On Healthcare as a Complex System: Ensuring Context and The Capacity to Adapt

There is a minor brouhaha in the world of healthcare quality and process improvement.  It started with a perspective piece written by Pamela Hartzband and Jerome Groopman published in a January issue of the New England Journal.  In the essay, (and I’m paraphrasing) Groopman and Hartzband argue that process improvement tools like LEAN are “medical Taylorism” that lead to to rapt attention to process measures
while resulting in unhappy patients and clinicians. They write:

Advocates lecture clinicians about Toyota’s “Lean” practices, arguing that patient care should follow standardized systems like those deployed in manufacturing automobiles. Colleagues have told us, for example, that managers with stopwatches have been placed in their clinics and emergency departments to measure the duration of patient visits. Their aim is to determine the optimal time for patient–doctor interactions so that they can be standardized.

The author’s main argument is that tools like LEAN encourage measurement of the wrong things and that clinical care is largely too complex for these types of tools to be effective.

We believe that the standardization integral to Taylorism and the Toyota manufacturing process cannot be applied to many vital aspects of medicine… We learn how to modify and individualize care in the real world, recognizing the variety of clinical presentations, the reality of multiple coexisting conditions, the variability of human biology, the effects of social and cultural contexts, and the diversity of patients’ preferences regarding risk and benefit, all of which defy rigid protocols.

Cue the ensuing bedlam:  Advocates for LEAN launched multiple attacks on the article, primarily arguing that LEAN has done great things and that the authors confused LEAN (Toyota production) with Taylor’s simple industrial strategies, or with “process improvement” efforts that aren’t actually LEAN.  John Shook, CEO of the Lean Enterprise Institute wrote a blog criticizing the article:

It is true that simple-minded, bone-headed applications of efficiency tools to healthcare situations are all too common. It is unfortunate that the doctors mistakenly associate these misapplications with lean thinking or Toyota. Indeed, there is much malpractice in the continuous improvement industry, with healthcare consultants peddling 12-step programs to be like Toyota.

Clinicians have always looked with some skepticism at LEAN.  I’m a believer, having witnessed massive operation improvements at places like Virginia Mason that took LEAN seriously. As I wrote last month, I’m convinced that operations science is the key to improving healthcare.

For some time, like Groopman and Hartzband, I’ve also been suspicious that some systems don’t lend themselves to improvement via engineering tools.  They aren’t linear and  they are too complex to be figured out using linear process analysis.

My thinking got a boost this week when I was introduced to the Cynefin framework and its original proponent, Prof. David Snowden.  The concept received a lot of attention in HBR a few years ago. The main take-away is that all systems can be categorized on a spectrum that runs from “simple” to “chaotic”.  To improving systems you need to recognize which type of system you are dealing with, and tailor your tools and strategies accordingly.   

Tools designed to improve simple systems can not be used to fix or interpret complex systems. 

If you have an hour, watch this lecture that Snowden gave at OSU a few years ago.  The guy is a genius and a superb speaker.

Snowden writes:

All too often, managers rely on common leadership approaches that work well in one set of circumstances but fall short in others. Why do these approaches fail even when logic indicates they should prevail? The answer lies in a fundamental assumption of organizational theory and practice: that a certain level of predictability and order exists in the world. This assumption, grounded in the Newtonian science that underlies scientific management, encourages simplifications that are useful in ordered circumstances. Circumstances change, however, and as they become more complex, the simplifications can fail. Good leadership is not a one-size-fits-all proposition.

… Using the Cynefin framework can help executives sense which context they are in so that they can not only make better decisions but also avoid the problems that arise when their preferred management style causes them to make mistakes.

Snowden argues that there are four primary system domains: simple systems, complicated systems, complex systems and chaos.  Simple systems are the domain of “best practice” and lend themselves to conventional process analysis tools.  Complex systems, in contrast are far more nuanced.  He writes:

This is the realm of “unknown unknowns,” and it is the domain to which much of contemporary business has shifted.

A complex system, unlike a simple one, contains multiple “agents” who continually modify the system through interaction.  Specifically, in the face of decision-making in the face of “unknown unknowns” people in complex systems develop the ability to borrow solutions from adjacent innovation- a process called exaptation. It is this “discovery” and repurposing of adjacent innovation that is how complicated decisions are best made.

In his OSU speech, Snowden argues that process improvement tools like six sigma work best for simple systems. They don’t work in complex systems.  Worse, by using simple tools to eliminate all excess capacity you lose adaptive capacity– the ability to run experiments and find solutions.  Snowden mentions that this was one reason why the innovative company 3M migrated from six-sigma a few years ago.

We need to create the preconditions under which exaptions can happen, which means creating inefficiencies.  One of the most evil things to have emerged in recent years is six sigma.  It’s a cult.  [With six-sigma] if you eliminate all excess capacity you lose adaptive capacity when the context shifts.  You over focus on efficiency at the cost of effectiveness. 

Snowden makes the point that the dominance of the engineering metaphor is the real problem in management science today.

People without medical backgrounds try to impose “best practice” in a medical setting, whereas the medical people understand that there are variations to how you do things and can’t remember until in the right context.  Humans don’t know what they know until they need to know it.  And the deeper the level of expertise the less likely they are to respond to consultant questions.

In fairness, I’ve never heard a LEAN expert argue that everything in healthcare should be standardized.  A LEAN black belt that I worked with a few years ago was clear that you should “standardize only what should be standardized, and no more.”  Her take, I suspect, was an implicit acknowledgment that stopwatches in the exam room is not at all what Taiichi Ohno had in mind.