Steve Burnell: Digital health refers to a broad range of technologies that make health information more accessible beyond traditional care settings. That includes telehealth, mHealth, consumer wearables, self-monitoring medical devices, digital diagnostics, digital therapeutics and clinical decision support solutions, as well as the advanced algorithms and artificial intelligence systems that are increasingly capable of supporting those solutions. At its core, digital health represents a much-needed connection between data science and healthcare, one that more comprehensively connects the dots for better, more informed health decisions and more personalised patient care.
Danelle Miller: The digital health applications Steve just described have already begun to transform healthcare delivery, and yet, the true potential of digital health goes much further than the sum of its individual parts. It’s revolutionary, really, because digital health can harness the power of connectivity and interoperability in ways medicine has never seen before.
Steve Burnell: Yes, by applying the innovation in digital technologies that’s happening, we have the opportunity to break down the information and workflow silos that exist in healthcare today. By filling in blanks between diagnostic, treatment and outcomes data, digital health allows people to better manage their own health and become more actively involved in their own healthcare. It also allows physicians to better manage care at the individual patient level rather than at a disease level, which enables more efficient and effective care.
Steve Burnell: Established regulatory frameworks – such as those for medicines, diagnostics and devices – have been around a long time. The development timeframe for most products like these is literally measured in years, sometimes even decades. The thing about digital innovation is that it operates at a much faster pace. Current regulations do a very good job of ensuring safety and effectiveness, but they are not optimal for the fast-paced innovation we see in software and digital health today. To bring safe, effective digital technologies into healthcare at a pace that matches the speed of what’s possible – and that patients deserve – we need regulatory frameworks that accommodate the shorter timelines and unique agility of software development.
Danelle Miller: When we’re talking about software – particularly Software as a Medical Device (SaMD), which is software that is intended to serve a medical purpose and performs that purpose without being part of a hardware device – we see a wide range of medical functions.1 The risks associated with SaMD technologies can vary significantly because the category is so broad, so a risk-based approach is and always will be foundational for any good regulatory paradigm. However, we also need to reimagine software regulation in ways that specifically align with the highly iterative and adaptable nature of its development.
Danelle Miller: The International Medical Device Regulators Forum (IMDRF) is a group of regulators from around the world who co-develop “global guidelines” on how to regulate certain things. The purpose of IMDRF’s work is to accelerate regulating authorities toward international regulatory harmonisation and convergence in areas that IMDRF sees as important. One such area is SaMD, and IMDRF has developed guidelines that can be adopted by regulatory authorities as a foundation for software regulation. In the U.S., the Food and Drug Administration (FDA) is leading the way toward adapting the IMDRF SaMD guidelines to the U.S. regulatory paradigm, through a program in which the FDA will pre-certify software development companies that demonstrate a verifiable commitment to excellence.2 This pre-certification has potential to offer a more tailored approach to software regulation, including focusing review on higher-risk software and a streamlined regulatory review pathway for software that needs FDA review. Ultimately, this will accelerate healthcare provider and patient access to digital health products that can safely improve health decisions and health outcomes.
Danelle Miller: It’s very innovative because it is testing a new and different approach to regulating software, one that objectively assesses excellence within an organisation that is developing medical software. The Pre-Certification program is in pilot phase right now. Roche is honoured to be one of nine companies chosen by the FDA to participate. We have been working collaboratively with the FDA and other pilot participants, industry organisations and other stakeholders on solutions that will improve patient and provider access to new, safe and effective digital health technologies.
Steve Burnell: From the perspective of a software developer, I want to commend the FDA for showing tremendous foresight and understanding that digital health requires a new regulatory paradigm. The pre-certification process is potentially game changing in terms of how quickly digital innovation will be able to reach patients and physicians. The pre-certification model aims to maintain standards of safety and effectiveness, but it is also testing the efficacy of a process that takes the emphasis off individual products, and instead places it very heavily at the organisational level. That allows software innovation to really keep pace with technology, with what we're learning in medicine, and with what we are learning from the vast amount of aggregated data that today is largely unused in healthcare. In this regard, the FDA approach is very novel and much welcomed. It does not lower the standards of regulation; rather, it sets an alternative pathway that better fits software.
Steve Burnell: Post-market data will be critical to any model of software-specific regulation, just as post-market data is critical to all good software development. Good software constantly monitors how it is used in practice, and one of the many benefits of digital solutions is that they allow a level of feedback and learning from the use of the product that regular solutions often don't allow. Roche is a leader in using real world data and evidence, certainly in the design of our medicines and diagnostics, and we intend to remain a leader in how we use feedback and real world data to inform continuous improvement of the software itself. What we’re really talking about here is the application of healthcare to real patients, so everything we learn has the potential to inform both the quality of future care and how quickly it can be delivered.
Steve Burnell: Healthcare generates vast amounts of data. Frequently today, that information is just siloed and archived, and it rarely gets used to inform future care. One of the biggest opportunities for digital health comes in analysing those vast, real world data sets. Such analyses are certainly enabled by data science – the ability to understand and interpret the data sets, but also by technologies like artificial intelligence (AI) and machine learning to filter information, fill knowledge gaps and use data-driven insights that can allow healthcare professionals to make more confident decisions at the point of care.
Steve Burnell: Current and future breakthroughs in AI, machine learning, and clinical decision support solutions have huge potential to change how healthcare is delivered, as well as the quality and cost of care. For example, today machine learned algorithms can deploy decision trees and suggest appropriate clinical trials for cancer patients. Algorithms can sift through countless publications almost instantaneously to find the most recent evidence-based literature relevant to a specific case – or aggregated data from large patient populations – to interpret therapeutic options and outcomes, which in turn supports treatment decisions or helps determine the risk of readmission for heart failure patients, for example. Machine learning, and deep learning models in particular, can also analyse images in radiology and pathology. Large data sets of imaging data in particular contain a vast amount of information that computers can interpret with an accuracy and speed that the human brain, no matter how well trained, simply cannot match. There are many worthwhile applications for deep learning and machine learning in healthcare, and I think we're only just touching the surface today.
Speaking on behalf of Roche, one thing that is central to our position on the use of AI in medicine is that AI and machine learning are meant to support human decisions. Ultimately people make decisions, and clinical decision support systems, whether they are AI empowered or not, support those decisions. We see AI in healthcare as a journey, and one of the biggest challenges is understanding how and when it is appropriate to introduce AI into clinical care.
Sorry to go on, but I'm passionate about this topic. The quality of the machine learned model depends on the quality of the data that it is trained on. Over time, as we get increasing amounts of well-structured healthcare data that include quantifiable measures of outcomes, we will see machine learning and AI models move further into the predictive realm – for example, predicting how an individual patient is likely to respond to a given therapy. We view that as something not to be rushed, but something to be carefully planned, carefully developed, and carefully validated for use only where the data can support it.
Danelle Miller: I love how Steve has stated that digital health in general, and clinical decision support in particular, involves software capable of connecting dots. When you think about how AI can evolve through continuous learning even to the point of making predictions, it’s easy to see how it presents a clear challenge for regulators who are looking at software to see what it does right now. There is only one health authority we know of so far that has issued a guideline on regulating AI – that’s Korea. China is looking at it, but they haven’t issued any guidance yet. In the U.S., the FDA seems likely to include AI within its pre-certification program, and frankly, that's where we think it should be. AI can be evaluated in part by looking at the processes of the developers, including how they will validate the software and how they will track performance specific to the changes the software is inherently making as it learns over time through its own data analyses. There are opportunities for regulators to look at this in a completely different way, such as taking a pre-certification approach, to enable the innovation that AI can bring while supporting safety and effectiveness.
Steve Burnell: Roche’s strategic vision for diagnostics information solutions and digital health innovation are closely tied – as an organisation we really are focused on doing now what patients need next. We are passionate about driving the next generation of personalised healthcare, and we believe digital health innovation is absolutely central to that effort. As digital health solutions connect more and more dots in healthcare, we believe those connections will create an ever-expanding ecosystem supporting more informed health decisions and improved patient outcomes.
Danelle Miller: We look forward to continuing and expanding our partnerships with regulatory authorities. Together, we can work to ensure the software products of tomorrow are regulated under a framework that allows rapid access to new technologies that give patients what they need now, and as Steve said, what they need next. Any new regulatory paradigm should be built from a foundational understanding that the software is going to be used with patients, or with healthcare providers who are diagnosing and treating patients. That’s where our focus is as well, on each and every patient. They are the epicentre of everything we do.
Steve Burnell, PhD, is Vice President and Head of Strategy for Diagnostics Information Solutions at Roche. With his strategic focus and vision for the intersection of technology and medicine, Steve has considerable experience leading and growing new, innovative healthcare businesses.
Danelle Miller, JD, is Vice President of Global Regulatory Policy and Intelligence at Roche Diagnostics. She is responsible for guiding Roche Diagnostics globally on all aspects of medical device regulation related to development, manufacturing, marketing and sales of the company’s products.
References
IMDRF SaMD Working Group. “Software as a Medical Device (SaMD): Key Definitions” (2013). Available at:
U.S. Food and Drug Administration. “Digital Health Software Precertification (Pre-Cert) Program” [Online]. Available at:
This website contains information on products which is targeted to a wide range of audiences and could contain product details or information otherwise not accessible or valid in your country. Please be aware that we do not take any responsibility for accessing such information which may not comply with any legal process, regulation, registration or usage in the country of your origin.