High Performance in Medicine

I was asked to give a keynote presentation this week at the annual Medical Informatics World conference in Boston.  I thought I would share a copy of my talk for those that were not able to attend.



Remembering Dr. Michael Rosenberg

Photo of Dr. Michael Rosenberg

Dr. Michael Rosenberg

I have struggled this week to accept the death of Dr. Michael Rosenberg, Founder and CEO of Health Decisions.  Michael was an inspiring friend to me, and one of a handful of people that regularly gave me confidence in innovation.

Michael and I would meet periodically at a local coffee shop.  These times for me were always treasured opportunities to connect with someone else passionate about our industry.  We talked about pharmaceutical and health industry trends, work, books, flying, and anything else that came to mind.  We had started making plans for some joint writing and speaking.  The coffee was always strong, the conversation was always great, and I can only hope he left these meetings with some sense of the rejuvenation that I always did.

As I’ve gotten older, I’ve come to appreciate the rarity of people like Michael.  An entrepreneur physician and business leader who deeply understands clinical research, analytics, and technology is of course exceptional.  But beyond his impressive ability to crosswalk so many disciplines, the things I admire most were his unquenchable thirst for improvement, his passion for technology-enabled advancement, and his unending desire to help organizations and people bring medical treatments to patients faster and more efficiently.  Michael ran a highly successful company, but he earnestly wanted a better tomorrow.  Some people have careers; Michael had a mission.

He and I had one of our coffee sessions a few weeks ago, and I started to ask him why are you still doing this?  Many leaders at his level would have retired by now; in the game of business, he had already won.  So why keep doing the hard work of change?  But I didn’t ask him; I already knew the answer.  The vision has not been fully implemented yet.

Mission work can be hard and seemingly endless, but it can also be very rewarding, as Michael’s life and work regularly reminds me.  I have no doubt his vision for more efficient, data-driven, and adaptive clinical research will be fully realized in the near future.  It might happen through natural evolution, but my money is on the disruptors, the innovators, the mission-oriented people like Michael who refuse to accept the status quo.  I hope I’m there to see it.

In the meantime, it’s going to be a while before I can return to that particular coffee shop.  I already miss the sense of connection with someone who sees the opportunity for change, and is undaunted by it.  I will miss conversations that flow effortlessly between business, research, analytics, and technology.  But mostly, I will miss the spark-like energy in the eyes of a bright, gifted leader — an energy he always seemed able to rekindle in me even in difficult times like these.

Live: health analytics on streaming radio

RadioI’ve been invited to appear on an AllAnalytics.com radio show tomorrow (Monday, 7-Oct-2013) called Making Medicine Smarter.  We are going to talk about health analytics, the book, and all things health care.  The show will air live at 2PM Eastern US, and you can hear it by going to http://www.allanalytics.com/radio.asp?doc_id=267367.  There will also be an after-show text chat on the AllAnalytics message board if you want to dive deeper on any topics.  Hope you can join us!

The 5 Most Annoying Data Excuses

data miningI recently had an interesting Twitter exchange with my friend Dan Munro over at Forbes regarding a KevinMD posting,  Quality is a Word that Lacks Universal Meaning.  The article touched on one of my book topics: the industry’s reporting-centric, manufacturing-oriented conceptualization of quality is ambiguous and unreflective of the problem space.  We need to look at quality differently.

Dan raised the concern of data — do we really have the data to support a different view of quality, especially in light of resistance to data collection?

Dan is absolutely right; from a quality perspective, we often are “flying blindfolded and handcuffed.”  But do we really lack data?

Consider that most health organizations have four available data sources:

  • Their own data (though it is often not readily consumable)
  • Partner data (though they may not be sharing it)
  • Purchasable data (though we often just pay for the answer, not the data)
  • Public data (though we often can’t determine how best to use them)

Between these four categories, we are drowning in data: EMRs, claims, referrals, device, billing, CMS, labs, imaging, clinical trials, public health, service utilization, reimbursement, health exchanges, costing, genomics, consumer sentiment, behavioral data, consumer health devices, e-prescribing… and it keeps coming.  It doesn’t look to me like we lack data; it looks like we lack insight.  So why do we lack insight?

Part of the challenge is focus.  Instead of getting smarter on how we use data, we continually shift our attention to the next set of data we need to collect, the next regulation we need to satisfy, the next benchmark we are handed, the next incentive to capture or penalty to avoid, the next dashboard report to build.

A second challenge is lack of process oversight, which I think validates Dan’s question.  We have plenty of processes across health care, and we do manage them.  But we fail to instrument or otherwise empirically characterize those processes in such a way as to offer process-related insights.  Note that meaningful use and HEDIS don’t do this either — they simply establish benchmarks.  Additional instrumentation does not imply we must collect more data — we could use derived or surrogate measures, for example.  But it just needs to be a focus (see above).

But a third challenge is the litany of status quo excuses that leaders hear daily about why we cannot be more data-driven in our decision making:

1. We don’t have enough data.

Of all the data excuses, this one gets the most play.  There are two assumptions behind this state.  First, it assumes that a complete inventory of data options exists, and that inventory does not offer enough to do what you need.  In my experience, most firms do not have a handle on all of their data options, and they do not understand (because they haven’t looked analytically) whether a given data asset is useful in understanding the problem at hand.  There is also an assumption you can determine how much data you need without analytics – which is untrue.  You actually need to do an analysis to determine if you have enough data.

2. The data we have isn’t good enough.

This excuse always interests me because, 99% of the time, it surfaces without any actual analytical work being done.  The data isn’t good enough…good enough for what exactly?  The perception is usually derived from either a) previously struggling with the data on an unrelated project, or b) physically looking at the data, which often appears messy, incomplete, and/or error prone.  Yet all data assets have limitations such as these; the utility of data can only be assessed in the context of the question being asked and the analytical method being used.  And the data never gets better until you use it and make it better.

3. The data exists, but we can’t have access to it.

Our obligations to patient privacy notwithstanding, we have to move beyond the rampant fears in using data.  Rules and contract terms should not prolong patient suffering and/or drive up health costs.  If an organization has data assets that cannot be used for agile innovation, then change the rules, re-write the contracts, change the consent forms, or provision new data sources to compensate.

4. We can’t use the data because of regulations.

This excuse is really a variation of #3, but it has the apparent added weight of the government behind it.  My same opinion applies.  There is no question that we need strong data protections and use provisions — we’ve been facing that for decades.  But no patient I know would rather suffer, die, or face bankruptcy than have their data used responsibly to improve medical decision making.  And if they do, that is fine, but have the patient or sponsor “opt out”, not “opt in.”

5. The data is not structured in a way that is useful; it would take too much time.

So it is too much work to innovate?  If you do not have easily consumable data assets, then maybe it is time to start treating data and analytics more seriously.  It is possible to create meaningful, agile analytical assets and insights, but it doesn’t happen accidentally, and it doesn’t happen without work.

The hard part of these five excuses is they all hold an element of truth to them.  But when we accept those challenges as barriers, we don’t progress.  And to Dan’s point, there is justifiable, growing resistance to performance metric pile-on.  Most practitioners I know do not believe there is a strong association between existing quality measures (MU, HEDIS, etc.) and real-world, patient-centered outcomes and costs.  And though that conclusion is overly broad, I think there will be growing evidence to support that view.  For example, two recent studies —  one in the Journal of General Internal Medicine, another in Health Affairs — call into question the association between readmission rates and quality.

The fact that we are using data inadequately does not mean we should not be using data.  And it doesn’t mean we need to re-double our efforts around meaningful use and physician education.  It means we need to become smarter about the questions we ask,  more focused in the priorities we set for our organizations, and predictive through the tools we give practitioners and patients to make more informed decisions.

In my opinion, we should be building and deploying comprehensive, predictive quality models that we know improve outcomes and costs; not justifying retrospective metrics that we hope might help.

Top 10 questions about the health analytics book

Health Analytics book coverWell, today is the day — Health Analytics: Gaining the Insights to Transform Health Care is finally out!  I promised to cover the book in more detail, so today I thought I would take the opportunity to answer some of the more common questions that I get asked about the book.

What is the book about?

The book is about transforming health care and life sciences through data-driven innovations.  It describes a roadmap for growing organizational capabilities across a broad range of insights such as health outcomes analysis, clinical research, financial management, customer engagement, and personalized medicine.  There is an About Health Analytics page on my blog that describes the book in more detail.

Who is the book written for?

I mainly wrote the book for industry leaders who are curious about how data can help them transform their organizations to become more innovative and competitive: value-oriented in delivery, and evidence-based in practice.  It is not a technical book; you don’t need to know anything about statistics or analytics to read this book.  It is mainly designed for non-technical professionals within providers, payers, pharmaceutical, biotechnology, and regulatory organizations who are trying to develop strategies and roadmaps for becoming more information driven.

Why did I write this book?

At the time I made the decision to write the book, part of my job was sharing the opportunities in analytics with current and potential customers.  Over time, it became apparent that there was not really a resource that I could easily point to that captured the landscape.  Also, I had spent more than 6 years studying the field of analytics across health and life sciences, and I had learned quite a bit about what was — and was NOT — happening in the industry.  I began to feel like I had something to say — maybe unique and valuable, maybe not — but something to contribute to the discussion of health transformation.  I was also looking for a vehicle to showcase some of the talent in my team (for example, Dr. Graham Hughes contributed a fantastic chapter to the book called “Best Care, First Time, Every Time”).  So writing a book seemed like a good idea at the time.

How did I get a publisher?

That was the easiest part of the process for me.  My employer at the time, SAS, has a well-established publishing group — they function both as an independent publisher, as well as partner to other publishers.  This book would not have happened without them, as I would have had no idea how to sell a book concept.  They did a great job in helping me develop the book idea, work through how to communicate the idea, and figure out where and how it made the most sense to take the idea forward.  If you are interested in writing any books — business or technical — covering the field of analytics and data, SAS is a great place to start.  In the end, Wiley opted to publish this, and it has been a really great experience.

How long did it take?

That’s hard to measure, as I tended to write in bursts.  In elapsed time, it took a little more than a year, which seems like a long time until you have a contract with deadlines, and then it becomes a very fast 12 months.  Since I already had tons of research on the topics I was writing about, my time was mostly spent actually developing the content.

What was the hardest part of writing?

The hardest part for me was managing the scope of the book.  I wanted a book that was a) easily readable by non-technical executives; b) covered the landscape of opportunities in analytics, not just a single perspective; and c) wasn’t purely conceptual, but instead offered some real-world perspective on the problem spaces.  Trying to strike the right balance between depth, breadth, length, and impact was really hard.  If I tried to cover all of the opportunities properly, I would end up with a book 2,000 pages long that no senior leader would have the time or desire to read.  But surveys are often superficial, and so much of the analytics opportunity requires showing the value hidden in the complexity (e.g., predictive modeling).  So I opted for a middle road — a survey-type book that dived into specific analytical examples, challenges, and case studies — and “bookending” the overall chapter flow with a common-sense plan for how to execute these ideas.

What makes this book different?

There are so many informatics books already out there, and a lot of authors have exhaustively covered topics many people associate with analytics: business intelligence, dashboards, quality metric reporting, etc.  I knew going into this project that my content would be somewhat contrarian to the prevailing winds in informatics.  It’s not that I don’t think quality metrics are a good idea; they are fine.  I just don’t believe that the time, money, people, and attention being devoted to retrospective, descriptive statistics will produce the fundamental insights required to understand the delicate, complex balance between outcomes, costs, safety, and personalized medical decision making.  I think innovative results come from innovative approaches, so I wanted a book that argues for more innovation in how we use data.

Will you make much money from this book?

I doubt if any royalties I might someday see will cover much of the time I spent creating the book.  But I didn’t write it to make money.  It was a labor of love.  Maybe my next book will be something everyone (not just health professionals) might want to read, like a sci-fi epic.  Then again, I may have just written a sci-fi book; we’ll have to wait and see. 🙂

Will I write another one?

Maybe.  I really enjoyed the writing process — even the mundane stuff like editing — so I could see myself doing another one.  When I realized that I had something to say, the idea of writing this book became something worth doing.  So I suppose if/when I have more to say, that will be the sign to do the next one.

Where can I get a copy?

Health Analytics: Gaining the Insights to Transform Health Care is available at Amazon.com, Wiley.com, SAS.com, Barnes & Noble.com, and other retailers.  The ISBN codes are ISBN-10: 1118383044 and ISBN-13: 978-1118383049.  Let me know how you like it.

Health Analytics vs. Business Intelligence

One of the questions I get asked frequently is the difference between business intelligence and health analytics.  And I struggle with a good answer; there is so much inconsistency in the use of terminology in this space, and to answer the question requires that you be able to cleanly define terms like “business intelligence.”  But I do think it is an important question, as how you answer the question implies quite a lot in terms of what you need to be doing operationally.

Business intelligence (BI) was a term first coined in the 1950’s, but many people would argue that our modern conceptualization of BI is based around Howard Dresner‘s 1989 definition as “concepts and methods to improve business decision making by using fact-based support systems.”  As the adoption of information technology has grown over the past two decades, the conceptualization of BI has evolved in parallel such that today, BI can be taken to mean just about anything.  This situation has led to the advent of the term “business analytics” (an equally ambiguous term) to characterize a more modern, advanced approach to insight development.

Of course, health analytics has an equally ambiguous history rooted in other overly generalized terms like “informatics.”  A few years ago, this terminology issue led me to explicitly define health analytics as a domain separately from health informatics and business intelligence:

Health analytics is the domain of advanced analytics focused on providing strategic insights into the inter-dependencies in health outcomes, profitability, and customer preferences and behaviors.  Health analytics target insights that support transformational programs and business growth opportunities, enabling organizations to improve medical care, strengthen financial performance, deepen customer relationships, and pursue medical innovations.

I’ve reflected many times since then whether this was a good definition or not, but I’ve not come up with a better alternative yet.  But in my mind, three key attributes that fairly clearly characterize health analytics are:

  • multi-market: health analytics intentionally blur the lines between payer, provider, and pharma.  A fundamental assumption in health analytics is that the insights needed to transform health require collaboration across these historically siloed market segments (i.e., convergence).
  • multi-dimensional: health analytics are designed to explore the relationships within and between clinical, financial, administrative/operation, and personal aspects of health.  As such, they cross many domains of business inquiry and data.
  • multi-method: health analytics provide both retrospective and prospective views of health through multiple channels.  So in addition to the typical descriptive health metrics (e.g., counts, averages, percentages), health analytics incorporates more sophisticated and powerful techniques such as predictive modeling, data mining, and optimization to help define what is going to happen in the future (not just what happened in the past).  And the insights it produces get deployed operationally in a broader set of channels within an organization (e.g., reports, alerts, rules, etc.).

So with that as backdrop, here is my attempt at describing the differences between business intelligence and health analytics.

Scope Usually domain specific: clinical, financial, administrative Domain specific or cross domain: designed to link, for example, clinical and financial information together into one model
Timeframe Mostly retrospective Both retrospective and prospective
Mathematical Concepts Descriptive statistics: sums, averages, means, medians, percentages, counts Descriptive statistics + inferential statistics: correlation strength, forecasting, prediction, simulation, optimization, data mining
Data Structures Standardized, usually in data marts Both standardized and emergent (based on research needs)
Hypotheses Implicit — included in the assumptions behind measure / metric Explicit — part of a formal iterative discipline of research, discovery, and validation
Project Delivery Linear: project scope can be well characterized before the project starts Iterative: project scope is designed to constantly evolve based on findings
Project Risks Mainly associated with data: completeness, accuracy, cleanliness, representativeness Data risks + Time risks + Finding risks: projects include research which by definition carries uncertainty
Insight Delivery Standardized paper reports and web pages that often include “dashboards” Standardized paper reports and web pages/dashboards.  Also includes custom reports based on research, and direct integration of insights into operational systems and processes (e.g., alerts, rules engines, decision support)
Business Impacts Best suited for operational performance measures against clear standards; rarely competitively differentiated Operational + Transformative: best suited for strategic insights into changes, growth, investments, outcomes, etc.

In summary, the difference is really based on how broadly you define “business intelligence.”  In it’s broadest sense, health analytics could be considered a subcomponent of BI, taking many of the concepts of BI and applying them to an industry-specific set of questions and issues.  In it’s fairly popular technology-oriented characterization, BI is seen more as a reporting framework; in that sense, BI is a subcomponent of health analytics.  And if you prefer the term business analytics, health analytics could be an industry-specific implementation of BA.  But regardless of how we define it, I hope we can agree that it represents an under-served aspect of health information technology today.

The Health Analytics Book is Coming

Book Mock CoverAs many of you know, I’ve spent the past few months writing a book on health analytics and informatics.  I’m pleased to be able to report that, as of last week, the fully-complete draft is in the hands of my editors!  So I thought now might be a good time to give you a sneak peek into what’s in store.

The working title of the book is Health Analytics: Gaining the Insights to Transform Health Care.  It is all about EMRs.  Just kidding.  I’d probably sell a few more copies if it were about EMRs, but it is actually about what we might do with all of that information we are now collecting.

The target readers for the book are executives and leaders in health and life sciences organizations who are curious how to better incorporate analytics and informatics as part of a more effective and efficient business strategy.  I do not spend any time talking about the technical aspects of analytical methods, and there is no software code in the book — it is designed completely for the non-technical reader.  Here is a draft paragraph from the preface that summarizes the topic briefly:

This book is about painting a tangible picture of a different future for health care — one where the business of health care is more closely connected to the evolving science of medicine and the evolving role of individual health care consumers (i.e., patients).  It builds the case for a fairly singular idea; namely, that using health-related information in new and creative ways can dramatically lower costs, enhance profitability, improve patient outcomes, grow customer intimacy, and drive medical innovation.  We call this opportunity “health analytics.”

Some of the topics that the book covers includes:

  • what is health analytics, and why should every health enterprise now be focused on developing their health analytics capabilities?
  • how is health analytics different than the retrospective, reactive, and metric-oriented status quo informatics and business intelligence approaches operating across health care today?
  • what are some key health analytics capabilities and scenarios that will be foundational for 21st-century health enterprises?
  • why health analytics requires us to share data and insights across organizational and market segment boundaries?
  • how does an organization get started, and what does a roadmap look like for becoming an information-driven enterprise?

The book is being jointly published by Wiley and SAS, and can be pre-ordered on Amazon.  I don’t have a publishing date yet, but I’m guessing it will be out in the early fall of this year.  Over the coming weeks and months, I’ll be gradually providing more information, including a few deep dives into the book content.  So stay tuned!

My Health Analytics Talk

This past summer, I was fortunate enough to be asked to deliver one of the keynote addresses at the first Health Analytics Symposium conference.  Afterwards, a number of people asked for copies of my materials, which I’m fine to provide except no one ever knows what I said around each slide, and the slides themselves are far from self-explanatory.  So I managed to get a partial recording of my presentation, and have posted it here for those that might be interested.  Note that some of this content will be in my upcoming book on health analytics. More on that later.

Seven Reasons Why That Beautiful Informatics Architecture Won’t Deliver

Every organization has one: the informatics diagram.  It flows from left to right, starting with a few source systems and ending with users doing incredible things in browsers and mobile devices.  It looks well thought out, comprehensive, and flexible.  I’ve created a fair number of these pictures myself.  They often make sense even to the least technical people in the room.

Typical Informatics Architecture


There is only one problem with the standard picture: real transformation from advanced analytics won’t work that way.  Here are the reasons why:

1.  Data will be federated.  I know we don’t like to hear that.  We like the control and convenience of our own data buckets.  And we will have plenty of data to manage.  But not all of it.

2.  Your data is not analyzable.  I know many people think it is, but it’s not.  The data are incomplete and inconsistent.  It lacks clinical context of some primary measures and endpoints.  The structure does not lend itself to real analytics, and you are missing information that your analysts will need.

3.  All data are not created equal, and you don’t know a priori which is valuable.  If you operate under the assumption that you amass as much information as possible on an ongoing basis, you should also assume that you will be amassing costs on an ongoing basis — costs that are not tied to any specific return on investment.  That doesn’t make sense.  You need a mechanism for understanding what data is actually important.

4.  Extremely high data volumes require extremely high analytical computing capacity.  If you really need high-performance computing (i.e., your questions require analyzing these high data volumes), that infrastructure is a very different architecture.  If you don’t really need high-performance computing (i.e., your questions can be answered with more modest data sets), then your plan (and diagram) needs to account for data triage (see #3).

5.  There is no provision in that diagram for a learning system.  Insightful intelligence in health — like any other discipline — is iterative.  You need a feedback loop from the users and their work back into the architecture and data, a way of capturing what we learn that should both influence and contribute to subsequent analytical exercises and insights.

6.  Clinical decision support in the future will not be purely based on business rules and published practice guidelines.  First, it takes too long to deploy those practices and rules (10 years or more by some estimates).  Second, the question is not whether practice guidelines and rules add value; the question is whether any standardized guideline can reflect the natural heterogeneity that exists in patient populations.  We already know that the multiplicity of factors influencing a given patient’s health outcomes and costs are beyond the cognitive capacity of single medical practitioners, and much more extensive than can easily be documented in standardized treatment plans.  The future of medical best practices will be driven by real-world, real-time analytics — predictive models that do not supplant business rules and policies, but that more accurately inform and guide collaborative care teams (including patients and family members) about treatment options and outcome propensities based on “patients that look like this one”, not broad-based patient populations.

7.  The challenges aren’t technical.  Though these diagrams serve well to educate people about the opportunity, the challenges are not related to hardware, software, or networking.  Organizations need new competencies.  Cultures will need to evolve: a tolerance for exploring the unknown, a comfort in risk taking, a thirst for knowledge, a willingness and ability to operate in real-time, and a discipline for managing electronic information and insights as evolutionary assets.

In a future post, I’ll share some ideas for what an advanced analytics architecture for health informatics might look like in the future.