- Speech by
Amy Abernethy, MD, PhD.
Principal Deputy Commissioner - Office of the Commissioner
(Remarks as prepared for delivery)
Thank you. It’s an honor to be here at the ONC’s 3rd Interoperability Forum. The FDA supports the smart, secure, and safe interaction among different medical devices and information systems, and the agency has been collaborating with hospitals, health care providers, manufacturers, standards development organizations, and other interested parties to promote medical device interoperability.
Needless to say, the exchange and use of information looms large over my work at the FDA in my roles both as the Principal Deputy Commissioner and the Acting CIO.
But before I talk about FDA, let me share a bit about how I got there.
My background is as an oncologist. I was on the faculty at Duke where I was a practicing clinician as well as a professor of medicine. At Duke, I ran an organization called the Center for Learning Healthcare.
At the Center, we focused on this question: How do we build data and technology solutions to simultaneously improve day-to-day clinical care delivery while also conducting research in the clinical environment?
Our group studied how to organize the data, we studied point-of-care data collection solutions embedded in the EHR or a patient’s tablet computer, and we studied the quality of data routinely collected in the clinic and methods to improve data quality.
Other questions focused on harnessing these datasets for clinical trials, participant recruitment, new statistical analyses and healthcare quality monitoring.
The patient was always the focus, and we organized datasets anchored by longitudinal patient journeys and patient-defined outcomes.
The summary goal was to pressure test the components needed to build a patient-centered learning healthcare system (take note of this, because it will come back up in our story later).
But we hit not one, but two, ceilings.
The first was related to our inability to aggregate data across institutional boundaries; it was a problem of interoperability. I could transition data in and out of the Duke data warehouse, develop solutions to structure it and clean it up, but one institution is only one institution.
The second was a ceiling on our ability to recruit and retain key talent such as software engineering and product managers.
As I wondered how to break these ceilings and create a learning healthcare system, I realized that we needed to get some unconventional partners involved – the tech industry and even venture capital. To really jump-start this, I left the ivory tower of academia and joined a tech start-up.
In the start-up, we built software for oncologists and patients -- for example an electronic health record – and then pulled data out of those software systems into a common data model.
Importantly, by building our own EHR we learned how to create workflow, data integration and other solutions to improve the lives of doctors working with the EHR as well as the quality of data that could be generated from the EHR.
We also built solutions that transcended any one particular EHR so that we could aggregate data from any system into a common data model ready for research-quality analyses.
A key goal was to clean the data - post-processing- and make it ready for secondary purposes like healthcare quality monitoring and—especially—clinical research. We were recapitulating the learning healthcare model that I worked on at Duke.
While there, I learned the in’s and out’s of venture funding, business development, agile software product design, data lakes and pipelines, and advanced analytics. We were proudly patient centric.
And, what was interesting about that experience was the importance of the FDA.
Regulations set the guideposts within which industry works. By clarifying what to do – and what not to do, regulations kept us from getting distracted and focused our energy on that which would be most impactful. What I learned was that the FDA both set the direction and clarified how to get there—particularly for the development of medical products like drugs and devices including technology-based solutions like software as a medical device.
ONC has a similar responsibility to develop policy and requirements— the guideposts— for health information. And ONC is charting a path toward interoperability.
Clear regulatory requirements form the guideposts within which industry works. In my private-sector experience, these guideposts kept us from getting distracted and focused our energy on that which would be most impactful. What I learned when I was in industry was that government – including FDA, ONC, CMS, sets the direction and clarifies how to get there.
If we want to maximize impact on public health, setting the vision and clear guideposts is a solid path. Joining the FDA made perfect sense to me. And, as you can imagine, I jumped in head first!
As Principal Deputy Commissioner, I am focused on a few core components. First, I want FDA to get our own technical house in order so that tech can “snap in” – we can be agile and efficient. We need to be able to have common interfaces with industry so we can pass data between our organizations, have collaborative review, etc. In the same way that ONC is establishing a vision for EHRs that clarifies the interfaces – the APIs – we need to do the same for FDA.
Second, we should use examples and use cases that allow us to see what it looks like when we put tech and interoperable data into play. Examples start to show our early steps, such as precision FDA and real time oncology review, which I will describe in more detail in a few moments.
And my third area of focus is to update how we interface with the technology community in a meaningful way. As a biomedical community and FDA, we have clear ways of interfacing with the drug and device industries, with patient groups, with trade organizations. But tech companies are largely seen as vendors being told what to do rather than creating critical solutions with collaboration and guidance.
We envision what we want from a single standpoint but don’t directly engage with those who will fund and build those books of work. We need to all move together, with better and more direct collaboration on what the data should look like, and what it can be used for. We need move the entire space forward by including tech at the table from the beginning.
Where do we go from here? Our longstanding goal for medical care is to enable the evolution of precision medicine by ensuring that physicians and patients have the information they need to match the right FDA-approved or cleared treatment or diagnostic to patients based on their needs and preferences.
Many of these opportunities to personalize patient health and medical decisions are being enabled by new technology platforms such as AI-enabled devices, targeted medicines, and regenerative medicine, including cell and gene therapies. These new technologies offer transformative opportunities that can make the health care system not only more effective, but also more efficient and patient-centered.
But all the participants in this medical product ecosystem, including FDA, have a challenge: How do we get there?
FDA is taking a thoughtful and proactive approach to help advance this story, including the appropriate (and scaled to fit) use of technology with a modernized review infrastructure – this is related to my second hat, as Acting CIO.
And we’re building a dynamic regulatory environment across the entire cycle of product development to help us meet the pace of medical innovation.
Let’s walk through a couple of examples:
First, let’s talk about the review process for a new drug. Our Oncology Center of Excellence is pressure-testing approaches to accelerate this process. In the Real Time Oncology Review (RTOR) pilot, a sponsor seeking approval of a supplement to an existing approved application can start sending data—pre-submission. This allows FDA to review—in real time—the pre-submitted data and assess it for sufficiency and integrity—and give the applicant early feedback.
By the time the sponsor submits the supplemental application with the FDA, the agency’s review team would already be very familiar with the data and the analysis.
This way, the FDA review team will be in a better position to conduct a more efficient, timelier, and thorough review. And sponsors will have benefited from early feedback on how to best analyze data to effectively evaluate key regulatory questions.
Our first submission reviewed under the pilot came a full three months before its PDUFA goal data, and just five months after it was submitted. But, this was still a static review package and dataset—the dataset was submitted the old-fashioned way.
As we start to clarify the interfaces for the FDA, we anticipate a landscape where the data are submitted via an API, the FDA reviewers can exchange messages with sponsors in real-time, getting quick answers to questions, and elements of the review package and response letters are generated in electronic environments.
This is a story of data interoperability – but as you can also see, it is FDA centric where most of the data are envisioned as coming from traditional clinical trial mechanisms.
Now let’s take this story a step further. In the context of real-world data, data may be coming from a variety of sources—including the EHR.
Data quality is key. A note about data quality—it is fundamental in the HER context. Quality will need to be continuously characterized and improved over time.
Traceability back to source allows for the ability to crosscheck, workflow solutions.
And, as the third example of where interoperability comes into play, and especially EHR interoperability, let’s talk about surveillance of medical products across time determining when something should be recalled or a product label should be adjusted.
In order to support effective and timely recalls of medical devices when needed, FDA has promoted several collaborative efforts to develop a common understanding and approach to adoption of a unique device identification system in the health care setting.
In addition, collaborations with organizations including the CDC and the National Library of Medicine helped lead to a final guidance intended to help manufacturers design and develop safe, effective, and interoperable medical devices by outlining important design considerations and providing clarity on the agency’s recommendations for submitting interoperability-related information in premarket submissions and labeling. We are proud to see UDI as a part of the ONC proposed rule.
In summary – it wouldn’t surprise you to hear that I believe that interoperability of all types including EHR interoperability is fundamental to a learning health system, and that FDA has a critical role to play in this. But it isn’t just FDA – it is FDA in partnership with our intergovernmental colleagues and all of you.
One way that FDA and ONC might collaborate is through partnered dialogue with stakeholders such as tech companies, manufacturers and other government agencies.
Another way we can work together is to pressure-test the “what happens if” scenarios to inform our guidances and industry.
If we continue with this dream, the engagement of FDA, ONC, and sister agencies could provide the incentives, if you will, to complement ONC’s policy development in the interoperability space. I believe this might provide an actionable roadmap for AI ready datasets, with a clear financial incentive for investing transparency and reproducibility.
Additional datasets could be added over time for training and validation and the AI and clinical communities could be able to provide real time feedback to FDA, provide additional verification of results, and identify opportunities to translate “lessons learned” to other use cases and public health applications.
I’d like to end with a bit of a teaser. In the next month or two, we will be rolling out a plan to modernize FDA’s approach to the use of technology for its regulatory mission, including the review of medical product applications.
It will include modernization of FDA infrastructure to make sure we can support emerging uses of artificial intelligence, blockchain, and other technologies. We will be ramping up activities that modernize how we use tech to work with the stakeholder community.
We will build on our early activities, such as real-time oncology review, and will be looking at ways to advance this interoperability story in service of a learning health system.
So, stay tuned in the coming months to hear more about this exciting action plan!
And with that, I am a happy to answer your questions.