When an individual is diagnosed with cancer, he or she goes from having a primary care physician to having a team of care providers, including nutritionists, surgeons, and radiologists. It can become overwhelming, to put it mildly. It’s precisely why Sarah Cannon, the Cancer Institute of HCA Healthcare, implemented a system of cancer navigators to help guide patients and their families by coordinating appointments, answering questions, and providing education. These cancer navigators, says CIO Andy Corts, “are our most precious resource.”
The challenge came in harnessing the data — which can be complicated in any area, but is infinitely more difficult in the “incredibly fragmented” oncology environment. Corts and his team have made it their key priority to combine data sets into a common warehouse and leverage analytics to be able to “view the entire patient journey.” In this interview, he talks about how they’ve been able to define a cancer data model, how they’re partnering with Digital Reasoning to automate manual processes and enable more personalized care, and the journey that brought him to HCA, and eventually, Sarah Cannon.
- About Sarah Cannon (cancer service line for HCA’s 185 hospitals, managed services for 50 cancer centers, clinical trial platform)
- Managing data in an “incredibly fragmented” environment
- Foundation of a data warehouse & EMPI
- “We didn’t have a defined data model for cancer.”
- Acquiring GenoSpace to focus on personalized medicine – “It allows us to bring genomic data in and map it so we have one standard.”
- Challenges with clinical trials recruitment
- Working with Digital Reasoning on AI
- “Now we have a model that can read & flag it, so we can act upon all those diagnoses.”
LISTEN NOW USING THE PLAYER BELOW OR CLICK HERE TO SUBSCRIBE TO OUR iTUNES PODCAST FEED
All of those systems roll up to us and we have a common database for all cancer patients. We have a common infrastructure where we can see exactly what’s going on in all of our cancer centers at any one moment.
You can’t just choose an EMR. It can’t be an Epic. It can’t be a Cerner. It can’t be a Meditech, because the actual beam is controlled by the EMR, and that’s where all the data goes.
Through this investment with GenoSpace, we now have the ability to structure and manage genomic data, and we’ve seen tremendous increase in the ability to enroll patient to specific clinical trials for immunotherapy and targeted therapies. That is a significant capability.
That data is critical for us to grab, but it always came back to our EMR system in an unstructured way. And when it came back unstructured and we had to either launch navigation programs or reach out to primary care providers who had ordered the colonoscopy, we had to read through each pathology result or report to actually find the patient. It was an incredibly manual effort.
Gamble: Thank you, Andy, for taking some time to speak with healthsystemCIO.
Corts: My pleasure.
Gamble: To start off, can you give a high-level overview of Sarah Cannon?
Corts: Certainly. Sarah Cannon is the cancer service line for all of HCA, which is a massive organization with around 185 hospitals. Sarah Cannon is a subsidiary that manages the cancer service line for HCA. We have around 50 cancer centers for which we provide management services. We also have a clinical trial platform where we work with medical oncologists, whether it’s revenue cycle management, compliance, FDA management, or the IT platform that they plug into, to manage clinical trials. We manage entire enrollment process from end to end.
Gamble: And there are different sites around the country and one overseas?
Corts: Yes, we have a site in the U.K that falls under our cancer care operations, and we also have a drug development unit there. That’s a phase 1 clinical trial unit — patients come to that center to get first-in-man treatments. And so, right as a drug comes out of preclinical and goes into clinical, that’s where patients are going to be enrolled. We have a number of those drug development units across the US as well.
Gamble: As far as HCA, what’s the relationship or reporting structure there?
Corts: I’ll give you some history. HCA is a lot like other hospital companies in that cancer care was predominantly taking place in the outpatient setting. They have a huge presence in terms of surgical oncology, but when it came to medical oncology, they didn’t have a big presence outside of a few markets. That changed in 2006 when HCA announced a joint venture with Tennessee Oncology, specifically the Sarah Cannon Research Institute. That joint venture was 100 percent dedicated to clinical trials and ultimately building a platform for medical oncologists. A lot of times, private medical oncologists are competing with academia, which is hard because academia has the research infrastructure.
And so we reached out to them and said, ‘Look, you can enjoy the benefits of private practice but also plug into our platform. And through that platform, we can give you everything that academia could in terms of the overhead so you don’t have to negotiate a trial with Pfizer or AstraZeneca; we can do that on your behalf.’ This happened until about 2012. Because HCA didn’t have the strongest presence in cancer, they reached out to Sarah Cannon and said, ‘We know you’re doing all this great stuff in cancer research. Can you take all of the findings that you’re doing in research and apply that to every day cancer care throughout the HCA system?’
At that point, we moved to doing not just research, but a refresh of all of our cancer centers that are now managed by my team. All of those systems roll up to us and we have a common database for all cancer patients. We have a common infrastructure where we can see exactly what’s going on in all of our cancer centers at any one moment. At that point Sarah Cannon had bought out Tennessee Oncology, and so now it’s a 100-percent wholly-owned subsidiary, 100 percent dedicated to cancer care for HCA.
Gamble: So obviously you’re dealing with some pretty huge amounts of data. Let’s get into that. First, what type of EHR system do you have in place?
Corts: This is what makes cancer care so challenging — it’s incredibly fragmented. On the inpatient side of the house, we have Meditech, so all of the cancer surgeries happen within Meditech. In the outpatient setting, we typically have two options for patients, radiation and chemotherapy. All of our radiation therapy systems are under the Varian EMR, and that is FDA-regulated, which means you can’t just choose an EMR. It can’t be an Epic. It can’t be a Cerner. It can’t be a Meditech, because the actual beam is controlled by the EMR, and that’s where all the data goes. And so we have Varian for all of radiation oncology.
With chemotherapy, it really depends on who our medical oncology partner is. Most common that we see is Flatiron, but there are other vendors out there as well. We’re bringing all three of those different data sets into a cancer data warehouse so that we can see the entire patient journey.
Gamble: How have you been able to do that? I imagine it’s a pretty complex process.
Corts: Yes. About 10 years ago, HCA made a significant investment in a data warehouse, as well as an enterprise master patient index — those two investments were the foundation we built upon. The piece that was probably the trickiest is we didn’t have a defined data model for cancer, and so we basically took it upon ourselves to define that standard. So we had to figure out, what’s the standard when we bring in radiation therapy data? What are the key data points we’re looking for when we bring in medical oncology data? And then when we bring in the surgical oncology data, what are those pieces that we need to be structured? All that comes and lands in our data warehouse. The trickiest part, I think, is defining the cancer data standards that ultimately make our analytics work and also drives our outcomes research.
Gamble: To your knowledge, are other cancer organizations are dealing with that as well, having to define those data elements?
Corts: I think it’s definitely behind where the rest of the industry is in terms of a data warehouse because the journey can be very fragmented from inside the hospital to outside of the hospital to all of the partnerships. You have to really align all those interests to get the data to flow into one place. I do think we’re ahead of the curve. One of the things that’s nice about HCA is that it isn’t very niched. If you go to some systems and they say they’re doing cancer care, a lot of times that rolls up under the imaging team, and so the imaging teams that do a lot of the cancer functions also end up owning other pieces of either the radiation oncology system or the chemotherapy system. I’m lucky at HCA, because I have a team of 90 people that report up to me. We’re focused fully on clinical trials for cancer or care systems for cancer. We’re really lucky that we have the bandwidth and the attention now. It helps that we have scale, but the size makes it feel like it’s not niched.
Gamble: Right. What about personalized medicine? I’m sure that’s something you’re working on.
Corts: Definitely. We made a big acquisition two years ago of a Cambridge, Massachusetts company called GenoSpace. That addressed a gap that my team had by bringing that genomic data model together. Typically, cancer patients want to get molecular-profiled during their journey. That molecular profile will ultimately tell the genomic disposition of your cancer tissue.
In our network, however, we use a lot of different labs. There’s Foundation Medicine, which does genomics and molecular profiling. We also have Guardant. We have local labs and regional labs, and we wanted to make sure genomic data was part of our cancer data warehouse as well, and so we acquired GenoSpace. What that did was, whether we were getting a feed from Foundation Medicine or Guardant, we could bring that genomic data in and map it so that we had one standard for all the data warehouse and all the analytics that function from it. That’s been a real differentiator, particularly in the clinical trial space, where 70 to 80 percent of the trials we offer in cancer require a molecular profile. It requires genomics. We weren’t able to find the patients we needed to enroll in clinic trials without having this analytics infrastructure to be able to say, ‘We’re looking for this type of patient with this genomic biomarker.’ By not having that data structured and mapped to a standard, we couldn’t query appropriately to find the patient, and so we were seeing enrollments on our side of the house go down. But now, through this investment with GenoSpace, we now have the ability to structure and manage genomic data, and we’ve seen tremendous increase in the ability to enroll patient to specific clinical trials for immunotherapy and targeted therapies. That is a significant capability that we now have today; the analytics that result from it, particularly from an outcome perspective, have been tremendous to what the organization has been able to do.
Gamble: Really interesting. And that’s just on the genomic data side.
Corts: The other big piece was the investment with an AI partner of ours, Digital Reasoning. What they do for us analyze that pathology data. In most cases when you come into an HCA hospital and you have a colonoscopy done or a biopsy of certain tissues, what typically happens is our pathology labs will send off that tissue and actually dictate, ‘This is the disposition of that tissue from a pathological perspective.’ That data is critical for us to grab, but it always came back to our EMR system in an unstructured way. And when it came back unstructured and we had to either launch navigation programs or reach out to primary care providers who had ordered the colonoscopy, we had to read through each pathology result or report to actually find the patient. It was an incredibly manual effort.
One of the things we did with Digital Reasoning was to put different pathological AI models on top so that without having a human to intervene, we could flag every report that actually had a cancer diagnosis. Now, keep in mind, HCA probably sees over 2 million pathology reports a year. That’s upwards of around 2 million procedures that are being done; and of that, we harvest about 100,000 cancer patients. When you have to use a manual process, the conversation goes like this: ‘Hey, do we have any cancer patients this week in the cancer center that have been recently diagnosed?’ ‘Well, we don’t know because we haven’t read those reports yet.’ Now we have a model that actually can read it and flag it, and then we can upon all those diagnoses.
healthsystemCIO’s Interviews and Podcasts are sponsored by: