Marrying Precision Medicine to EHRs is a Priority at UH Cleveland

University Hospitals Cleveland's technology platform is a key part of a long-term strategy to improve clinical outcomes.

Healthcare organizations can make patient genomic data actionable, incorporating it into the EHR.

So says Maulik Purohit, MD, associate chief medical information officer at University Hospitals Cleveland, a 25-hospital system with 6,000 affiliated providers covering northeastern Ohio.

In this interview with HealthLeaders, Purohit describes ways to evaluate organizational information technology and clinical readiness for strategic precision medicine programs, moving genetic information from tests to the lab to the EHR and then to the patient. This interview has been edited for brevity and clarity.

HealthLeaders: What kind of problem will this solve?

Purohit: Precision medicine is about using large sets of data to understand pathophysiology solutions, diagnoses, and treatments, and how they affect the patient. We have a lot of potential, where nearly everybody is on electronic records of some type, in converting that data to knowledge and using that, with the hope of improving care for patients. How did each individual do? Did the individual have side effects? Did the individual get the same level of improvement as somebody else? Those are questions that research studies couldn't answer because they're done on a population level.

HL: When you set out to get this genetic data, do you go for an opt-in approach from patients? Do you go for an opt-out approach? Are there pros and cons to either approach?

Purohit: Our approach is opt-in at UH right now. The provider will have that discussion with the patient. If it makes sense, they'll opt in for getting a genetic test. It would be nice if we get to a world where it is opt-out. I say that not with the intent of making somebody do something they don't want to, but I want to get to a place where we feel comfortable with genetic testing and what that means and what the implications are.

Maulik Purohit, MD, associate chief medical information officer at University Hospitals Cleveland. Photo courtesy UH Cleveland.

Genetic data is quite sensitive, and it can reveal things we may not have intended to reveal. There are implications for that in terms of insurance coverage. But it would be nice if we could be in a world where we can get genetic data without having to worry about some of those logistical issues, because I think the potential to help somebody with their genetic data is tremendous.

HL: With clinical trials, there's a very specific kind of aim and outcome that is desired. So it's kind of easy to identify that genetic data is actionable when it's a clinical trial. But if it's a more general practice of medicine, and standard of care issue, how does one make genetic data actionable?

Purohit: There's several ways that I think are in play right now. Our team has worked on pharmacogenomics, which is when you get the genetic testing done, then it matches that genetic code to different medications. One area is for depression treatment and a class of medication called SSRIs, selective serotonin reuptake inhibitors. The standard practice has been that you pick one of the SSRIs and do sort of a trial and error. That's a very dissatisfying process for a patient. It would be nice to get a better starting point.

With pharmacogenomics, what UH has been able to do is get a genetic test, and then look at how that matches with the SSRIs that are out there. Now you have a much better starting point to achieve a good outcome of treatment for the patient. That's not to say that it's perfect, because we don't have all the information from genetic code right now. But it's a much better starting place.

HL: Is the bulk of the actionable data pharmacogenomic in nature, or are there other forms of actionable data as well?

Purohit: For example, cancer treatment. Maybe you can look at a genetic code for designing chemo treatments. There's a world of possibility. We're evolving with our knowledge of what's out there in terms of the genome.

HL: When can genomics be used to bend the cost curve of healthcare? For example, by indicating alternate remedies such as exercise instead of prescribing a drug?

Purohit: It's more of an investment than it is a return. We're not there yet in terms of bending the cost curve tremendously. But if we don't make the effort to invest in it and explore it, then we won't get where we need to go.

HL: Are there particular disease management programs that are benefitting first?

Purohit: Pharmacogenomics is the hottest one, not only for depression, but also anticoagulation, like what is the best method of blood thinning to avoid clots, which can be important for preventing heart attack and stroke, some of the most disastrous illnesses and diseases out there.

Genetic counseling is probably up there as well. And I think we're getting to a point that, even for oncology and chemotherapeutic agents, it is helpful to have that in matching the type of cancer and the specifics of that with the treatment protocol. So those are probably the three biggest, but certainly evolving on a fast pace.

HL: To what degree do you bump up against the shortcomings of EHRs as you try to integrate genetic data into them?

Purohit: A lot of labs might send back a PDF, which is great, but it's not data that can be processed within the EHR. How do you combine that with existing knowledge of genetic code and what that means, converting that raw genetic data into a decision support tool that the clinicians can use? There are many steps in the process. It's certainly not easy, but these are the challenges that all of us are working on, along with the scientific knowledge about the genome itself. We don't have the solution worked out for the majority of this, but we are working on it.

HL: To what extent are you challenged by interoperability issues, whether it be patients who go outside your system for other care or payers whose data is not transparent to you, or other issues?

Purohit: What we had to do initially was have every lab out there build an interface to have the data transfer. Each interface is resource-intensive. Now we're using 2bPrecise to solve some of these issues.

HL: You have a variety of job responsibilities as associate CMIO. Where does this genomic data initiative rank vs. the other priorities that are on your desk?

Purohit: This falls into high priority. One thing that UH is undergoing right now is converting the EMR from the existing state to a full Epic integrated suite. That is priority number one. But if we ignore the R&D component and focus simply on the current [IT] aspect, then we're not going to move ahead.

HL: The biggest criticism we hear about vendor partners is the proprietary nature of what they offer, with algorithms that aren't necessarily shared because they're trade secrets.

Purohit: I certainly understand that. If you're a company and you come up with a solution and you give it away, it's not always the best business plan. I don't need to know what the algorithm is. What I do need to know is what's the performance metric? How well does it work? What's your sensitivity, specificity, positive predictive value, negative predictive value? Those kinds of things should be public, because it's a public safety issue.

We don't need to know how the sausage is made. But we do need to know whether the sausage is safe for people. Let's see if it performs the way we expect it to. And then if it does, great, then we can use it more. And that, you know, I think is a happy medium between giving away the secrets, which nobody wants to do, but at the same time, ensuring safety for the public.