A leading scientist once claimed that, with the relevant data and a large enough computer, he could “compute the organism” – meaning completely describe its anatomy, physiology, and behavior. Another legendary researcher asserted that, following capture of the relevant data, “we will know what it is to be human.” The breathless excitement of Sydney Brenner and Walter Gilbert —voiced more than a decade ago and captured by the skeptical Harvard geneticist Richard Lewontin – was sparked by the sequencing of the human genome. Its echoes can be heard in the bold promises made for digital health today.
The human genome project, while an extraordinary technological accomplishment, has not translated easily into improved medicine nor unleashed a torrent of new cures. Perhaps the most successful “genomics” company, Millennium Pharmaceuticals, achieved lasting success not by virtue of the molecular cures they organically discovered, but by the more traditional pipeline they shrewdly acquired (notably via the purchase of LeukoSite, which ultimately yielded Campath and Velcade).
The enduring lesson of the genomics frenzy was succinctly captured by Brown and Goldstein, when they observed, “a gene sequence is not a drug.”
Flash forward to today: technologists, investors, providers, and policy makers all exalt the potential of digital health . Like genomics, the big idea – or leap of faith — is that through the more complete collection and analysis of data, we’ll be able to essentially “compute” healthcare – to the point, some envision, where computers will become the care providers, and doctors will at best be customer service personnel, like the attendants at PepBoys, interfacing with libraries of software driven algorithms.
A measure of humility is in order. Just as a gene sequence is not a drug, information is not a cure. Getting there will take patience, persistence, money and aligned interests. The most successful innovators in digital health will see the promise of the technology, but also accept, embrace, and ideally leverage the ambiguity of disease, the variability of patients, and the complexities of clinical care.
We’ll also need to incorporate four key lessons of the genetics experience:
- Don’t confuse data with insight: it can be difficult to extract robust, clinically-relevant conclusions from reams of data;
- Don’t confuse insight with value: many solid scientific findings, while interesting, do little to inform existing practice or significantly improve today’s outcomes;
- Don’t overestimate your ability to forecast from data: even the best data often afford only limited insight into health outcomes; a lot may depend upon chance or other factors;
- Don’t underestimate the implementation challenges: leveraging data successfully requires a care delivery system prepared to embrace new methodologies, requiring significant investment of time and capital, and the alignment of economic interests.
Digital health will ultimately revolutionize medicine, but it will get there through a series of evolutionary phases. These won’t be tidily sequential – some disease areas and some delivery systems may offer more fertile ground initially and see early successes.
But for the healthcare experienced by the vast majority of providers and patients and influencing a meaningful share of the dollars spent, the process will take much longer. Ten to fifteen year adoption cycles are typical (even rapid) in healthcare, and digital health is well advanced in only one domain (digital capture of data in today’s workflows through electronic medical records [EMRs] and digital diagnostics) of the many required for far-reaching impact.
As a whole, we anticipate the revolution happening in three broad phases:
Phase 1: Consolidating gains so far
The ongoing digitization of today’s clinical information will – over time – create a clearer understanding of how medicine is currently practiced; it will provide a coherent longitudinal record of each patient’s health and enable codification of existing treatment practices.
True, most EMRs are still piecemeal, incomplete, and inaccurate – and resisted by many physicians. But the short-term benefits – simple, descriptive analytics to evaluate basic measures of quality, such as whether specific patients are getting necessary treatments, or whether interacting drugs have inadvertently been prescribed – are attractive. Many companies (from startups to behemoths) are pursuing data consolidation — to make records increasingly complete and “analyzable” — as well as clinical decision support – to expose and harvest the obvious departures from acceptable clinical practice.
Nor do the opportunities to exploit digitally captured data have to be strictly clinical: addressable unforced errors may be as basic as hospital supply chain optimization – an area where best practice is already well understood outside of health care. Not surprisingly, the “internet of things” is dramatically revealing how far health system logistics often are from best practice.
A key challenge that healthcare will need to address in this phase might be termed “algorithm creep,” the tendency of well-intentioned “best practice” initiatives to become artificially precise, recommending and often mandating a specific course of action when a range of alternative approaches may be equally reasonable.
Insisting on a uniform approach is instantly recognized by patients and providers as artificial, and imposes a dangerously fragilizing solution that is neither required nor ultimately beneficial. Healthcare will need to reign in the algorithm mania of some hospital executives who think more like capacity managers than physicians, a problem likely to only get worse as the industry consolidates. Rather than aggressively force providers into contrived boxes, health system leaders would do well to consider the “via negativa” strategy discussed by Taleb, and focus more on eliminating the clearly wrong than on dogmatically designating a single approach as definitively right.
Working out the right balance between computer driven and physician insight represents an ongoing challenge healthcare must wrestle with as it seeks to fully exploit the multiplying new data streams from digital health. Approaches to care improvement that emphasize continuous evolution and iterative improvement, and are explicitly designed around active physician participation – such as the SCAMPs program at Children’s Hospital in Boston – or ones which (like UpToDate) offer such compelling value that even computer-phobic physicians embrace them, point to the path forward.
Another major challenge will be getting economic incentives aligned: digital health will ideally take aim at unnecessary, potentially harmful procedures, meaning fewer dollars for those who perform them. Success requires the continued transition to a reimbursement system that rewards quality rather than volume – an incentive structure that is arguably taking shape for larger delivery systems, perhaps less so for the remaining small, unaffiliated practices. As digital health places new demands on providers to invest in new capabilities and adapt practice patterns which may turn procedure-focused revenue-generating stars into occasional consulting specialists, widespread buy-in is likely to occur only after the transition to meaningful quality-based reimbursement reaches critical mass.
Phase 2: Learning to Drink From The Digital Firehose
After digitizing today’s data flows, the next phase of digital health will drive towards dramatically denser physiological measurement exploiting winning models coming out of today’s explosive investments in health data acquisition approaches, especially “intelligent sensors.”
The idea itself is compelling: right now, physicians have very limited visibility into the health of most patients – care is episodic, and there’s usually not much solid data on what a patient is doing in the real world; granular physiological data would seem helpful.
Yet turning more measurement into better health will require significant advances in data analysis and care delivery.
On the data front, existing and emerging predictive analytic platforms will help distinguish signal from noise, and ultimately deliver scientifically significant insights, a small fraction of which will actually have the potential to improve patient care in a meaningful way. This journey from raw data to demonstrated clinical benefit, however, promises to be long, difficult, and expensive, with a number of false starts and many dead-ends.
Nevertheless, the ready availability of more comprehensive measurements, especially if tied to more precise characterization of different interventions, will enable the crafting of better, more effective treatments by supporting iterative clinical learning and the rapid evolution of improved clinical practices.
Inevitably, a significant limitation in this phase is implementation: New instrumentation, data flows, interventions and patient adherence will impose a compounded “last mile” problem, characterized by three major obstacles: figuring out which data are likely to be dispositive; processing these data so that a provider or patient can easily understand and act upon it; and continuing the redesign of healthcare workflows and delivery system to support, communicate and implement the new clinical procedures.
These are difficult challenges: how in the world are already overwhelmed and overworked care providers supposed to even begin to think about, the dizzying amount of new information, much of still-to-be-determined clinical value? We will be reminded again and again in this phase just how hard behavior change really is, for providers as well as patients.
The good news is that digital health innovators recognize and are starting to concentrate on these challenges, as evidenced by both the intense entrepreneurial interest in user engagement (signaled by the ubiquitous invocation of Stanford social scientist B.J. Fogg), as well as the progressive recognition that healthcare urgently requires design thinking – an approach to innovation that, in the words of design guru and IDEO founder Tim Brown, is “human-centered, iterative, creative, and practical.”
Phase 3: Integrating genetic and physiological data streams
The most exciting opportunity associated with digital health also, from a scientific perspective, may be the most difficult: integrating the new physiological data with emerging genetic data and evolving biological understanding to achieve profound new insights into the underlying basis of important diseases.
Today, we are surprisingly ignorant, and lack a fundamental understanding about what causes most human diseases. Integrating physiological measurement and genomic data in a consolidated network representation of physiology might offer an original perspective on health and disease, suggesting potentially unique portfolios of targets and perhaps bringing us a step closer to realizing the powerful scientific vision that originally inspired Brenner, Gilbert, and so many others.
In addition, the opportunity to better segment patients by combining genetic with physiological, and ultimately even social (see here) characteristics, could represent an important opportunity for disease management by making interventions more precisely focused, and hence much more cost effective, while also maximizing the likelihood of clinical impact.
As digital health affords medical researchers the ability to approach science in a less reductive fashion, more readily incorporating insights at the level of the whole organism, frontline care providers – and patients themselves – are likely to find themselves an increasingly central part of the scientific process, often providing the pivotal data and generating the animating insights. Progress in science will also bring further challenges for the healthcare system, as the benefits of more comprehensive monitoring must be balanced against the privacy issues and fundamental inconvenience. A whole new infrastructure of liability and regulation will need to be developed.
While we believe deeply in the promise of digital health, our optimism is tempered: human health is complex, our understanding is incomplete, and change – for both individuals and systems – is very, very hard. While media reports often focus on exceptional examples of early adopters, we would be foolish to use these to calibrate our expectations (a specific example of a more general publication bias).
We are likely to discover that even if we could acquire all the data we could imagine, there are fundamental limits on what this might reveal. It’s unlikely we’ll ever be able to “compute the whole organism.”
Even so – and with humbled mien — we should push digital health technologies hard, and leverage the resulting data as best we can to improve the human condition.
 Lewontin reference (including his explanation of “compute the organism”) is from the “Genes and Organisms” essay in The Triple Helix; see also Lewontin’s “Dream of Genome” essay, originally in The New York Review of Books, and reprinted in part in Biology as Ideology, as well as more completely in It Ain’t Necessarily So.
 Digital health can be defined (via the World Economic Forum, 2012) as “applying the most advanced information and communication technologies to the collection, sharing and use of information that can improve health and healthcare.”
David Shaywitz is co-founder of the Center for Assessment Technology and Continuous Health (CATCH) in Boston. He is a strategist at a biopharmaceutical company in South San Francisco. You can follow him at his personal website. Tory Wolff is a founding partner of Recon Strategy, a healthcare strategy consulting firm in Boston. This post originally appeared on Forbes.com.