Skip to main content
Premium Trial:

Request an Annual Quote

Cleveland Clinic Rolls out EHR-based Program to Alert Docs about PGx Tests for Azathioprine, Abacavir


The Cleveland Clinic is a month into a pilot program to test whether electronic health record alerts can serve as an effective way to integrate pharmacogenetic-informed prescribing into standard clinical practice.

The clinic last month launched its EHR-based Personalized Medicine Program across all of its facilities, including its main campus, seven regional hospitals, around a dozen family health centers in Northeast Ohio, and branches in Toronto, Florida, and Las Vegas.

The program currently covers two drug/test combinations: TPMT testing for the immunosuppressant azathioprine and HLA-B*5701 testing for the antiviral abacavir. When doctors anywhere within the Cleveland Clinic system order one of these drugs through the Epic EHR, an onscreen alert recommends that they consider ordering the test. If doctors choose not to order the test, the system prompts them to explain why. In cases where the patient has already received the recommended test, that information is already included in the health record and, instead of the alert, the Epic system provides prescribing recommendations based on the results of the test.

Kathryn Teng, director of the Center for Personalized Healthcare at the Cleveland Clinic, said that the goal of the project is to develop a system to seamlessly integrate pharmacogenetics into healthcare and to "influence the culture of clinical practice."

While improved patient outcomes are the ultimate goal of the effort, "the initial metrics are more focused on process," she told Pharmacogenomics Reporter. "It’s a lot of process measures in terms of what do the clinicians do with that information? What do they hit for overrides on the alert? Was there acceptance? All of these different process parameters."

The effort is structured into two phases. In the first, currently underway, physicians are alerted to order a PGx test at the time they prescribe the drug. Drug/test pairs are chosen for this phase of the project based on "the amount of evidence available and the amount of clinician buy-in," Teng said.

In the second phase, now in the planning stages, a population of patients will be prospectively genotyped with a PGx test panel. In this phase, all the genotyping results will be entered into a database but only the PGx drug pairs that are already approved for the first phase of the program will be entered into the Epic system. In this scenario, when a physician writes a prescription in the EHR for one of these drugs, the system will provide real-time decision support based on the patient's genotype.

Teng said she is currently building a business plan for the second phase of the project that she hopes to have completed by the end of the year. She noted that both arms of the program will run in parallel for the foreseeable future. So much is still unknown about the genetic basis of drug response, she said, "that you would only pull into the electronic health record what you know is your standard of care, and that's what part one is all about. So part one is an essential piece of being able to do part two."

She noted that prospective genotyping will ultimately provide long-term cost savings compared to multiple single-gene tests and that it will also provide an advantage in terms of the timing of the information. "You'd have the information up front before the patient is ever put on the medication," she said.

In a presentation outlining the program at the Molecular Medicine Tri-Conference in February, Teng said that prospective genotyping should be of particular interest to cardiovascular doctors, who she said have been resistant to adopting PGx testing.

"They do tell me that if they had the information in advance of the drug prescribing, they would use the information," she said. "So they don't want to pay for it, they don't want to order the test, but if they had that information they would use it."

In a follow-up interview with PGx Reporter, Teng emphasized the tremendous amount of planning that went into the project prior to its launch. She and her team conducted an evidence-based review of the literature and created a PGx advisory board comprising clinicians from each unit of the clinic as well as pharmacists, pathologists, and IT staff. Smaller PGx drug workgroups then helped prioritize the drug/test pairs to include in the program. Another workgroup focused on the development of the EHR alerts, considering questions like the availability of the test, the language used in the alert, and how to track clinician behavior.

"There was a lot of change management that went on along the way," Teng said. "It was about how do you elicit culture change? How do you build consensus for entire Cleveland Clinic? Not just each individual doctor doing their own thing, but how do you really decide what the Cleveland Clinic's standard of care is going to be in relation to these specific pharmacogenetic tests?"

In terms of the criteria that the advisory board considered in choosing azathioprine and abacavir for the first drugs in the program, Teng said that recommendations from the US Food and Drug Administration played some role, but the group ultimately made its decision based on a range of criteria. "We basically said to them, 'Which do you consider to be standard of care? Which do you think has enough evidence and justification for use? And which is really going to impact patient care?'"

The next drug to be added to the program will be the anticonvulsant carbamazepine, for which the FDA recommends testing for the HLA-B*1502 allele because it is associated with the risk of developing Stevens-Johnson syndrome.

Teng said the Cleveland Clinic is hoping to add carbamazepine to the program by the end of 2013, but noted that this example highlights how "every gene/drug pair is going to be a little bit different because of the clinicians involved."

In the case of carbamazepine, she said, "it's a little bit tricky because you have a lot of doctors who are not necessarily cohesive in terms of how they work together. Carbemazepine is prescribed for headaches, general neurology, mood disorders, chronic pain. So you have doctors from so many different types of practices that you're trying to bring together to build consensus."

Teng said she hopes to add one or two drug/gene pairs to the program per year.

As the program rounds out its first month, "it's been very well received," Teng said.

"The clinicians for whom the alert has fired and who have answered our surveys have told us that they find the language is easy to understand, they don't have any questions about the alert or why it's being fired or what do to with it."

There have been a few hiccups, however. For example, "we did have some nurses and nurse practitioners who didn't quite realize that this was rolling out and they had questions about why they were getting this alert."

In addition, a "design flaw" in the Epic system allows doctors who decide not to order a test to ignore the alert altogether without providing feedback on their decision. Teng said her team is currently reprogramming the system so that doctors must provide a reason if they choose not to provide a recommended test.

On the other hand, the system has demonstrated a clear benefit for doctors who prescribe the PGx-guided drugs very rarely and may not have been aware that a test was recommended or known how to order it.

For example, several clinicians said they knew there was some sort of PGx recommendation for azothioprine but didn't prescribe the drug enough to really know the details. "And when they got the alert they realized, 'Oh, it really is something we need to do,'" Teng said. "So that was a quality issue right there. That was our way of communicating to them that this is a quality standard for us and you should do this."

Furthermore, she said these doctors appreciated the fact that they could order the test directly through the EHR because it saved them time.