How Penn Medicine primed its IT infrastructure for precision medicine
Penn Medicine has positioned itself at the forefront of genomics and immunotherapy, but capitalizing on the promise of personalized medicine has required extensive investments in the medical center's IT infrastructure, said Brian Wells, Penn' associate vice president of health technology and academic computing.
Wells, who will speak June 12 at the Healthcare IT News Precision Medicine Summit in Boston, explained some the technology innovations Penn Medicine has made as it works to improve integration with its electronic medical record, use analytics tools to mine unstructured data and offer real-time decision support.
Q. Talk about Penn Medicine's journey to where it is today.
A. Penn Medicine has been around a long time: first hospital in 1751, first med school in 1765. I guess all of that leads up somewhat to where we are today. In 2010 or so, Penn began to emerge as a powerhouse in the genetic engineering and immunotherapy space.
In 2013 we capitalized on that when we published our strategic plan, one of the big pillars of which was precision medicine. We brought in a vice dean for precision medicine and really began to focus the energy of the organization, around building the technology infrastructure and the clinical and research expertise to excel in that space.
Our cancer center director, in conjunction with the chair of pathology, who came from NYU, put together what we call the Center for Personalized Diagnostics. This was one of the first local area efforts to build a CLIA-certified genetic testing lab that was going to focus on testing of solid and blood-borne tumors to find the genetic makeup of the tumor and tailor the treatment based on those findings.
That, along with immunotherapy work that was being driven by Carl June (director of translational research at the Penn's Abramson Cancer Center), really were the things that launched that initiative and created a focus on precision medicine.
Around that time I also was asked to focus on the infrastructure and data for the school of medicine. I was asked to bring together common systems, common approaches and centralized management of infrastructure and data for the school of medicine.
My to-do list was building and buying technologies that the school would need on the research side, that would intersect with the clinical side. That is a lot of what I'm going to be talking about in Boston – what we've done, and what we have left to do. I'll be focusing on the research aspect of IT and what it takes to build a research enterprise that can support precision medicine, and then what it takes on the clinical side to build a clinical infrastructure to support precision medicine.
Q. When you first took at the scope of what you had to do, what were some of the gaps you saw?
A. I think one of the first big gaps within the school is that there was not a centralized team. Typically in schools of medicine, IT is a regional thing – it grows up within a lab or a division or a department within the school. Maybe radiology has their own people and pathology has their own people and the cancer team has their own people and they're not integrated.
One of the first things we did was to say, look, we're not going to get down this road to precision medicine if we don't have centralized support and a holistic view of IT within the school. And that team must report to the CIO.
We have one CIO, and one consolidated budget for IT investments across the school and the health system. So that's what we did, and that's what I worked really hard for. By the middle of 2012, we were able to announce the formal creation of this team we called Penn Medicine Academic Computing Services. And now it's grown to 100 people who are dedicated to the school's IT needs, from desktops to servers to storage to applications.
The other thing we did was ask, what are the high-priority applications that the school didn't have? For example they didn't have a common laboratory information management system. They didn't have a common sample management, sample inventory, sample tracking system. They didn't have a data warehouse to aggregate and collect all the data and put it in one place to then link back to the clinical data. It was difficult for researchers to get access to clinical data in an easy way.
We didn't have a way to share our de-identified data with industry partners who could search that data and identify patient cohorts at Penn that could be used for bringing sponsored trials to Penn.
We didn't have an enterprise-wide clinical trial management system that was integrated with our EMR. We just accomplished that, in fact, about two or three months ago.
So there was a litany of things. Some of them were small, some of them were big. But the biggest thing was the culture. We didn't have a culture of common systems and central management and collaboration with our faculty to really build this suite of tools that we were going to use and share and contribute to.
We were able to prove that by putting all the data in one place it created an exponential increase in value. Sort of like Metcalfe's law: The more nodes on a network, the more valuable the network is. The more genetic data we put in one big repository, the more useful that repository was to everyone.
There was some resistance. There were folks who said, "I'm not going to share my data because this is my research and I haven't published it yet." And so you can't know what my samples are and you can't have my genetic results. We had to prove to people that we could protect that data, we could limit access to it, but still put it in one central place that everyone could get to. At least to look through and find cohorts of interest to them – not actually take the samples and not actually use the data for anything, but at least be aware that it existed to stimulate additional research.
Q. What were some of your priorities with regard to EMR integration?
A. There were several places where we wanted to integrate with the EMR. One is in tracking patients on trials: You want to know who's enrolled, and the fact that they're on a study, and which study are they on, who's the principal investigator, who do I contact?
That was a big piece for us. We just went live with that a couple months ago. We're actually the first in the country in an academic environment to have a two-way interface between our EMR and our trial management system that we can track enrollment. So you can enroll or unenroll a patient in either system and they keep each other in sync. That's really key, all the clinicians know what patients are on what trials and they can be aware of their participation in a trial while they're working in the EMR.
Another big piece – because we have to make sure we're compliant in how we get paid – is the whole research billing aspect.
So we're now doing our research billing reconciliation within the EMR. Every day our study coordinators are looking at the charges that have come in for the patient so the right charge gets billed to the study and the right charge gets billed to the patient and/or their insurance company. That has become infinitely simpler now with the integration with our EMR.
Another big area of integration that we're still working on, and no one has completely solved yet, is the ability to bring genetic results into the EMR discretely and store them in a discrete way so that you could write rules against them: You could provide alerts to providers that say, 'This patient has a certain genetic makeup that makes them a high-metabolizer of nicotine so if you are considering giving them a patch or Chantix, you might want to go with Chantix because it provides more nicotine replacement than a patch might.
There are those types of genetic results that we still can't today find a place to store in the EMR, but we're working with our vendor to try to fix that and we're hoping we'll find a solution soon.
Another big piece is that we still do send a lot of or genetic testing out to external labs, and that data is not flowing electronically, either outbound or inbound to us as far as what's been ordered and what the results are.
Those are things that are on our docket to work on, but we do need the help of our vendors to make that happen.
Q. What advice do you have for smaller organizations looking to try some similar precision medicine initiatives?
A. I would be inclined to outsource the whole thing. I would go find an academic medical center and either outsource it to them, or find a partner you trust. For example there's a company called Tempus in Chicago that's attempting to be a pretty high-powered genetic processor and precision medicine company.
If you're a small community hospital or someone without an academic arm, you need to find a partner that can manage and explain this to you, because it's not easy. Even our own clinicians at Penn that are not actively working in the pathology and genetics space need help interpreting the findings.
I would not try to build this myself. There's just too much knowledge that you just don't have in your local environment.
Q. What's next for precision medicine at Penn?
A. The EMR integration of genetic results is number one on my list. I really want to prove that that can be helpful. I'd like to build what we call molecular decision support so we have the logic and the rules in place to do several things.
One, if we have genetic data in the EMR, we can do a better job of recruiting patients into trials, because of their genetic makeup, at the point of care. We can do better decision-making around testing and prevention and additional sequencing work if the data is in the EMR.
But it has to be presented in a way that's clear and understandable. Providers need to understand it. It can't just be a bunch of numbers, there has to be some textual explanation of what this means and what the options might be.
Another one is continuing to refine how we capture data in the EMR to make it more and more discrete all the time. There's a very healthy tension between capturing more discrete data and slowing down the visit, so we're trying to find ways to mine more data out of the EMR using natural language processing and unstructured data analytics tools.
And we're also trying to find out if we can strike that balance between how much is discrete and how much is not. Even things like figuring out the stage and grade of the tumor is not consistently stored in a field that is discrete and easily reportable.
A lot of patients don't get their care all in one place. They may get their care outside of the enterprise and then come to Penn bringing paper reports with them. That stuff is hard to work with as well. So as we look at health information exchange, and trying to get more discrete data electronically, those are all areas that are on tap to be resolved.
Q. How do you expect precision medicine to evolve in the coming years? Will it get easier with the right technology infrastructure in place, or get more challenging as genomics knowledge evolves and becomes more complex?
A. I don't think it's going to get easier. It's going to get more and more complicated. There's going to be more and more discoveries, more and more genetic data that's going to be determined, more and more markers that are going to be identified, there's going to be combinations of markers, and there's variations in timing and sequence – you had this marker early on in your cancer and it moved to this – I don't think it gets any easier.
I think we need standards defined with regard to genetic data and how it's stored discretely and how it's transmitted discretely to various enterprises, and I don't think we're really there yet.
We've got regulations and ethics issue about data sharing – how much do we have to share with your children and your parents about you, and can we get access to your parents' genetic data or is that considered protected health information that we're not allowed to have because we're not treating them? There's all kinds of challenges in the ethics and legal space as well.