If you think you have trouble fitting your message into 140 characters for a Tweet, imagine trying to keep a pharmacogenomics alert to fewer than 15 words. That is the challenge Mark Dunnenberger, M.D., faces at NorthShore University HealthSystem. “We have been told that any more than that is too much for a clinician to read while they have a patient in front of them,” he said.
Dunnenberger is senior clinical specialist in pharmacogenomics in the Center for Molecular Medicine at Evanston, Ill.-based NorthShore, an integrated system with four hospitals and 2,100 affiliated physicians. In a recent interview with Healthcare Informatics, he described some of the challenges all healthcare systems face integrating genomic data into EHRs as well as how NorthShore has approached integrating pharmacogenomics alerts into clinician work flow.
Dunnenberger said most large integrated health systems are struggling with the multi-faceted challenge of getting complex genomic data into formats that clinicians can understand in an intuitive manner so they can integrate it into their current decision-making process.
“Genomic data only makes sense for a patient and patient care when put in context of everything else,” he said. “If we are really going to get implementation of personalized medicine or genomic medicine into practice, we can’t be taking the clinician out of their normal work flow, making them think about this data in a different way through a different process and then bring them back into their current work flow. That is just not going to work. Either they are going to make the wrong decision because they don’t have all the data in front of them or ignore it because it is outside their normal work flow.”
He added that harmonizing data standards of laboratory information systems and EHRs is another problem, along with figuring out how much data to send across. The amount of data you generate in a laboratory system is very different than the amount you would deliver to a clinician. “I have a feeling if I were to give them all the data I had, their eyes would glaze over,” he said.
In many health systems today, a lab report with genomic results comes to the clinician in PDF form, which is a big problem. “There is a tab in the medical record where scanned documents go and that might as well be a black hole because it is very hard to navigate through there and find anything you want, especially if it is something that was done a couple of years ago,” Dunnenberger explained.
With his area, specialty pharmacogenomics outside the realm of cancer, the data collected at any point in the patient’s life is applicable at every point after that. “Imagine you had a laboratory test done five years ago. Now you are seeing a new clinician, and that data point could be really important to your care, and that new clinician has to know that the test was done, where to find it, and know what to do with it. We can’t rely on PDFs for that,” he added. “Instead, if we can get this data into the record discretely and build tools around it, we can then deliver it to the clinician with the right amount of information at just the right time. They may never have known you had that testing done five years ago; it is just delivered to them in a nice little package.”
Partnering with ActX and Customizing Alerts
To deliver its clinicians actionable alerts about pharmacogenomics, NorthShore initially tried to use the native rules engine in its Epic EHR. “What we found was to deliver the amount of complexity and detail we wanted to, it was going to be cumbersome and almost impossible to do it in that system,” Dunnenberger said. “It would have been possible, but doing it that way would require a lot of resources.”
Instead, the health system partnered with a company called ActX that provides a service that enables real-time checking of each prescription against the patient’s genetics. “They give us the framework to build the rules and logic for how to present this data in an actionable way to clinicians,” Dunnenberger said. “A lot of EHRs have interruptive alert systems with rules behind it. ActX gives us a logic engine designed to deal with genomic data. We send them the patient results and the genetic variance; we use their rules engine and customize it. It operates via a simple web service and API callout to their server any time an order is placed for a high-risk medication that has a link to pharmacogenomics, where you would want to make some change to therapy.”
The system calls out to ActX, which send back an answer that is an interruptive alert that pops up in the clinician’s work flow just like a drug-drug interaction alert would. NorthShore has access to a pharmacogenomics profile — a list of medications that are a problem for a patient: what the problem is, how to get around it, and what the genetics are. “The beauty of this list is that it is updatable all the time,” Dunnenberger said. “Every time a clinician clicks on it, they get the most up-to-date information as we put it in the system. We are getting away from that PDF, which is a snapshot in time, to a dynamic list. If we learn something new about a genetic variation you have, the clinician gets that information and is not stuck with what we knew a year ago. The cherry on top is that it is embedded in our EHR. All a clinician has to do is click where they would click to see a lab result, and it pops up. They don't’ have to log into a different system or close the EHR. It is right there and is starting to get embedded into their work flow.”
This process does require NorthShore to maintain that knowledge base: ActX can do it to some degree, but Dunnenberger said that to hone in on what clinicians want to see at NorthShore, customization is required. For instance, both cardiologists and neurologists could be alerted on one particular drug-gene pair, depending on the indication. But the strongest data so far is with cardiology. “Our neurologists don’t want to see those alerts, so we have had to go in and customize the alerts to state that they are not preferred by neurology or limit it so it doesn’t trigger for neurologists,” he said.
NorthShore has doing pharmacogenomics in clinical care since March 2015 and has been using the ActX system since September 2016. This has allowed it to take the ordering of genomic testing out of the pharmacogenomics clinic and put it into the hands of clinicians. “We targeted primary care physicians at the beginning so that has allowed us to exponentially grow the impact of pharmacogenomics,” he said.
“Previously, when all we had were the PDF files to send to providers, we weren’t sure they would be able to get the full context and incorporate the results,” Dunnenberger explained. So they would have patients visit the pharmacogenomics clinic for both education about risks, benefits and limitations of testing and the actual testing and sharing of results. “Now we have it set up so the primary care provider has the tools at their disposal. We can deliver results to the clinician in their EHR. Patients no longer have to come to our clinic; however, they may want to come and get a deeper dive into what the data means, but that is up to the patient.”
NorthShore has worked to get the clinical decision support alert distilled to less than 15 words. They get a 15-word alert, then a button that says click for more information. If they click on that, they get a two-minute summary and references to go further from there if they want even more information. “We are really trying to meet their needs and it is not clear what those needs are,” Dunnenberger said. We are taking an iterative approach. We build something, test it, see how it works, get that feedback and improve, and get it back out there.”
Looking to the future, Dunnenberger said that how and what data his clinic provides to clinicians is always going to be evolving. “We have to be nimble in creating new solutions to get this in the hands of clinicians in ways that are intuitive for them to use. Also, we are good at looking at one gene at a time but the reality is there are multiple genes at a time that we have to evaluate to make a recommendation. We have to figure out how to do that and put it in the context of care.”
Dunnenberger said he realizes that genomic data is new and that it can be scary. “But the ways you apply it to care are no different than the other clinical factors you are already evaluating,” he said. It is just more information to help make better therapeutic decisions. It is there to help clinicians, not create new therapeutic dilemmas. As long as we can deliver it in a way that is helping them, I think we are going to have real success delivering personalized or precision medicine.”