Parsing the Data-Driven Opportunities to Improve Oncology Practice
Can investment in technology, analytics, and improvement processes transform how oncology is practiced and delivered on behalf of cancer patients? The answer is yes, even as the landscape remains complex, say oncology leaders and others in the trenches.
That was the consensus of the discussants participating in a panel entitled “Closing the Loop on Precision Oncology Care, on Monday afternoon, March 7, the first full day of sessions at the ViVE conference, being held at the Miami Beach Convention Center this week, and co-sponsored by CHIME, the College of Healthcare Information Management Executives, and HLTH. This is the first joint conference for the two organizations.
As the session’s description noted, “Medicine is not a one-size fits all solution, particularly not for cancer patients. Individualized treatment planning is essential for each unique therapeutic response even for patients with the same form of cancer, and individualization requires data. Precision oncology care is harnessing patient-specific information to change cancer management, healing, and outcomes. Join this panel of leaders and explore how technology is enabling individual data curation from family history to social preferences to cancer’s genetic makeup information. Learn the promise technology holds for efficacious cancer care planning that will deliver meaningful interventions, research, and personalized care journeys.”
The panel was moderated by Leonard Kalman, M.D., executive deputy director and chief medical officer at the Miami Cancer Institute. The other panelists were Dana E. Rollison, Ph.D., vice president and chief data officer and director of data science at the Tampa-based Moffitt Cancer Center, Jennifer Wesson Greenman, CIO of Boca Raton, Fla.-based Cancer Treatment Centers of America, C.K. Wang, M.D., chief medical officer at the New York City-based Cota Healthcare, and Abhinav Shashank, co-founder and CEO of the San Francisco-based Innovaccer.
Early on Dr. Kalman asked the other panelists how they viewed the journey into precision medicine and individualized care. Cancer Treatment Centers of America’s Greenman said that “Precision care is essential to our vision, because we see it through the lens of patient empowerment. Having an understanding of genetic alterations and variations, and being able to provide insights for the patient and clinician. I consider it our job in IT to provide those insights to them. We were recently acquired by City of Hope. We share a vision of an equitable healthcare environment, in which these innovation, disruptive treatments, many of which are premiered at Cancer Treatment Centers of America, can be provided, that’s core for us.”
Still, it’s important to remember, Dr. Wang noted, “The vast majority of cancers don’t have ana actionable result. We believe that all patients affected by cancer deserve a clear path to the correct care.” He went on to say, with regard to “learning from patients,” that “I think one of the big issues I had in my years in practice, is, I had a patient who showed up and asked how a patient like him fared without treatment; he didn’t want any treatment. And I had to tell him I actually didn’t know. Imagine if I had the answers. This is what we need to focus on, not just a narrow definition of targeted therapy, it’s about how we get the best care to the patient.”
So, how do we get the best treatments to patients? “Trying to abstract the data” remains a formidable challenge, Wang opined. “Currently, our medical recorded documentation system was never designed to facilitate care, it was designed for billing. And in terms of documentation, I just followed someone else’s lead. And a lot of the clinical documentation is quite messy and is sometimes unstructured. There’s a lot of technology being developed—machine learning, AI, NLP [artificial intelligence, natural language processing]—that information is quite messy, and you still need people parsing that data out. Something as simple as defining cancer staging—what stage a person is in in their cancer, is not actually straightforward. It’s still a very intensive process. We’re working on a few categories. As much as we can get data into a structured format, is best. At the same time, I don’t think the problem will be solved until physicians document differently. We’re working with one of our partners to develop a tool to layer on top of their clinical document system to capture the relevant information not only to get paid but to facilitate patient care in a more structured way, and that populates the note. Technology not only has to solve a problem, it has to work within the daily workflow of folks using it. How do we source and curate data faster and better? And how do we test out the newer technologies like AI and NLP? We’re using them to help guide people through the data abstraction process.”
“I do a lot of onboarding,” Kalman noted. “I think some physicians would be willing to use structured documentation processes, but our EHRs [electronic health records] aren’t there yet.”
“I’m an epidemiologist by training,” Moffit’s Rollison said, “and in my 18 years, I’ve been totally focused on how we drive the data into patient care. And working for years, partly with Jennifer, we’ve been building out our data warehouse, and trying to work with unstructured data.” Recently, she reported, “we’ve been working with cloud-based data to directly ingest data. The hope is that the technology, combined with our investment in the data science faculty, will allow us to detect those patterns in the data that we as humans cannot do ourselves. Of course, this requires rigorous process methods, but there is power in the technology, and we’ve just opened an artificial intelligence unit. Some of the AI and ML splashes in the headlines have been in the areas of chronic diseases, involving conditions like heart disease and diabetes, which are very clearly measured over time using discrete lab values. Meanwhile, cancer, which is many diseases, involves patient-reported systems, plus documentation of lab values… Image-based tumor markers, liquid biopsies, other elements, can inform the research. So it’s an informatics challenge, but also a research one, and the faster we can meet that challenge, the faster we can find a cure for cancer.”
“Oncology has its own set of challenges, in terms of how data is stored,” and how the science is evolved forward, Shashank emphasized. “Curation is an incredibly hard challenge, and is something incredibly hard to scale. Because of the curation process being so manually intensive, it becomes very difficult to scale any precision medicine downstream. But as it begins to scale—at Innovaccer—we have 500 engineers focused on connectivity, and making sure everything is mapped to a single data model. And there’s no central standard—and hopefully, FHIR [the Fast Healthcare Interoperability Resources standard] will come into play—where there are more standardized ways of storing data—it’s hard.”
Indeed, Shashank said, “Before we made a single dollar of revenue, we had 350 engineers focused on all the major EHRs, and everywhere, the data is stored differently. How do you stored the data correctly and derive the insights? Oh my God, it’s hard. You really figure out what the minimum common denominator is that you can solve for, and start incorporating into the documentation system. We started looking at things like coding gaps—if you put more data in front of the doctor and care team, you start seeing behavioral changes. You start convincing doctors to document things in a more standardized way. Our general thinking has been that you start driving that behavioral change a tiny bit at a time. We can’t start trying to solve the entire problem from day one. Customers and partners—everyone wants everything on day zero, but making incremental steps towards population and then precision, is right. Also, for far too long, everyone has felt that the electronic health record is useless. Separating the data layer and the record layer, will make it easier to operationalize downstream.”
Further, Kalman asked the panel, “Once you’ve managed to source and curate your data, what happens next?” “We’ve encountered all the challenges you’ve encountered here,” Greenman said, gesturing to Rollison. “We’ve invested time and energy in the molecular data files. We’re bringing together results from genomic testing from different vendors, and aggregating those. But the file types and formats have been different. So we’ve created a modernized data lake, in which each of those lab sources is in one container, and we land the data points into the lake And we want to create better linkages between clinical and molecular data. We look at feasibility, trial feasibility, matching, quality accreditation; most importantly for patients, we can proactively patients with alterations or mutations who may respond favorably to proactive therapy.”
“Obviously, we work with multiple oncology providers, so we bring all that data together to be used in a pool fashion,” Shashank said. “Now, we’re taking our data and using it to generate data insights. Three or four years ago, we found that testing for biomarkers was woefully low—it was happening in less than 2 percent of cases. At the same time, when we curate data, we’re combining the clinical and the biomarker testing data, and are giving it back to our partners. And on an experimental level, we’re bringing in fully curated clinical data with genomic data and aggregating that together, and are actually sharing that with pharma companies, to explore why patients are responding to particular biomarkers.”
What about risk categorization? “Like Jennifer, we’re comingling our molecular data, clinical data from the EHR, cancer registry data, as well as information from our clinical trial management system, and other sources,” in order to stratify risk among cancer patients, Rollison reported. “This helps us to profile our population overall, and finding disparities. Are we doing a good job. And around clinical trials—triple-negative breast cancer, we know is a greater burden in the African-American community. Do we have targeted trials in that area? And in actual trial enrollment, are we seeing adequate proportions of under-represented minorities? Are transportation, child care, or other groups. Or are the co-morbidities prevalent in those communities knocking them out of the trials? So using precision data can help us look at the population and determine whether or not we’re serving the cancer patients of today with he portfolio of services that we provide.”
What are clinicians asking for? “All of the above,” Shashank said. “And coordination of care is an incredibly important thing. If you think about someone who’s going through cancer, it’s hard enough already. And then, navigating the oncology clinical care network is not easy for anyone. And apart from other use cases, coordination of care for an oncological patient, is one of the biggest things happening today. And Medicare is helping… And cancer patients generally need someone to help them navigate the system.”
And, what about applying value-based care principles in this area? Kalman asked. “The problem with applying the value-based care paradigm to cancer care is that you’re starting to put value on people’ lives; and the drugs are very hard to manage the cost of,” Wang said. And the challenge remains “Really understanding the clinical care. Payers don’t have much access to clinical information,” he noted.
“How are you trying to contribute to a better oncologist experience?” Kalman asked Rollinson. “I turn again to my digital colleagues in the audience,” she said, gesturing towards Moffitt colleagues who were in the room. “We have a number of initiatives underway to improve the interaction with the EHR; we use Cerner. Timeline. Understanding what treatment was given, and the sequence, relative to what treatment was given, is one big ask from the oncologists. Also, they often have to rummage through past records to find test results, which are stored as PDFs. And they want fewer clicks, so we want to provide higher-quality data downstream that doesn’t put the burden on the oncologists. Technologies that interact with Cerner that aren’t native to Cerner,” have to be corralled in ways that lessen the challenges of working with EHRs, and her informaticist colleagues are helping to lead the way in that regard, she emphasized.
The ViVE conference will continue through Wednesday.
