Renowned Oncologist & Cancer Data Researcher Walks You Through the Future of Big Data in Healthcare
As part of our special PMC series, Slone Partners presents the unique perspective of four leaders driving progress in the personalized medicine space.
Dr. Amy Abernethy is internationally recognized for accelerating cancer research through her distinctive outlook on data use in healthcare. In this exclusive interview, Amy outlines a path to solve the huge infrastructural challenge in healthcare data and discusses the importance of patient input in improving outcomes.
Slone Partners: When approaching the use and sharing of health data, what are the biggest obstacles going to be over the next 2-3 years? The next 5-10 years?
The next 2-3 years?
Dr. Abernethy: Our biggest challenge when approaching the use and sharing of health information is the immediate need for better organization of the patient data that resides in the electronic health record. Most of the discussions around data sharing focus on the need for better interoperability in order to transfer medical records – and supposedly to aggregate data. But, in reality, moving medical records doesn’t equate to amassing meaningful data sharing that can lead to new knowledge in the aggregate. In order for data to be used, it needs to be organized and analyzable. When I say “analyzable”, I mean that you can conduct computations with the data, derive insights about what is better when compared with something else, or match patient characteristics to requirements in an algorithm; useful, analyzable data looks like an organized spreadsheet where mathematical calculations can be conducted on the rows or columns. Meanwhile, over half of the important datapoints in the medical record are embedded deep in “unstructured documents” – e.g., doctors’ notes, radiology reports, or even the condolence card which might be the most reliable signal that a person has passed away. We must transfer the information from these documents into digitized datapoints that are analyzable. Efficient, scalable, tech-enabled methods for this are needed. And, as a side point, the solution isn’t to ask doctors to become data scribes; we need doctors focused on treating patients, which is an interpretive complex art – not a data task. So we need technology to solve this problem while letting doctors focus on what they were trained to do.
Once the data is cleaned up, the other major near-term challenge is being able to integrate datasets. Information needs to be linked from disparate data sources, and this is best done through unique patient-level identifiers. There isn’t a single common “Master Patient Index”, and data linkage itself is really hard, especially when trying to maintain privacy. Overall, these are huge infrastructural dilemmas that we are looking to solve at Flatiron Health.
The next 5-10 years?
Dr. Abernethy: The infrastructure challenge is the most immediate, but as big as the problem is, it’s only the beginning. Think about what happens after a highway is built — cars can then easily travel from one place to another. But now you need cars and people who know how to drive them. Once the system of data is in place (the highways), the next task is deep analysis (cars and drivers). In the next 5-10 years, we’ll need to learn how to do trustworthy analyses that everyone finds credible.
Right now there is an exciting conversation going on about the application of machine learning and identification of unknown patterns in the data in service of improving health. But let’s stop to think about that for a moment. Do you want your doctor to choose a new cancer treatment for you instead of the “tried and true” based upon computer algorithms? What would it take for you to trust your doctor in making that choice? This is the next big problem – how do we make sense of the data in a way that leads to credible evidence in support of choosing one treatment over another? We need experimental models that can be applied to the data, statistical tests, quality assurance, the ability to peer review, and solid clinical interpretation. We need a regulatory framework that supports all of this as well so that the FDA can provide guidance as to which treatments are safe and effective in this land of big data. The list goes on.
The exciting part is that we are farther along in getting over these obstacles than many people realize. At Flatiron Health, we have developed methods to generate analyzable data from electronic medical records for people with cancer, including all of those unstructured documents. We are working on new analytic frameworks to help us generate credible evidence. We are working with the FDA to make sure that the data, analytic approaches and quality assurance methods are appropriate. This is all happening now.
Slone Partners: What advice would you give a patient today that wants to collect and share their health data for the sake of research or to build a profile of family risk?
Dr. Abernethy: Patients have a limited understanding of what it means to share their health data. So for starters, they need to be better informed and educated about data sharing. One way to do that is to partner with a patient advocacy organization like Friends of Cancer Research (FOCR) or LUNGevity to gain more insight on what patient data sharing really means. This information will increase confidence on being able to speak on their own behalf. There’s an amazing program that Flatiron Health is a part of called Big Data for Patients (BD4P), organized by the Reagan-Udall Foundation, a nonprofit organization formed by the FDA. BD4P is a training program to teach patients and advocates about how data science can improve patient health. Not only do we need to ensure that patients understand the importance of data sharing, but it’s also important for them to be able to interact in this complex data space. We need to be thoughtful about how we make patient interaction doable.
Slone Partners: What should small start up data and app companies be aware of as they work to improve data sharing and enable patients to control their data?
Dr. Abernethy:These companies should first understand that data and data collection happens in context. It’s critical that they are aware of the big picture – knowing what else is going on in that person’s life at the point in time that the data is collected. App developers should be cognizant about matching data collected from the app with the rest of the healthcare context. For example, if the app is collecting information about symptoms; is the patient on a medicine that might be causing the symptoms and what time of the day is the patient taking the medication?
Another important element for app developers is the algorithms that are used to translate between data generated and information about personal health data. For example, if biosensors and their related apps are going to be used to generate credible healthcare data, we need transparency about how underlying algorithms translate the sensor information into a score and what that score means. For example, if we look at a “steps” monitoring app, we need to understand exactly how the number of steps you took measures up. Generally these biosensors aren’t true checking every time your foot hits the ground, but rather they translate the distance you cover into a count of number of steps. What if your stride is bigger than mine? What is the algorithm that calculates this? A company needs to disclose these algorithms in order to be credible.
In terms of enabling patients to control their data, I believe patients need the ability to pull their data back if they feel it’s being misused. If patients can’t do that, how are we going to develop a trusted system?
Slone Partners: What do you see as the biggest roadblock in cancer research today, and what can we do to change it?
Dr. Abernethy: There is so much excitement about how “big data” is going to help us advance cancer research. I am excited too! But, the biggest roadblock we face in cancer research today is the infrastructure problem I described above coupled with “magical thinking.” Those who don’t understand the infrastructure problem think that a simple set of solutions are going to fix the issue with a press of the button; the comment that I always love is “we just need interoperable electronic health records”. But there is no magic in this; we need to do the good old-fashioned work of sorting it out – we need to first build the highway. The infrastructure issue is a messy, overwhelming and an incredibly hard problem to solve, but it must be dealt with in order to move forward in meaningful ways. And then we have to be sure that we can analyze data in a responsible and credible way. And perhaps most importantly, we need the patient voice in this every step of the way.
I am really proud of where we are today. Strides are being made across the industry. The steps that we’re taking at Flatiron Health to organize, clean and process unstructured data will help overcome this obstacle. It is really fun to see how the analytics space is moving forward.
Slone Partners: In your opinion, what is the most pressing issue that must be resolved in data management for Personalized Medicine to reach its potential?
Dr. Abernethy: I’ve outlined above what needs to be solved in terms of infrastructure. Now let me share why it’s important for personalized medicine. Imagine that I’m a 48-yr old woman diagnosed with early-stage breast cancer, a family history of breast cancer, unknown genetic mutation status, and that I am a non-smoker who eats a low-fat diet, drinks alcohol occasionally, and has a notable past medical history for melanoma on my arm, and a family genetic history of neuroendocrine tumors. It is really hard for my doctor to keep up with all of those facts and use them to determine precisely the right treatment for me. That is where data can help. In order to personalize my care, all of those details can be rapidly transferred into digitized, analyzable data that can be matched with algorithms to provide a recommendation of the right treatment aligned with my personal characteristics. But I need my data cleaned up and standardized before any of that work can be done. And then, as we have more (and better!) information about diagnostic tests, biology, and important personal details, this information can be layered in to progressively personalize my treatments. In other words, getting the data right is the critical foundational step.
Slone Partners: How does Flatiron Health fit into all of this?
Dr. Abernethy: These are all of the problems that we are trying to solve at Flatiron Health. I am an oncologist and a clinical researcher. I came to Flatiron Health because I knew that in order to help move the needle, I need to get my hands dirty. Being able to harness data for our future means that we have to get the underlying infrastructure right. We need to take out the time for dataset preparation in order to be able to focus our time on analyses. And the analyses have to be accurate – and clinically actionable.
About Amy Abernethy
Amy P. Abernethy, M.D., Ph.D., is the Chief Medical Officer/Chief Scientific Officer and SVP Oncology at Flatiron Health, a healthcare technology company focused on organizing the world’s cancer data and making it actionable for providers, patients, researchers and life sciences. At Flatiron, Dr. Abernethy leads the Oncology and Science parts of the organization. She is a hematologist/oncologist and palliative medicine physician, and an internationally recognized cancer clinical researcher.
With over 400 publications, Dr. Abernethy is an expert in cancer outcomes research, clinical trials, patient reported outcomes, evaluation of healthcare quality, health services research, clinical informatics and patient-centered care. She is an appointee to the National Academy of Medicine’s (formerly the Institute of Medicine) National Cancer Policy Forum, on the Executive Board for the Personalized Medicine Coalition, and Past President of the American Academy of Hospice & Palliative Medicine.
Before joining Flatiron, Dr. Abernethy was Professor of Medicine at Duke University School of Medicine and ran the Center for Learning Health Care in the Duke Clinical Research Institute and Duke Cancer Care Research Program in the Duke Cancer Institute. For more than a decade, she has pioneered the development of technology platforms to spur novel advancements in cancer care, including the development of systems by which big data can support tracking cancer care, drug development, personalized medicine and scientific discovery. Joining Flatiron was the obvious next step.
Dr. Abernethy went to the University of Pennsylvania as an undergraduate and then medical school at Duke, where she also did her Internal Medicine residency, a year as Chief Resident, and her hematology/oncology fellowship. She has her Ph.D. from Flinders University in Australia, focused on evidence-based medicine. She is also on the Board of Directors of athenahealth, Inc.
In Slone Partners’ newsletter, we interview thought leaders from top companies to provide expert advice to leaders in the diagnostics, clinical trials, contract research, healthcare information technology, life science tools, and laboratory testing industries. Subscribe to our newsletters on our Home Page.