Part III: Moneyball Medicine, Narratives, Probabilities and Making Better Medical Decisions

Lotteries work and Vegas was built on this simple notion: we believe we can beat the odds, we all believe we are a little special. 

Part of our notion of being special, of our ability to beat the odds comes from our notion of our own narrative, our own story of our life in which we take a starring role. If we can imagine it, it seems more plausible and more possible. Economics Nobel Prize-winner Daniel Kahneman calls this the “Plausibility vs. Probability” in his latest book, Thinkng, Fast and Slow. Our minds tend to turn plausibilities into probabilities, inflating and overweighting both our chances of success as well as our risk of losses, depending how risks are framed.

Kahneman points out that he simple fact is, we live in a probabilistic world and we’d all make better decisions if we made decisions over the long run with perfect rationality and decision-making based on the long term, but it’s not easy. We’re not wired for it. 

The fact is, we’re all very bad at making statistical decisions, including doctors and patients, and it costs us. Even those well trained in statistics are susceptible to these and many other types of statistical errors.

Part of physician training is to make them ultra-confident and infallible. Who has time to argue when someone’s life is on the line?

Yet physicians consistently misinterpret statistics. It’s been well-documented that we and our doctors will make very different decisions based on whether an option has a 90% survival rate vs. a 10% mortality rate, yet the choices are the same. 

Or take this classic example in I.C. McManus’s “The Arithmetic of Risk“:

You are asked to advise an asymptomatic woman who has been screened for breast cancer and who has a positive mammogram. The probability that a woman of 40 has breast cancer is about 1%. If she has breast cancer, the probability that she tests positive on a screening mammogram is 90%. If she does not have breast cancer, the probability that she nevertheless tests positive is 9%. What then is the probability that your patient actually has breast cancer?

Go ahead, scratch out some numbers….

It’s not easy.

If you’re like most doctors, you estimated that the woman who tested positive on a mammogram has about a 90% chance of having breast cancer. But you (and me) and most doctors are way off. 

The correct answer is 10%.
Think of 100 women. One has breast cancer, and she will probably test positive. Of the 99 who do not have breast cancer, nine will also test positive. Thus a total of ten women will test positive. Now, of the women who test positive, how many have breast cancer? Easy: ten test positive of whom one has breast cancer—that is, 10%.
When it’s framed the right way (in terms of a narrative rather than pure mathmatics), and we have a little bit of the right kind of framing in our decision support, the answer’s easy.

These poor decisions lead to unnecessary treatment and increased costs. In a related example:, the majority of physicians draw unfounded conclusions from screening statistics and this leads to increase in costs, an increase in information, yet no improvement in results:

How we frame the question and how we calculate the odds has an enormous influence in how we get an answer, and what recommendation we might make. 

The opportunity in health care, reiterated by several physicians on the panel with Esther Dyson at the eCollaboration Forum at HIMSS12, is that we need to provide both physicians and patients to tools to help them make better decisions, but ultimately they are their decisions to make. Much of accountable care models will hinge on both physicians and patients abilities to make the right decisions and how we can assist them in the process of interpreting statistics.

That’s why we need better decision-making tools for physicians and patients, to help us make decision based on probabilities rather than plausibilities (our legal system, for better or for worse, is built upon plausibilities, which leads to other bad health care decisions, but that’s a story for another time).  We expect physicians to be infallible and certain, and their training often reinforces this mindset, but the right apps and dashboards could do a better job of interpreting and making recommendations based on statistics. Read Kahneman’s book and you’re unlikely to trust an expert ever again (without a little statistical help.)

As the saying goes, we are all special, just like everyone else. As we enter an age of Big Data, we’ll need more help than ever going from data to meaning (with the right framing) to decision-making in practice. The success and failure of these new models may will hinge on our ability to do so, and may well be the new basis of competition in health care.


ACOs and Moneyball Medicine Part II: A New Era of Network Science in Health Care

Dave Chase (@chasedave), CEO of Avado, spoke at the Collaborative Health Consortium‘s weekly Pilots and Collaborations call last Friday.

Dave led with the quote from Dr. Josh Umbehr: 

A good scalpel makes a better surgeon. Good communication makes a better doctor.”

Communication (and understanding how information flows) is the critical tool for physicians to get the information they need to make the best choices they’ve been trained to make.

On the same day, Dr. Jim Yong Kim was named as President Obama’s choice as the new head of the World Bank. Just last night I watched an interview that Dr. Kim did in 2009 with Bill Moyers that focuses largely on health care. During the interview, Dr. Kim talked about Southwest Airlines and how what they’re thinking might do for medicine:

“they have taken seriously the human science of how you transfer simple information from one person to the next…What we need now is a whole new cadre of people who understand the science, who really are committed to patient care. But then also think about how to make those human systems work effectively. We’ve been calling it, aspirationally, the science of health care delivery.”

Dr. Kim is talking about many things here, social networks, health care experience design, network science and in other parts of the interview he talks about the need for reducing variability with such thing as Kiazen or Lean. It’s risk reduction in care delivery through continuous feedback, learning and understanding information flows in networks of people, systems and care delivery processes.

Patient Engagement Means Helping Patients (and their Physicians) Make Better Decisions

It struck me that Dr. Kim and Dave Chase are talking about something very similar, if not the same thing: networks, communication flows and decision-making.

Dave Chase talked about the need for patient engagement and “control” in this brilliant slide from Avado:


You can see the rest of the slides Dave has given us permission to post here (great presentation).

When we talk about patient engagement and control, we’re largely talking about who has control over decision making, how to engage them, and how to help influence that decision-making through information with flows and feedback. 

In the graph above to the left in the “At Home/Low Acuity” area, this is where ACOs and new payment models’ success or failure will be determined. As Dave notes, this is ultimately 75% of health care costs that are behavioral, based on decisions that patients and families make day in and day out.  Understanding behavior networks in health care is critical. Fowler and Christakis’ Connected that does a deep dive into how behavior patterns transfer through real social networks (not just the online variety).

While Dr. Kim and the Dartmouth school of Care Delivery are be focused on the right side of the graph, in a patient’s accute phase when they are in the clinic or in the hosptial, but the realization is the same:

We need to understand what engages people, what information is needed and when to help patients and their care team make the right decision and reduce the risk of a bad outcome.

It’s Network Science for health care.

Network Science in Health Care

The fledgling field of Network Science has been studied and applied in fields such as UX design, big data (scaled learning) CRM systems (be sure to read Michael Wu, @mich8elwu,’s posts on this), social network analysis, lean (and lean startups), Kaizen, open source, open innovation and supply chain logistics, and yes, Moneyball and now ACOs. It’s a network perspective to systemic learning, with accelerated, scaleable learning as the critical measure of success. It’s a matter of understanding connections, decisions, engagmement and human factors, information flows and how they all fit together all enabled through feedback. It’s the new dialectic.

(For more on this subject, along with Connnected, I highly recommend The Information by Gleick, Thinking, Fast and Slow, by Kahneman, Where Good Ideas Come From by Steven Johnson, and the collected writings of Clay Shirky and John Hagel III to pursue this subject further.)

The Moneyball approach to baseball is a network and systems approach where statistics and critical measures, such as on-base-percentage, or how to approach a specific batter in a specific situation are found to matter more than nuance physical or mental characteristics of the players.

Payment reform, at long last aligning incentives of patient, physician and insurer, we hope, can only be successful with a continuous learning and statistics driven approach similar to what Moneyball did for baseball. Network science and big data are the science of figuring out what works on a large scale. Once systems are connected and health information is digitized, both through payment reform and initiatives such as Stage II meaningful use, we’ll be able to make better decisions in medicine. As long as everyone is covered, that’s a good thing. It means insureres will focus on keeping well rather than restricting coverage on those who need it most. It a networked world, it means personal responsibility and universal coverage, something both conservatives and liberals could love!

We Need People and Systems that Understand Network Science in Health Care

Dr. Kim is right, we need a whole new cadre of people who understand netowrks in health care.  This is the science of reducing risk through the understanding of how information moves with various types of networks. We need people who understand the fledgling field of network science in health care. Network Science is the key to driving innovation while reducing costs because it is ultimately about accelerating organizational and systemic learning. New payment models need to find out what works in health care and fast, and both Dr. Kim and Dave Chase understand that what’s happeing is that networks of organizations, people, care delivery and technologies can lead to better care and reduced risks —  if we understand how they work.

Looking at network effects and information transfer across a wide variety of disciplines, we are rapidly gaining new insights into risk reduction from a network perspective.

Network science, the science of care delivery, patient engagement, social network analysis, customer/patient relationship management, user experience design, big data, and the algorithms being used now in baseball, are all part of a whole new way of looking at risk and probabilites across a wide variety of disciplines. In medical school and hospitals, as Dr. Kim pointed out, we currently do it ad hoc. For better outcomes, we need to understand these information flows better, so we can learn what works.

ACOs will need to understand very well Jim Yong Kim’s research on variation, Dave Chase’s work at Avado and even Moneyball. Scaleable learning is rapidly becoming the new basis of competion, even in health care. It’s the rocket science of health care.

Next up: Why We Need Decision-Support in Health Care (Hint: we’re all really bad at intuitive statistics)

Big Data, Algorithms and Moneyball Medicine – Part I

I finally got around to watching Moneyball this week. Great film. Roger Ebert points out that the film “isn’t so much about sports as about the war between intuition and statistics.” (I’ll let you guess who wins if you haven’t seen the movie). The main character was neither Billy Beane (Brad Pitt) nor Peter Brand (Jonah Hill) so much as a set of algorithms the Billy and Peter characters implemented to build a winning baseball team at low cost. I’ll go out on a limb and declare it the greatest statistics movie ever. (It’s a short list of great statistics movies.) 

And it’s timely.

Algorithms seem to be everywhere in the news. A few pieces really nailed the trend this week, driven by O’reilly’s Strata conference. 

Tim O’reilly aptly described the trend in the context of business intelligence in an interview. “The role of visualization, of BI (business intelligence), is to help someone design an algorithm…the end game here is the design of systems that do the right thing in response to data better than we can.”  (quote is approximately minute 20).

The key for healthcare will be defining the right contexts where algorithms can really help. We, and physicians, can use a lot of help, but we’ll need to go after the low-hanging fruit first (there’s a lot of it), and be very careful where we draw the line between intuition and algorithm.

Another great piece on the subject of big data was in Forbes from from Jerry Michalsky, also inspired by the Strata conference. He points out that we’re heading toward big data and algorithms at a time when we’re discovering how little we know about ourselves. 

Notable quote about big data for Michalsky:

It’s eroding the culture of expertise. Read Daniel Kahneman’s new book and you’ll stick to statistics. Add a pinch of Taleb and you’ll never speak to tie-wearing experts again.”

Hmm. What experts in health care might be seeing their culture eroding? The key here is to enable the experts with better #UX-centered tools, I think, like dashboards for docs and patients. Another key point: “big data is nurturing a culture of collaboration.” To get big data, we need to pull from multiple sources, and that means collaboration among all players, participating patients, physicians and payers. This is much of what’s driving meaningful use and accountable care, we need to get to big data to achieve the triple aim.

And finally a piece by Ezra Klein notes that in 25 years we’ll be loading ongoing personal health metrics continuously up to the cloud.

This kind of monitoring may be the future of health care. It will pay for itself in real dollar terms, but how much of our privacy are we going to be willing to put up for it? This is a question we’ll be answering for decades to come: who owns our data and what are we willing to sell it for?

I don’t know who said it, but one of the quotes out of SXSW this week was that “data is the new oil”. Indeed, it’s driving a lot of industry right now.

We’re going to start seeing a Moneyball/statistical approach to just about every arena. Moneyball, algorithms and big data (and oil) are becoming synonymous. We’re looking at new ways to use data and connect data to find out what actually works in a wide variety of arenas. Human intuition is on the ropes, but we have to be very careful about how far we extend the reach of algorithms to make decisions.

I’ll be writing a series on what that this all means for health care. I’ll discuss where we need help with algorithms in health care (decision-support, patient engagement and more), what’s the dark side of algorithms in health care, and how big data, energy policy and health care really are related. We’ll need to begin to be very careful about what we mean by “patient-centricity” in this new realm, just as Michalsky warns about “customer-centricity”.

Accountable care means BI for health care systems (something that’s relatively new) and the need for a statistics-driven approach. As I’ve written before, it’s no accident that Blues, Aetna and United are buying up health IT companies that focus on care beyond the clinic.

It’s a whole new ballgame. More to come.