The Early Bird Gets the CFO
This has been an exciting time for the field of artificial intelligence (AI) and data-informatics in the clinical healthcare space. We’ve seen bullish market signals in this sector with acquisitions like Flatiron Health by Roche, but also stern warning signs with challenges facing IBM Watson at the same time. Needless to say, pharma and biotech, and even the FDA, are now getting serious about applying innovative approaches to drug development leveraging data and analytics. Why now? To help us understand the dynamics behind this phenomenon and to predict what’s still to come, Chasm Partners brought together a panel of three recognized CEOs directly involved in the race to help life sciences, payers, and providers achieve clinical transformation through data and analytics.
Matt Dumas – Founder and Managing Partner
Colin Hill is CEO and Co-Founder of GNS Healthcare. He is a leading voice in healthcare technology and precision medicine and brings impressive leadership experience in commercializing machine learning technologies in the biopharmaceutical and managed care industries. Colin co-founded GNS Healthcare in 2000 and has since served as Chairman & CEO.
Miki Kapoor serves as CEO of Verana Health, a data and technology company building a cutting-edge platform for life science innovation that utilizes regulatory-grade specialty data sets, initially featuring real-world data from the largest specialty clinical database in medicine. Miki has deep expertise in scaling health data and software-as-a-service businesses through building high-performing management teams. Miki serves on the board of several healthcare organizations and has taught on the topic of private investment into public health at two universities.
Ken Tarkoff is CEO of Syapse, which partners with health systems to implement precision medicine programs across their organizations, enabling oncologists to deliver the best precision care to every patient who needs it. Ken is a health technology veteran from RelayHealth Intelligence, a data and analytics business owned by McKesson, with deep experience in facilitating physician, health system and health plan relationships.
With the recent acquisitions of major data and analytics players like Flatiron Health to Roche, and SHYFT to Medidata, what is the market telling us?
TARKOFF In healthcare, and in precision medicine more specifically, it is becoming clear that companies operating in silos will not help us reach the end goal of improving care for patients. That is why we are seeing companies across the healthcare ecosystem join forces and recognize the need to work and evolve together. Making targeted treatments available to more patients requires a broad range of capabilities. Those health systems, Life Sciences companies, payers, labs and other players who are willing to collaborate, and in some cases formally join forces, to combine strengths will likely be in better position to lead the efforts. We believe we will see more of this trend of broader ecosystem partners coming together in the near future.
HILL Building off of what Ken said, I think the market is telling us that it believes in the ability of data and technology to transform healthcare by generating insights to drive precision medicine and ultimately improve outcomes. You are seeing more consolidation as large organizations look to make real, substantial investments in better, broader data assets, infrastructure and technology. This push into big data is becoming part of the core strategy of many healthcare companies and mirrors what other industries have been doing for years.
KAPOOR First, these acquisitions are clear indicators that deeper healthcare data than historically available truly matters. Second, these acquisitions establish data rights as table stakes. While data rights matter, there are several rights-holders on the exact same healthcare data element in this country, including patients, physicians, specialty societies, EMR companies, payers, and self-insured employers. In other words, Roche and Medidata, like many companies in healthcare, understand that rights to a rich data source is just a requirement to have a seat at the table today. Data aggregation, on the other hand, is the new axis of competition: organizations that can aggregate massive volumes of data, especially those that can incorporate the official data set of patient care – EMR data – have a disproportionate ability to disrupt the market and derive deeper and more unique insights than has been possible in the last several decades.
Over the past few years we have been finally seeing Life Sciences companies, particularly in biotech and personalized medicine, embrace more progressive approaches to big data and artificial intelligence. What has sparked this demand? Why now?
HILL There is now unprecedented access to fast, high throughput data, from molecular to large scale EMR data which is enabling our ability to investigate patients’ response to drugs both in clinical trials and the real world. There is also increasing pressure on Life Sciences to prove the value of their drugs to health plans. By applying artificial intelligence to big data, we can now discover what works for whom, which is essential to not only Life Sciences, but patients, health plans and Providers as well.
TARKOFF As more targeted therapies and diagnostic tests are administered in the real-world setting, Life Sciences companies are recognizing a significant opportunity to apply real-world evidence across their organizations. Whether it involves designing an outcomes-based payment model or understanding the clinical utility and value of a particular drug, harnessing data will be a core differentiator for Life Sciences companies. Over the past few years, this work has been accelerating in Oncology, where we have made huge strides in ingesting and integrating clinical, molecular, treatment, and outcomes data on a massive scale.
KAPOOR I view Life Sciences companies as inherently evidence-based entities that rely on data for clinical and business decision-making. The recent and growing interest of these companies in big data has been enabled by regulatory and policy measures that have supported the adoption of technology and real world data in healthcare. The HITECH Act of 2009 incentivized the adoption of EHRs, moving patient records onto technology-based platforms. With rapid advancements in data analytics and artificial intelligence, these massive data sets can now be used to generate deep data insights, something that was not possible just a few years ago. The FDA has also begun to recognize the value of real world data and real world evidence when monitoring post-market safety and making regulatory decisions.
In your conversations with your customers and partners, what are their unmet needs that you see “deep data” and AI today having not yet addressed, but will eventually get to as this space evolves in the next 3-5 years?
KAPOOR Market demand today is trending towards predictive analytic capabilities and personalized medicine, which enable us to use existing information on patient health to understand and address future outcomes. Google is one company that has made some progress on this front with the development of an AI algorithm that can predict heart disease by examining a patient’s retina. As AI models continue to advance over the next few years, the real barrier to progress will be data silos. Verana is already making progress on this front by unifying discrete specialty data sets through its multi-specialty data platform, and this is an area we intend to heavily invest in over the coming years.
TARKOFF We see improving interoperability and data sharing as a primary challenge in cancer care that we need to address — it is still common practice for test results to be faxed or downloaded in such disparate formats that oncologists have a hard time using all the data we could tap into. Oncologists consistently mention to me the burden of accessing clinical data that could help patients get the treatment they need. We are making progress, but it is time for oncologists to have a standardized, portable way to access test data and make the most informed treatment decisions.
HILL It’s our view that there are many organizations that are still mining a very narrow set of data. But patients exist in the real world and are impacted by so many factors beyond biology. I think that as this understanding progresses, we’ll see more incorporation of data sources outside the traditional health sources, such as socio-economic, geographic, mobile, behavioral and real-time data. And the more data we can transform into insights the better we can personalize care.
With all of this excitement in AI, some industry players like IBM Watson have run into challenges. What lessons can we learn from this?
HILL From the technology point of view, I think one lesson learned is that our current understanding of disease and human biology is very limited. We need to be careful on relying too heavily on what’s currently known and available in the literature. Instead we need to continually drive towards discovering new insights without relying on our own hypothesis or existing knowledge. The other lesson learned is that we need to recognize that the platform and the data is a means to an end and we need to focus on solving specific problems. Medicine is a field rooted in scientific and clinical validation because of the gravity of dealing with human health, and that needs to be recognized and appreciated.
TARKOFF Health systems are starting to tell us they need more than just technology; they need strategic guidance to launch and scale an effective precision medicine program. They also want the most complete picture possible through data; whether a company uses AI or not, incomplete or inaccurate data will not give them the best treatment recommendations. Clinicians want clear, trusted treatment insights, delivered at the point of care, to guide their care decisions. Again, the lesson all of us in this industry are learning is that we need to collaborate across many parts of the healthcare ecosystem to tackle data completeness and quality issues before we can enable providers to deliver the best care to their patients.
KAPOOR Adding to what Colin and Ken have shared, a key lesson we have learned from large-scale AI projects such as Watson Health is the difficulty in implementing analytics on free-form clinical text in a fully automated fashion. At Verana, we focus on a joint approach that includes both human and programmatic components to train and continuously improve our models. We have also learned how difficult it is to generalize the application of artificial intelligence in a way that is meaningful across different contexts. To this end, we have built a team of data scientists with deep domain expertise in ophthalmology to train these AI models with our massive data sets. AI models rely on the deep learning that comes from providing a massive amount of data to each new situation. Ultimately, we know AI and machine learning are only as powerful as the human intelligence that fuels it, and we invest heavily in having the right people in place to ensure our AI models are extracting the most valuable insights possible for our stakeholders.