With all the talk about Big Data (Big B, Big D) in all technology and business circles and with Big Data nearing the peak of the Gartner Hype cycle, are we ready to start talking about it in health care? Can we have Big Data in health care ahead of real connectivity and Meaningful Use? Is there a business model for big data in health care?
After attending the first conference on Big Data in health care, StrataRx, I’ll answer all the above an unequivocal “yes.” In fact, I’d say Big Data in healthcare probably isn’t hyped enough. While there are many roadblocks, a whole new kind of medical science is emerging.
Here’s why: to me, big data isn’t really about data, it’s about seeing, it’s about what John Hagel called, in the Power of Pull, “learning at scale.” As a one-time systems neuroscience researcher, I tend to see technology systems through brain and biology metaphors, and the one I keep coming back to, as did a few others at the conference, was that of big data as a visual system. It’s a visual system that is beginning, by connecting data, to see further and with higher resolution than ever before just how the health care system works.
We’re just beginning to uncover the causal relationships between actions we take and the reactions they create. Keeping in mind the learning health care system, here are a few things I learned and or had reinforced at StrataRx last week:
Big Data sounds cold, but in working with patients and the folks who need the data, we have the opportunity to become more human.
Sun co-founder and tech venture capitalist Vinod Khosla made a big splash a few weeks ago when he said that 80% of what physicians do will be obsolete. He clarified this a bit to say (paraphrasing) that machines do a better job of analyzing vast amounts of data, so the physician’s role will become more of a human relations role, or a guide. It’s the 80% of the mechanistic things we make physicians do that will go away.
Big Data could become Big Brother.
This will be a battle we’ll fight for years to come. I heard several speakers and had several conversations on the theme of “with great power comes great responsibility.” Big Data will undoubtedly create vast opportunities to cure disease, to reduce health care costs, and to find out what really works in care delivery. As we learn, many sacred cows will undoubtedly be killed, fortunes won and lost, people saved, and attention focused on new insights. I often talk about incentives as being key to changing the system for the better, and it is true now more than every that we’ll need to align incentives and create a culture of transparency to keep these opportunities moving forward. We’ll need to ensure that everyone is on the same side, including our potential big brothers.
While a few speakers touched on this topic, my “must see” is the talk by John Wilbanks on “Choose your Monopolies Wisely.” Of all the talks, this is my “must see” because we are at a unique time where we have to choose very carefully how this data is managed. Watch this video to see we how we can keep this in control of the patients through policy and careful management of our digital rights.
EHRs are one important source of data in the health care system, but they aren’t the only one.
Vast new oceans of data are opening up. I heard the phrases “blue ocean” “uncharted waters” “untracked powder” and “greenfield” more than a few times. The point is, whether the data is coming from claims data, referral data, mobile patient data, e-prescribing, EHR data, or some other emerging source, there’s plenty of opportunity for creating value. As one example, there are more than 7 billion medical claims transactions becoming available in various forms, state-by-state, annually. It like someone just invented a healthcare microscope (or perhaps, telescope) and we can now see how things work on a whole new level.
Highlights: Fred Trotter released his “mystery data set.” It turns out that it’s referral data between physicians from CMS. He’s opened this data to the public to see what kind of analysis can be done on referral patterns, perhaps overlaid with other data, such as claims data or outcomes data. We don’t know what we’ll find, but, now, we can begin to see how these things happen. In the hands of the right data scientists and combined with other data sets, we’ll be able to learn a lot about a variety of relationships between all the moving parts in health care. John Freedman talked about all payer claims data (APCD) that is now being released state by state. Each state is a little different, but Colorado appears to have one of the most open policies. I’m looking forward to the opportunity to research such claims data in the coming months.
Frederica Conrey of Booz Allen Hamilton showed that, by using claims data and social network analysis, “provider connectedness, or coordination of care, was more strongly and consistently related to how many different claims patients had rather than how much their care cost once they were there.” Look for interesting insights as people from all over start to analyze this data and match it against sources such as clinical data and referral data.
EHRs will enable “evidence-generated medicine.” EHRs and clinical data are becoming gold mines of clinical insight.
Bharat Rao, PhD, of Siemens highlighted that “a new form of evidence is emerging from rapid-learning systems that will mine vast amounts of electronic patient data collected in routine care to create “evidence-generated medicine.” Rao showcased some very impressive work done using clinical cancer data to find more personalized treatments. The results are publicly available at predictcancer.org. They’ve developed multiple key insights that are now clinical trials.
Rao highlighted the difficulty in accessing this data, but I imagine as their results continue to show value, processes will emerge where clinical data analysis will be written into the DNA of health care organizations (see closing the loop, below).
We’ve relied for a long time on human ingenuity and serendipity to come up with some real breakthrough hypotheses, but big data is flipping this notion around. Insights can come from the data itself. Meaningful Use is opening up a whole new body of what could be considered “big data,” but it is, as the ONC says, just foundational, and we’re only at the beginning. Data science is becoming a new form of hypothesis generation and may rapidly accelerate insights in the emerging science of care delivery.
The data scientists and technologists that enable them will drive the future of health care.
To get a handle on this new science of evidence-generated medicine, healthcare will need help. Several presentations by those with deep analytics and actuarial backgrounds show they are generating pretty incredible insights.
Carol McCall, chief strategy officer at GNS Healthcare and an actuary by training, showed how, through big data analytics, they create knowledge that companies need but aren’t looking for through “hypothesis-free, cause-and-effect relationship discovery at scale.” In working with one healthcare company, they “rediscovered” a drug interaction that the company had a hypothesis about by analyzing data from about 110,000 patients over 3 years. They also found a possible adverse affect for a commonly prescribed drug. The company is now in the process of validating the finding.
A highly recommend video of McCall’s talk is online and a good example of what’s possible through evidence-generated medicine.
Many folks with expertise from outside health care, like former LinkedIn data scientist Scott Nicholson, now at Accretive Health, are moving into health care because they want to do something meaningful and help our health care system. It’s good to see that some of the smartest analytic minds are beginning to work on something besides getting us to click on a link somewhere. As McCall said, “these hypothesis spaces are not going to get any smaller,” and we’ll need their help.
Integration is also key to enabling these hypothesis spaces to get bigger, and Shahid Shah explained how to overcome our integration challenges.
I ran into several other former and current neuroscientists at the conference including David Santucci, Scientific Solutions Manager from GNS Healthcare. This may seem like another seemingly, odd, mufti-disciplinary group that is getting involved in big data, but it makes a lot of sense. It’s also a group with a combined knowledge of statistics, analytics, cognitive science, biology and a smattering of machine learning, the perfect crossover skillset to enable big data in health care and enable systemic learning in health care.
Closing the Loop turns Big Data into real value.
In several of these conversations, the idea of “closing the loop” came up, in other words, ensuring feedback to the system from what we see with data. With all the predictions of what data will enable in healthcare, it doesn’t have much value until it changes the behavior of the healthcare system. Generating insights have to become part of the feedback in workflows. Big data can only help healthcare learn if the feedback is built into changing the system.
“The smartest people don’t work for your organization.”
Several speakers, including Jonathan Gluck from Heritage Provider Network and Stephen Friend at Sage Bionetworks channeled and/or quoted that line made famous by Sun Co-Founder Bill Joy. No matter how skilled, how much of an expert, or how competent your employees are, they are still a very small part of the cognitive capital that exists beyond your organization. Good ideas and good solutions can come from anybody: patients, nurses, doctors, developers… anywhere. Why put limits on your organization’s problem-solving ability? Connecting systems and opening up data stores isn’t a one-way street — you aren’t just letting go, you also can receive. Opening up patient data to a patient might just increase the accuracy or provide new insights. People outside your organization allow you to see deeper and farther than your own employees can, just by sheer number. Innovation Challenges and open data sets are just a few ways to unleash the power of the crowd.
Another key point here is that many of healthcare’s problems will likely be solved by people without a healthcare background and we need to give them that opportunity.
John Kansky from the Indiana HIE reported that about a third of ACO patients will receive their care from outside the ACO. Connectivity and access will become paramount to ensure quality care and coordination.
While always a bit scary, opening up data and connecting systems may not just be required for Meaningful Use, it may quickly become a strategic imperative to harness outside knowledge. Data is a resource that you’ll have to give to receive.
Quantified Self or Quantified Them?
There seems to be a lot of confusion about quantified self and where its definition begins and ends. Much of the confusion seems to stem from the idea that quantified self and self-tracking seems to involve a lot of work. One venture capitalist from Cambria Health Solutions said that he had seen far too many quantified self applications, but that he could see only 5% of the marketplace ever embracing such technologies. He said he’d like to see technologies that enable better “behavioral markers.”
To me, there’s not much of a line separating the two. At it’s best, “quantified self” applications such as fitness app RunKeeper do a great job of generating data to uncover new behavioral markers. With the right incentives and the right amount of transparency of the technology, who’s doing the quantifying seems almost irrelevant. Ideally, it should be up to patients what information is collected. To me, all quantification should be essentially quantified self, because we control what’s quantified and how.
And there’s so much more, be sure to check out the YouTube channel. I see health care data science as just the beginning of a new kind of medical science, a new kind of discovery. ♦