X-rays, 41 percent in bronchodilators and 22 percent in viral testing.

At the end of the day, however, education and marketing won’t drive adoption if end users conclude that the software tool is clunky and slows workflow.

Brian Edwards, an independent validation consultant for artificial intelligence vendors, says software developers typically don’t spend enough time making sure their CDS solutions work for end users. That’s because the standard software-development process doesn’t involve input from end users to the extent that is necessary in healthcare, he says. “The question of—Is it useful—is mostly unanswered by any clinical decision-support vendor,” he explains.

Selva agrees, saying that software vendors often focus more effort on meeting a specific requirement—such as complying with a federal regulation—than on usability. “You have a product that meets the requirements of the federal government, but the end user finds it very clunky, and it is not very helpful. Then what happens is people build workarounds,” he says.

For reprint and licensing requests for this article, click here.