At long last, we seem to be on the threshold of departing the earliest phases of AI, defined by the always tedious “will AI replace doctors/drug developers/occupation X?” discussion, and are poised to enter the more considered conversation of “Where will AI be useful?” and “What are the key barriers to implementation?”
As I’ve watched this evolution in both drug discovery and medicine, I’ve come to appreciate that in addition to the many technical barriers often considered, there’s a critical conceptual barrier as well – the threat some AI-based approaches can pose to our “explanatory models” (a construct developed by physician-anthropologist Arthur Kleinman, and nicely explained by Dr. Namratha Kandulahere): our need to ground so much of our thinking in models that mechanistically connect tangible observation and outcome. In contrast, AI relates often imperceptible observations to outcome in a fashion that’s unapologetically oblivious to mechanism, which challenges physicians and drug developers by explicitly severing utility from foundational scientific understanding.
The recently-announced acquisition of the oncology data company Flatiron Health by Roche for $2.1B represents a robust validation of the much-discussed but infrequently-realized hypothesis that technology entrepreneurs who can turn health data into actionable insights can capture significant value for this accomplishment.
Four questions underlying this deal (a transaction first reported, as usual, by Chrissy Farr) are: (1) What is the Flatiron business model? (2) What makes Flatiron different from other health data companies? (3) Why did Roche pay so much for this asset? (4) What are the lessons other health tech companies might learn?
The Flatiron Business Model
To a first approximation, Flatiron has a model that can be seen as similar to tech platforms like Google and Facebook – delight (or at least offer a useful service to) front-end users, and then sell the data generated to other businesses. For Flatiron, the front-end users are oncologists (mostly community, some academic), and the data customers are pharma companies. In contrast to Google (and also in contrast to the less successful Practice Fusion, recently acquired at a loss), Flatiron doesn’t sell access to front-end users themselves (e.g. through targeted ads), but rather access to de-identified, aggregated clinical information.
Success of this model requires that the Flatiron platform is attractive to oncology practices, who must feel that they’re getting distinct value from it and believe that it helps them fulfill their primary mission of taking care of cancer patients. If this is true, then the Flatiron platform will enjoy continued traction from its current base, and may more easily win over new users (including practices that use a different EMR system, like Epic, but still want access to the Flatiron network and analytics).
Despite (some might say, because of) a raft of new biological methods, pharma R&D has struggled with its EROOM problem, the fact that the cost of successfully developing a new drug, including the cost of failures, has been relentlessly increasing, rather than decreasing, over time (EROOM is Moore spelled backwards, as in Moore’s Law, describing the rapid pace of technology improvement over time).
Given the impact of technology in so many other areas, the question many are now asking is whether technology could do its thing in pharma, and make drug development faster, cheaper, and better.
Many major pharmas believe the answer has to be yes, and have invested in some version of a by-now familiar data initiative aimed at aggregating and organizing internal data, supplementing this with available public data, and overlaying this with a set of analytical tools that will help the many data scientists these pharmas are urgently hiring to extract insights and accelerate research.
Increasingly, I’m convinced that the underappreciated challenges of implementation describe the ever-expanding gap between the promise of emerging technologies (sensors, AI) and their comparatively limited use in clinical care and pharmaceutical research. (Updated disclosure: I am now a VC, associated with a pharma company; views expressed, as always, are my own.)
Technology Promises Disruption Of Healthcare…
Let’s start with some context. Healthcare, it is universally agreed, is “broken,” and in particular, many of the advances and conveniences we now take for granted in virtually every other domain remain largely aspirational goals, or occasionally pilot initiatives, in medicine.
Healthcare is viewed by many as an ossified enterprise desperately in need of some disruption. As emerging technologies shook up other industries originally viewed as too hide-bound to ever change, there was in many quarters a profound hope that advances like the smart phone or AI, and approaches like agile development and design thinking, could reinvent the way care is delivered, and more generally, help to reconceptualize the way each of us think about health and disease.
Perhaps the normally measured physician-economist Aaron Carroll best captured the reaction and sentiments of the healthcare community in response to a recent JAMA article demonstrating that subjects in a weight reduction study using activity trackers lost significantly less weight than those in the control group:
“I TOLD YOU SO!!!!!!” (Emphasis in original.)
These results were cheered for several key reasons.
First, many in healthcare are irritated by the idea of simplistic technical fixes for complex medical (and social) (and cultural) (and economic) problems–like obesity.
Second, as Carroll has pointed out, exercise is healthy for many reasons, but weight loss is probably not one of them; changing your diet seems to matter a lot more.
However, it’s important to critically evaluate research even (especially) when it seems to produce an ego-syntonic conclusion–a conclusion with which we so strongly agree.
My initial reaction to the result was that perhaps it reflects an example of the concept of “moral licensing” that Malcolm Gladwell discusses so thoughtfully on his Revisionist History podcast–i.e., when you deliberately act morally in one context, you may be more likely to act less morally in another context, having already demonstrated to yourself your moral bona fides.
Lyons is the author of Disrupted, the buzzy new book about what happens when a curmudgeonly fifty-ish tech writer gets unceremoniously dumped from a plum role at Newsweek and takes a job as a “content generator” at Hubspot, a white-hot Boston startup selling marketing software.
Best known for creating a “Fake Steve Jobs” blog, and more recently for his work on the writing team for HBO’s achingly funny Silicon Valley, Lyons has a taste for the absurd, and his prologue (excerpt here)–describing his initial experience at Hubspot–is a laugh-out-loud takedown of tech startup culture.
The fun only lasts a few chapters, however (captured perfectly in this review by Erin Griffith), as Lyons hopes to convey a more serious point (conveniently summarized in an op-ed in today’s New York Times): that the excitement around technology companies is largely empty hype, enthusiasm used to sucker naïve young adults to work for peanuts (and candy), and to enrich savvy founders and venture capital investors, and the investment bankers who enable them, at the expense of the gullible mom and pop investors who buy shares of these fast-growing but often profitless companies after they go public.Continue reading…
“People hate pharma,” my Forbes colleague Matthew Herper observed recently–and at times I can understand why. There’s not much to admire about executives like Martin Shkreli, or businesses like Valeant.
But I’ve started to worry that the “pharma = evil” narrative has become so ingrained that it’s taking on a life of its own, as readers instinctively anticipate this storyline, and journalists reflexively provide it. Coverage of a recently announced innovative training collaboration between Johns Hopkins and MedImmune (a subsidiary of AstraZeneca), for instance, focused primarily on potential conflicts of interest.
(Disclosure/reminder: I work at DNAnexus, a health data management company in Silicon Valley and Boston; our partners include universities, government agencies and private companies.)
This narrow view, however, not only fails to capture the urgent need for effective, new therapies, it overlooks entirely the vital role played by companies in translating fragile but promising scientific ideas into robust medicines for patients.
The latest assessment is courtesy of an editorial in the New England Journal of Medicine (NEJM), written by editor-in-chief Jeff Drazen, along with Dan Longo.
Essentially, Longo and Drazen argue that while the Platonic ideal of rich data sharing is lovely, reality is not so pretty.
First, Longo and Drazen allege, researchers who weren’t involved in gathering the original data often lack essential appreciation for how it was gathered, and thus may misinterpret it, as they “may not understand the choices made in defining the parameters.”
Second–and this is really the heart of the issue–Longo and Drazen worry that a new class of research person will emerge—people who had nothing to do with the design and execution of the study but use another group’s data for their own ends, possibly stealing from the research productivity planned by the data gatherers, or even use the data to try to disprove what the original investigators had posited. There is concern among some front-line researchers that the system will be taken over by what some researchers have characterized as “research parasites.”
Can an impassioned band of savvy, battle-tested techno-optimists save our healthcare system from its worst instincts, and deliver at least a soupcon of real interoperability?
That’s the question I found myself seriously contemplating after last week’s JP Morgan Healthcare Conference in San Francisco, and in particular after an inspiring data science roundtable discussion led by Aneesh Chopra (the first Chief Technology Officer of the United States, now at Hunch Analytics) and Claudia Williams, a senior advisor in the White House Office of Science and Technology Policy (OSTP).
The FHIR Engine
The (not so) secret sauce in question is the FHIR (pronounced “fire”) interoperability standard that has emerged from the non-profit HL7 organization, and functions as a universal adaptor, allowing some types of clinically-relevant data types to be shared easily and securely. Remarkably, most major electronic health record (EHR) vendors – including athenahealthATHN +0.00%, CernerCERN +0.00%, and yes, Epic — have signed onto this concept, and agreed to support an early effort at implementation, known as the Argonaut Project.
The key to driving behavior change, a seasoned marketing executive turned digital health investor told a panel on patient engagement that I moderated this week, is to get beyond the demographics of customers, and to understand the “why” – what are their distinct motivations and drivers?
Customers with similar demographic characteristics might be motivated in very distinct ways, he explained; sophisticated, quantitative market research can help define the different “personalities” present in a particular market.
Healthcare businesses, he emphasized, need to recognize these differences, and customize their approaches based on this nuanced understanding.
On the one hand, it occurred to me he was describing the behavioral component of precision medicine; in the same way it’s important to match an oncology drug with the right biochemical pathway, it’s also essential to customize the motivational approach to the characteristics of each individual.
On the other hand, I realized there was something that seemed a little sad about the idea of developing extensive market analytics and fancy digital engagement tools to simulate what the best doctors have done for years – deeply know their patients and suggest treatments informed by this understanding.
Instead, it seems, we’ve slashed the time physicians get to spend with patients, protocolized and algorithmitized almost every moment of this brief encounter, and insisted the balance of time is used for point-and-click data entry and perhaps a rushed dictation. We’ve industrialized the physician-doctor encounter – the process and the paperwork — but eviscerated the human relationship; it’s value, unable to translate easily to an excel spreadsheet, was discounted and dismissed.
As I look at the extensive analytic efforts to categorize patients, and the many digital health platforms designed to motivate behavior, it’s hard not to ask whether we’re painfully trying — at scale but without heart — to re-create something we might have been better off not destroying in the first place.
David Shaywitz is based in Mountain View, California. He is Chief Medical Officer at DNAnexus, a Mountain View based company and holds an adjunct appointment, Visiting Scientist, in the Department of Biomedical Informatics at Harvard Medical School.