Why are so many flights delayed?

Although my flight made it home from Baltimore, my flight there was canceled, and on the way back the two flights on adjacent gates to mine were canceled.

Northwest Airlines, with a hub in Detroit, seems to have led the pack, with 14% of its flights canceled two weeks ago, stranding over 100,000 passengers.

An article in this morning's paper confirms that it's getting worse. It also notes something I realize I should have seen myself, since it's one of those scale-dependent thingies: the delays counted by airplane are nothing compared with the delays experienced by passengers. The airline calls it a 1-hour delay, but it causes a missed connection and an overnight stay, or even longer, waiting to get re-booked, because all the other flights are already full too, and you're not the only one who got bumped.

Here's some numbers from the Times:

Ugly Airline Math: Planes late, fliers even later
New York Times
Jeff Bailey and Nate Schwebber
July 5, 2007

As anyone who has flown recently can probably tell you, delays are getting worse this year. The on-time performance of airlines has reached an all-time low, but even the official numbers do not begin to capture the severity of the problem.

That is because these statistics track how late airplanes are, not how late passengers are. The longest delays — those resulting from missed connections and canceled flights — involve sitting around for hours or even days in airports and hotels and do not officially get counted.

Researchers at the Massachusetts Institute of Technology ... determined that as planes become more crowded — and jets have never been as jammed as they are today — the delays grow much longer because it becomes harder to find a seat on a later flight.

But with domestic flights running 85 to 90 percent full, meaning that virtually all planes on desirable routes are full, Cynthia Barnhart, an M.I.T. professor who studies transportation systems, has a pretty good idea of what the new research will show when it is completed this fall: “There will be severe increases in delays,” she said.

Over all, this could be a dreadful summer to fly. In the first five months of 2007, more than a quarter of all flights within the United States arrived at least 15 minutes late. And more of those flights were delayed for long stretches, an average of 39 percent longer than a year earlier.

Moreover, in addition to crowded flights, the usual disruptive summer thunderstorms and an overtaxed air traffic control system, travelers could encounter some very grumpy airline employees; after taking big pay cuts and watching airline executives reap some big bonuses, many workers are fed up.

If a flight taxies out, sits for hours, and then taxies back in and is canceled, the delay is not recorded. Likewise, flights diverted to cities other than their destination are not figured into delay statistics.

About 30 percent to 35 percent of Continental’s passengers make connections between flights

A spokeswoman, .. added that many delays are caused by weather and thus do not reflect the airline’s performance.

...That is a typical level of missed connections, but Continental’s flights that day were 89.6 percent full, so finding seats on later flights for some passengers was difficult.

Continental also has a new system that sends e-mail messages — and, beginning next month, text messages to cellphones — informing connecting passengers on late flights how they have been re-booked.

It also is moving ticket kiosks inside the security area so passengers can print new boarding passes without going out to the main ticketing area or having to wait in line for a gate agent to help them.

The system, however, re-books people on the next available flight with a confirmed open seat and that is not always as soon as people might expect. Some are told their new departure is in three days.

“That causes them to go berserk,” said David Grizzle, a senior vice president at Continental. Often, on standby, people get out sooner, he said.

I also noticed that Northwest Airlines had attempted to solve this problem by institutionalizing the response. They now had entire special carts to make it easier for large numbers of passengers to attempt to make new bookings faster.

From the point of view of "lean" practices, and the Toyota Production Model, this represents one of the worst wastes possible - trying to become more efficient at doing work that shouldn't even be done in the first place. The risk is that the "workaround" will partly work, and then dig in for the long haul and become part of the new "normal" process, replicated 500 times in other places. New vendors will spring up to build even "more convenient" re-booking carts, and to lobby for sustaining this practice.



What might be done instead?

The first thing is to identify what the problem is. The problem is not thunderstorms or a feud between the traffic controllers and the FAA, although those contribute. The problem here is one of those pesky physical laws that I've been writing about, and what "the Yarn Harlot" pointed out as man's persistent desire to make "ten less than nine" and the delusion that maybe it just hasn't been rotated the right way yet and somehow this will "fit."

The law in question is called "Little's Law", and it looks innocent enough. It says that for any system the "cycle time" to process one unit (or passenger) goes up towards infinity as the system becomes full, and goes up much faster if there is more variability in the processing time for any individual step.

I can't easily find an authoritative textbook online, but here's the key info from a wafer fabrication newsletter "fabtime". (The same law applies to semiconductors as to passengers.
WIP = Work in process)

The relationship between cycle time and WIP was first documented in 1961 by J. D. C. Little. Little’s Law states that at a given throughput level, the ratio of WIP to cycle time equals throughput, as shown in the formulas below:

Throughput = WIP / Cycle Time

In other words, for a factory with constant throughput, WIP and cycle time are proportional. Keep in mind that Little’s Law doesn’t say that WIP and cycle time are independent of start rate. Little’s Law just says if you have two of these three numbers, you should be able to solve for the remaining one. The tricky part is that cycle time and WIP are really functions of the start rate.

Oh, and that tricky part is the devil in the details. What this really says is that as you try to jam more and more stuff through the same process, as it fills up the process starts to run into conflict and congestion costs, and the actual throughput starts declining rapidly, while "work in process" (passengers waiting for a flight) climbs towards the sky.

Fabtime's tutorial, shows a graph of the result, that shows that effect.

What this shows is that not only does the "cycle time" expected for a unit in this system (a passenger) to be processed (get home) go up, it goes up rapidly to multiples of the time it would take on an uncrowded system. So, for a very consistent, low variability process, the blue line,
trying to operate at about 90% full capacity will cause the process time to be six times the time it would take at 10% full. If the process has more variability (thunderstorms), this knee can be reached much sooner - at 65% capacity.

This is as true for service work and management work as for producing widgets or silicon wafers. Past a certain point, trying to shove more work through the system only slows down everything. So, the right thing to do is to find the sweet spot where the most work actually gets done, and resist the temptation to now try to fill every open space with more work. For wafer fabrication, this is about 85% "full". In other words, at the maximum throughput, 15% of the system will be empty, just "sitting there". This drives management crazy.

What typically happens is that people don't believe this result, even if they know it. (The delusion factor is strong, and surely 10 can be made less than 9.) None of the outside stakeholders, or visiting brass from the parent company understand this law, and a piece of idle equipment is surely a mistake and needs to be doing work! Or so it seems.

So, now, our friend, the feedback loop, comes into play. Once this knee in the curve is passed, and output starts to slow down due to congestion, the typical response of management is to go ballistic and push harder, trying to jam even more work through the system. This slows the system down more, which leads to management pushing even harder and starting even more work in process.

Then psychosocial factors come into play. Management becomes convinced that the employees must be goofing off, and become irate. "Surely that is true, because the total throughput is going down!" they think. Meanwhile, the swamped employees, seeing more in their in-boxes than ever and becoming exhausted trying to deal with all the internal delays at getting the simplest thing done, also become testy and hostile.

Meetings are held to discuss why so little is getting done, which takes more time, further slowing down the process. Labor strikes. Management retaliates, further cutting production and sales and revenue, which makes stockholders even more desperate to make up the losses with even more bookings. We end up with a positive feedback loop that rises until something breaks.
That's where it appears to be today.



If you click on that diagram, you can zoom it up to a readable size.

(That diagram is most of a Causal Loop Diagram, as developed by Systems Dynamics folks like Worcester Polytechnic Institute, or MIT Professor John Sterman (author of the tome Business Dynamics), drawn with Ventana Software's Vensim software that could put in numbers and actually run the simulation to see how this unfolds. This sort of reasoning is described by Peter Senge in his book The Fifth Discipline, where he uses an example of a beer production and distribution system to show how things can fall apart even when everyone is doing a good job, as they see it, because of "system factors" and "feedback loops". People interested in that would be interested in the whole Systems Dynamics Society. )

This occurs in a great many companies today. Unfamiliarity with Little's Law loads the gun, psychosocial factors cock the hammer, and every new thunderstorm or glitch pulls the trigger as everyone involved - stockholders, management, labor, and passengers, blame each other for the problem -- which is really a "system problem" not a "bad person" problem.

When this kind of thing happens to any health care delivery system, such as a hospital, it becomes a public health problem. When this kind of thing damages nerves and business effectiveness which leads to more pressure which damages nerves and leads to obesity and heart attacks and layoffs and no health insurance, it becomes a public health problem.

This is the kind of "systems thinking" competency I'm hoping the new ASPH Core MPH Competencies will lead to, so people can see this effect and head it off at the pass.

The biggest single controllable step here is to lower the blame factor, and realize we're all in this together. Myths and delusions and a norm that management's job is to crack a whip and push harder and harder come into play in a bad way when it is the system that is slowing down, not the employees. (Take it up with God, I guess, if you don't like Little's Law.)

Going around and around this loop is one of the major factors winding us all too tight these days, both humans and corporations. Maybe understanding what we've run up against can help defuse it and lead us back to a saner world for everyone.

At least, that's what Public Health hopes, in my view.

Wade

Disaster Prepareness using Web-based tools

I put another paper online:
Web-tools as potential assistance to Public Health Preparedness and response tasks.

Abstract

Faced with the prospect of natural and man-made emergencies for public health, it is worthwhile to see whether any of the new software tools might help. The new "web-2" tools have changed in a surprising way - they are easier and more fun to use. These tools are often free of charge, require no installation, consume no disk-space, and require no IT-department support -- all of which meet constraints that public health workers have faced in the past. The focus of these new tools is on cost, agility, simplicity, ease-of-use, and collaborative work. This paper reviews what tools are available, and how they might fit into the set of public health tool-box. Finally, training issues and other barriers to adoption are assessed, with an eye to figuring out what University-based Disaster Preparedness centers might do to make this technology legitimate, more accessible, and better utilized.

There's also a discussion of rules-based "plan generators" instead of "plans" in the appendix.

The Sixth Discipline of Learning Organizations - part B

Yesterday, in my post The Sixth Discipline of Learning Organizations, I reviewed a few of the lessons Peter Senge's book The Fifth Discipline teaches that we can learn from thinking in circles, not in lines.

There are other properties of loops that are critical, but as subtle as the difference between the behavior of a spinning bicycle wheel (a gyroscope) and a stationary one, or attempting to throw a plate or a playing card that is spinning rapidly versus one that is not spinning. At first glance you might say - it's just spinning, so what? But the behavior of trying to throw a plate and a "Frisbee" is quite different - the plate may go 20 feet and the Frisbee 100 yards.

Spinning rapidly in a circle matters. All feedback is not the same. The speed of feedback in a feedback loop also matters. The feedback rate matters ( loops per second or per day or per year).

But this morning I want to start looking at vertically oriented loops in hierarchically structured organizations - for which a triangle or pyramid shape is more helpful than a circle for discussion.
(Imagine the pyramid shown on the back of every US dollar bill.)

Say that the "boss" is the eye on the top of the pyramid, and that the boss's orders come down the right side, through the "chain of command" (which is actually a branching tree shape.)

At the bottom of the organizational pyramid, where it actually touches the reality and "ground truth", employees attempt to carry out those orders, and imagine that activity moving us from right to left across the bottom of the pyramid. Finally, status reports ("mission accomplished!") move back up the chain of command being consolidated at each level all the way back to the boss at the top. So, we have a vertically oriented loop, or cycle, because now new orders come down the chain and that loop pattern repeats.

So far, so good.

In a static, simple world, if all employees except one named "Joe" report success, and Joe keeps reporting failure, the classic model would say that the action management needs to take is to replace Joe. The model says all employees are interchangeable machine parts and if a part fails to do its job, the part is broken and should be replaced. This is a simplified version of McGreggor's "Theory X" of management, very popular in the machine age, from 1850 - 1950.
Another implicit assumption is that the boss completely understands the tasks to be performed, and is the resident expert. If people don't "perform" it must be because they are "lazy" and what is needed is a "bigger whip." Employees are told to "jump" and they don't need to understand why or agree -- they just need to ask "yes sir, how high sir?"

That model worked for early industrial models, such as workers in textile mills, or slaves picking cotton.

But, in a dynamic, complex world, that model breaks down and doesn't work. Actions and responses that worked yesterday suddenly no longer work. The "cheese has moved." The organization has to learn new responses to the same old inputs. The response of the outside world to an action is no longer predictable, and has to be judged based on rapid-feedback and a quick poke to see what happens and learning from that. We move into McGreggor's "Theory Y" of management where the expertise is now on the bottom of the pyramid, where front-line
troops are as likely to reply "What bridge? The bridge is gone!" as "OK, yes we crossed the bridge." Now an ever-changing set of facts or dots of information have to be aggregated upwards and "reporting" has to change into continuous "sense-making" of shifting patterns and images of the battlefield truth.

Again, this model is not that strange. It's the basic model we use when we have to move a bit of food from the table to our mouth on a very windy day - we move the hand a little, see where it is now, move it a little move, see where it is now, etc., in a very rapid sequence that automatically adjusts for the wind. If we don't adjust for the wind, the hand and food will miss the mouth on the downwind side. We don't "compute" wind velocity and use Newton's laws to figure out what to do - we just do it and watch while it's happening. It's no big deal. It's the basic "cybernetic loop" of tiny intent, tiny action, tiny perception, and repeat the loop rapidly over and over. It's a loop we can use to cross an unfamiliar room in the dark. Move slowly, stay alert and aware, and adjust as you run into things. It works. It doesn't require quantitative analysis or calculus or a computer or a PhD in robotics. It just requires using a very basic action and sensory loop over and over.

And, like any feedback loop, causality disappears in the normal sense. Motion alters perception and perception alters motion and the two become one, in a very real sense, a single motion-perception action and a loop as an actor.

Again, no big deal. So why is this important?

The big deal is that our society is in the middle of adjusting to this change from "Theory X", and a stable, static world with expertise at the top to "Theory Y" with a very dynamic, unknown world and the expertise at the bottom. In fact, because of the property of loops, there really is no longer much of a "top" and "bottom" in the classical Theory Y sense of the terms.

Just as the level of the water could be seen to control the hand on the faucet, the staff at the bottom of the chain of command can be seen to be controlling the General at the top of the pyramid -- and both those models are wrong, because it's actually the shape of the feedback loop that now has taken on a life of its own, on a whole different scale, and is controlling both of them.

Senge's point, and mine, is that most of the organizational problems we see around us are because we haven't managed to get that much right. In some health care organizations, an extreme case of the expertise being on the "bottom" of the pyramid, the top management still thinks in "Theory X" terms and tries to see itself as the expert in everything and "gives orders" to move in a certain way. The body reports back "No -- what bridge?" and the boss sees this as stubbornness, stupidity, or hostility and things just get worse from there.

Arguably one of the best "learning organizations" around is the US Army. I've mentioned many times before role of Doctrine in FM22-100, the US Army Leadership Field Manual. The pyramid model I just described is the theoretical basis for the doctrine, and every field action is supposed to be followed with a "lessons learned" session. News, particularly surprising news about a misfit between upper management's concept of where the battle or bridge should be and what actual boots on the ground see in front of them, is supposed to be free to travel upwards. Management, as it were, is supposed to listen to the staff and learn what's actually going on, not what management imagined yesterday was going on. It's not insubordination to say "Sir, What Bridge Sir?"

By simple trial and error experience, repeated millions of times, the Army has finally figured out what works and what doesn't and come to some conclusions that are startling to the Theory X old guard, but not at all surprising to the Theory Y thinkers. For one thing, listening has to go upwards, at every level. It's as important that superior officers listen to junior officers as vice versa. If new conditions at the bottom don't result in a new picture of what's going on at the top, the whole pyramid will simply drive off a cliff or otherwise carry out actions that bear no resemblance to reality.

And, because the picture of reality is not perceived directly, but has to come up the chain of command and be re-filtered and consolidated at a dozen different levels, that process has to be incredibly accurate, frank, honest, and unbiased. Even a 10% "adjustment" in facts, repeated over and over at each level of consolidation, can result in a reported "reality" at the top that is 180-degrees out of whack.

In a profound sense, the key word is integrity, and not just integrity when the going is easy, but integrity when the going is tough - not because of enemy action but because of "friendly fire from above". That kind of integrity is also part of the other key word in the doctrine - character.

If the information flows freely and rapidly and can spin up to a high rate of rotation, as with a bicycle wheel or gyroscope, this whole design pattern becomes very stable, agile, nimble, and capable of navigating the most bizarre terrain as events unfold in surprising and unexpected ways. BUT, if there are pockets of resistance to the flow of information, such as cover-ups, that model breaks down. Or, if there are superiors who think "superior" means they know everything and they don't need to learn from their men, the model breaks down. So, another few important words are honesty and humility.

See US Army Leadership Field Manual FM22-100
and What relates Public Health and the US Army?
and the whole posting from my Capstone slide 7 Theories are Changing which has twenty more references to the literature on high-reliability organizations in nuclear power plants and chemical plants and aircraft cockpits and hospital intensive care units, and what makes them actually work in practice. It just keeps coming back to the same thing and the same model that's right in front of us be we haven't finished mastering.

And, again we have a place where our religious heritage has been observing what makes society work for thousands of years and has more wisdom to offer on this than scientists, although the science is beginning to catch up at last. Our religions have been stressing virtues - integrity, honesty, compassion, humility, etc. - for centuries but we haven't really been listening or haven't thought that "mattered any more in the modern age." Actually, the basic cybernetic model is ageless, and true at any size and scale. It's going to be something we have in common with aliens from other worlds when we meet. It's a universal truth every bit as solid as other physical "laws" we rely on.

These are truths that are seen by Hindus, by Muslims, by Christians, by Jews, by atheists, and by learning organizations like the US Army. They can serve as a basis for unity among even such diverse groups and cultures. They can link science and religion without either side having to admit they were wrong about something and lose face.

Grasping and implementing that truth certainly looks like it could give us far more "bang per buck" than investing in new technology, new weapon systems, new gizmos and gadgets, and other ways to shift the detail complexity around.

Also, see my early post Virtue drives the bottom line with many links at the end to such literature. (excuse the formatting near the top of that post - I'm technically challenged by the html editor.)

Another author's take on this subject is "Spirituality in the Workplace - The Sixth Discipline of a learning organization, by Harish Midha at the University of Toronto.

Peter Senge's latest book is Presence: Human Purpose and the Field of the Future and readers interested in that book might also be interested in Stephen Covey's book The Eighth Habit. All these books teach the same gospel - that we are going to have to come to grips with the nature of community to "make it" through our social problems of this century, and that community requires us to realize the power and impact of "virtues" when amplified by the feedback properties of complex systems.

Another post I wrote exploring the role of community, virtues, and organizational learning and agility is The Importance of Social Relationships (short)

I also recommend: Pathways to Peace - beautiful slides and reflections to music on the value of virtues

A general summary of what I think are my best dozen posts on related subjects is here.

This is also relevant:

Spiritual solutions for technical problems

Enjoy, and please, for reasons this whole post embraces, send me feedback! A human can't sustain a thought without some measure of social support! Criticisms and objections are welcome. Use the comment box below, or send to my email in my "profile" box above.

Wade

Subtle nuances matter

It's not obvious what matters.

Whether we are just thinking, or doing fancy math, there's the stuff we leave in our model, and the stuff that we throw out, because it doesn't matter. Sometimes we think things don't matter because they have such a small effect that they are "negligible."

Sometimes we are wrong.

A classic example was the mistake that pouring toxins, like mercury, into the sea would dilute them to the point where they didn't matter. We forgot that nature has natural filters and amplifiers that recollected all those dilute molecules in one place, namely, the tissue of fish, so that the concentration was again dangerous or lethal to humans.

Or sometimes we forget some other factor. Nuclear scientists at Dugway Proving ground computed how much fallout would land on the ground and how much would wash away and "go away", and figured it was safe. After the sheep died and many people got cancer they found out, oopsie, that grass is remarkably good at harvesting water and holding onto it, so instead of the toxins washing away, they were recollected.

So, sometimes things that look like "small" effects do "go away", and sometimes they don't.

One conceptual problem we have is that we're not used to math where the answer depends on what time scale or geogrphic scale we're working in. So, yes, in the short run, "rock" is stronger than "water". In the long run, "rock" is demolished and destroyed by "water".

Or, in the short range, electromagnetic forces dominate gravity. A balloon, rubbed on the sleeve, will stick on the wall, not fall. For many purposes, gravity "goes away." But, if you look on longer time scales, it's the "strong" force of electromagnetism that "goes away", and the end state of the world is determined by that "weak" force of gravity, on a cosmological scale.

Or, if you look at an M.C. Escher painting of a staircase or waterfall, locally, there is nothing wrong, aside from a very slight noise or error -- but that error accumulates and on the larger scale, the total painting is absurd, even though locally any small part of it makes sense.

So we need to be careful about not "throwing out the baby with the bathwater."
It's not always obvious which is which.

Then there are other effects even more insidious or subtle. As the philosopher "Snoopy" observed one day, lying on top of his doghouse in the cartoon strip "Peanuts",

"Did you ever notice, that if you think about something at 2 am, and then again at noon the next day, you get two different answers?"

Or, another example I love, the story of two stone masons working on a church in the 1600's. One was doing very good work, and the other was doing work that needed to be redone often. The supervisor came to talk to each and asked them what they were doing. The one with poor outcomes replied "I'm building a wall." The one with great outcomes replied "I'm building a cathedral."

So, at least to human beings, it seems to matter a great deal whether the work they are doing makes sense in a larger context, whether it has "meaning" to them or not.

Is this true for people who write computer programs or "provide" health care services as well? Probably. How would we know for sure? And if it does matter, are we designing our systems in light of that effect, whatever it's called?

And is this just some "mental" or "psychological" effect, or is it an effect so "real" or fundamental that it would show up even if the agents building things were robots not people? Does this sort of thing matter to ants or bees or termites or bird swarms or swarms of viruses or bacteria?
Does it matter to the US Army?

Do real, tangible outcomes depend on "meaning"?

Certainly, from the model I described yesterday of nested contexts, the outer, distant contexts matter a great deal, although, again, the effects may take longer and longer as the context gets more distant. So, as many computer system designers and nation builders have discovered, "culture matters", and the survival of some change imposed from outside on a system depends, in the long run, on whether it fits with culture or not. If it fits, or can transform the culture to fit, it will remain. If it doesn't fit, the cultural equivalent of the body's immune system will identify it as "foreign tissue" and reject it. You can take that one to the bank.

Today's International Herald Tribune has an opinion piece on this subject at the scale of nation building, reflecting on Iraq and Afghanistan. Here's a brief snippet.

Do Not Neglect Culture
International Herald Tribune (on-line)
May 8, 2007
by Nassrine Azimi (Hiroshima, Japan)

The Rand Corporation recently published a study called "The Beginner's Guide to Nation-Building." It covers the basics with clarity and objectivity, defining the roles of the military, the police and the judiciary; distinguishing humanitarian relief from economic stabilization and development, explaining the complexities of governance and democratization.

But the book has almost nothing about what is clearly the Achilles' heel of recent nation-building adventures: culture. No single chapter is devoted to it - nothing on the role of culture in countries being rebuilt and, just as importantly, nothing on the culture of the nation-builders themselves.

Though we are reminded that six of the seven cases of nation-building initiated in the last decade by the United States were in Islamic countries, we do not learn much of the lessons of this extraordinary experience.

How, for example, did it inform the dispatch of some 120,000 mostly Christian soldiers to Iraq - a Muslim country and one of the most ancient civilizations on earth?

Neither do we learn much about what kind of cultural preparations, if any, were undertaken in advance of embarking in Afghanistan, also an ancient and proud land, with subtle values and vulnerabilities not readily accessible to the Western mind.

The fault, however, may not lie as much with the Rand book as with nation-building operations themselves. In most, culture has been at best an afterthought and at worst a shallow and cynical exercise in public relations.

This was not always so. The U.S. occupation of Japan between 1945 and 1952, so often cited as a model for Iraq, was quite different. American planners then appeared to have asked themselves some hard questions about dealing with a country they barely knew or understood, with which they had fought for almost four years, and which lay in ruins....

Perhaps this same effect is as evident in the many failed efforts across the country to install "Electronic Medical Record systems" where the system did not fit the culture or "the way we do things here", and the hope that the culture would "come to the system" was dashed by the fact that the system yielded to the culture.

This phenomenon is very well known and studied in public health, after a century or so of attempting to impose behavioral patterns on indiginous people who tended, as soon as the intervention team was gone, to keep the goodies and discard the behaviors that the strangers had imposed. The natives happily nodded "Yes!" while thinking to themselves "In your dreams!"

The lessons are that lasting change has to be rooted, and, in a mixed metaphor, rooted "deeper and deeper" upwards into the hierarchy of contexts that surround the point of intervention, or the unit of the hierarchy of life that is being tinkered with.

This effect is dimly and incorrectly perceived by many in McGregor's "Theory X" camp as "resistance to management", and as something that needs to be attacked, proponents of such resistance located and rooted out and fired, and overcome by brute force. In the short run, rock beats water. But, in the long run, water beats rock. If the intervention is "not me", the culture will ultimately find some way to reject it, or perhaps the culture will simply collapse under the conflict.

I think the prophet Yogi Berra once said "You can hear a lot by listening" , or words to that effect. It seems advisable that those messing with systems behavior at any scale should first investigate the system's "culture" before investing a lot in a particular change that seems, from the outside, to make sense. There are subtleties that are not obvious, "small things" that don't fit that turn out not to be so small after all, as the mercury or the fallout or the stone mason examples showed.

Whether a piece of the developmental puzzle "fits" or "is good" or "goes there" needs to be assessed at the cultural level, after all the "small things" have been given a chance to accumulate and add up again. This is a "complex adapative system" and the behavior at large scale is not reflected, in any obvious way, by the behavior at small scale.

The very fact that that's the problem is not widely understood.

There is no way that, for example, the CCHIT assessment of electronic medical record systems, at the individual user level, can possibly reveal whether this overall system will "fit" or "work" if "installed" at a particular site, in a particular "culture".

Collaboration-ware needs a completely different scale approach than classic IT software.
Again, that's not recognized as a problem to even fret about.

We are desperately short of good tools and accepted practices in this area. Maybe public health informatics can address that in the coming decade.

W.

Credits: Photo above is "The Hierarchy of Consciousness" by slark on flickr.

The hierarchy of life and implications for interventions

Apparently, we don't exist.

Every day more studies come out showing something that we'd suspected all along - namely, we actually have very little control over our own lives and even over our own decisions.

The people around us and our neighborhoods, at work and at home, are increasingly seen as the main cause of our beliefs, our decisions, and our actions.

Well, that just messes up everything, thank you. Our whole system of justice, and education, and rewards at work, and "the American way" are all based on the concept of rugged individualism, on one dominant person surrounded by a sea of "environment", making decisions, navigating the shoals of life, and deserving rich rewards for success or punishment for "being bad."

But that concept doesn't seem to survive the light of day, or a careful look at the evidence. And much of the evidence lately is coming from public health, including studies of the "health" of the "healthcare system" itself.

A very "robust" finding of the field of "social epidemiology" is that the physical health of a person seems to be very strongly associated with his or her "connectedness" with the tissue of society around them. The more someone is connected to the social fabric, the healthier they will generally be. The more someone disconnects and drops out of social interactions, the worse they will tend to be, across the board, in terms of almost every measure of morbidity and mortality. They'll be more depressed, more fatigued, less successful, less wealthy, more likely to be obese, more likely to have depression, diabetes, heart disease, asthma, the flu, common colds, etc.

But, does disconnection cause disease, or does disease cause disconnection?

The answer is "yes" to both, because this is not a linear chain of causality, but a causal loop. That means it can spiral downwards or upwards.

That's familiar. The more a person becomes depressed, the more likely they are to fail to cope, to get into trouble at work and home, and to worsen their situation at work and home. And, the worse their situation becomes, the more depressed they become. It's a "vicious cycle."

The ultimate end of that death spiral is, in fact death. There is complete disconnection and isolation, total dropping out, followed by catching the next excuse to die, from natural disease or neglect or violence, or violence against others (death by police). Just as a human cell, removed from the body, will lose the will to live and commit suicide ("apoptosis"), humans,
disconnected from the social body, lose the will to live, and find a way to die.

This is a real bummer in several ways. One unexpected way is that almost all research studies are based on statistics developed by a guy (Sir R. A. Fisher) studying crop yields where the causality only goes one way. The crops do not realizing they aren't growing and make midnight raids on the fertilizer shed. People, however, do. In fact, almost everything people do, or collections of people, are just drenched and dominated by feedback loops. And feedback loops invalidate classical statistics based on lines, not circles. (It's based on the "General Linear Model"). So, it's hard to study. So, people don't study it and go study something else.

Of course, there are tools that can easily handle such loops, including electronic circuit design or "system dynamics" or "feedback control system engineering." But those are almost unknown in public health so don't hold your breath.

Despite that, the evidence just leaps off the page. The most successful interventions in health care, as described in "Health Program Planning - An Educational and Ecological Approach" (4th ed) by Green and Kreuter, apologizes for abandoning classical models on page 3, with the comment that

"Ecological approaches, however have proven difficult to evaluate because the units of analysis do not lend themselves to the random assignment, experimental control, and manipulation characteristic of preferred scientific approaches to establishing causation."

Which is a long way of saying that the old set of "linear" tools and linear thinking really doesn't work, if you try to apply it to the real world that people, not billiard balls, deal with daily - a world dominated by feedback.

But, all is not lost. Even despite that, the healers of the healers, the designers of the health care system itself, have studied their own problem and concluded that the right unit of intervention is the small team on the front lines, which they call a "microsystem." In between the one doctor who is hard to change, and the hospital, which is hard to change, is the small practice team, which, fascinatingly, the Institute of Medicine has found easy to change.
(See Crossing the Quality Chasm.)

And, ta da!, big surprise, the recommended method of changing that unit of life, the small team, turns out to be "feedback." Well, of course it's feedback - that much becomes obvious once you shift lenses and realize that everything, at every scale, is more defined by what's outside it than what's inside it. (Mach's principle in cosmology.)

So, a single doctor or staff member can't really be changed by an intervention, because their behavior isn't really "theirs" -- it is a feedback property of the small team they work with. So, if a doctor or nurse "makes a mistake", it usually turns out that the place to fix isn't the individual, it's the larger structural team around them that effectively forced them to make the mistake. The system buys the gun, loads the gun, cocks the gun, hands it to the person on the front line who pulls the trigger.

And, on the flip side, there is no such thing as "the patient." Patients are people, and people come with a posse, an aura, their own small team of friends and family that mutually influence each other. So, ta da!, if you want to change how "a patient" behaves, or go a step further upstream and change what they believe, you have to address how the patient's "microsystem" behaves. The IOM didn't make that leap, but the rest of health education has realized that "family-centered" interventions are way more effective than "patient" interventions.

Of course, this really only changes the geographic and time scale, something the IOM hasn't yet realized.

This property of being defined by the outside peers is not restricted to cells or to people - it's a universal property of living things or any regulatory control system.

So, it's "scale invariant". That means if we flip to the next lens on our microscope and stand back another hundred yards, now we see the unit we are messing with is "the microsystem" but it is swimming in a sea of other "microsystems", and is ultimately dominated by the other microsystems as a peer group. Now, the time constant is much longer, so it may take months not days, but simply changing one small team and leaving its environment unchanged will sooner or later result in the change being undone, rejected like foreign tissue, and discarded by the larger living tissue of the body of the health care system. People will revert in hours. Clincial services may take months or years to revert, once the intervention pressure is released.

Man, how far does this thing go? Well, according to many people such as myself or Ken Wilber, it just keeps on going upwards. Wilber refers to one of these structural ladders of the hierarchy of life as a "holon." Norm Anderson, when at the NIH, refered to the same hierarchy from cells to tissues to organs to people to groups to neighborhoods to populations -- but nobody really wanted to hear that, so Norm left. The tissue rejected the novel idea.

Well, that math just gets impossible then, doesn't it? Not really, it just rotates. Large, tall, hierarchical structures have their own basic modes, as does anything else. There are almost certainly solutions that can be found, or descriptions, based on combinations of scale-invariant (symmetric) properties as basis vectors. And one such scale-invariant property is the concept of a regulatory feedback loop. At every level of this nested hierarchy, exactly the same problem has to be solved - how to maintain the equivalent of homeostatis in a sea of change. Cells do it. The pancreas does it. The Endocrine system does it. The body does it. People do it. Small teams (microsystems) do it. Hospitals do it. Health care chains do it. Whole cultures do it. Nations do it. They're all doing the same abstract dance, of seeking to reestablish their own feedback loop that works for them.

So it's kind of a fractal, a Christmas tree shape, where each branch is the same shape as the tree itself. The question is, what are the fundamental modes of vibration of such thingies? If it were made of steel and you plucked a branch, what would it sound like? (There would surely be harmonics of harmonics of harmonics.)

And, do such things have "resonant frequencies"? Is there some speed of change that will work far better than other speeds, or one that is far easier to "fall into" because it "aligns" with the larger resonance of the larger system around it?

Those are the interesting questions. In the short run, we have some immediate insights that don't need years of theoretical simulation and wisdom, based on this model or framework or lens, whatever you call it.

Here's a few:

1) To change a person, you have to change their peer group. They can move to a different peer group, or the peer group itself can be altered, but it has to happen.

2) etcetera. That is, you can't change that peer group, stably, without clicking up one more rung of the ladder, using a new power lens, and finding the peer-group's peer group.

3) Therefore, either you have a cascading, exponentially growing evangelical type of change, or you have a diminishing, exponentially decreasing, tissue-rejection kind of change. There is no such thing as a stable change of one "unit" at any scale. Life doesn't support constants, only growth or decay.

4) Our whole system of justice, education, rewards, and punishments is based on a flawed model of the world. That's all going to have to be rethought. All this emphasis on individual education has already run into the increasing emphasis on "teamwork" and "groupwork" and a realization that the unit of research, of discovery, of industrial production, of making or preventing errors is not a person, but a "Microsystem", a team, a cockpit crew, an operating room team, etc.

5) We're going to have to "bite the bullet" and start using the right tools to address these problems. They don't fit into the general linear model. All linear statistics break down and all linear thinking leads to erroneous intuition.

6) Collaborative IT systems are feedback loop generators, not huge replications of a single human-machine interaction. The "electronic health record", viewed this way, is part of the feedback loops that a patient uses to control his own life, or a doctor uses to control and manage their care for the patient, each side also calling on their own "microsystem" team to support this activity. Such systems cannot be evaluated or tested as if they were an Excel spreadsheet with a Graphic User Interface -- the human factors are feedback loops that can't possibly even show up in single user testing. The system will be made or broken on how the larger social fabric changes feedback loops when the system is put in place. That won't be revealed by the current CCHIT test suite.

7) This model would say that the right thing to be tracking for hospital adminstrators would be microsystems and teams, more so than individuals. The "dashboards" should reveal whether the microsystems are working, and, moreover, the people who need the dashboard aren't just the management outside the team, which is post-hoc, but the team members themselves for real-time self-management, steering and navigation. (That's straight out of the IOM's Crossing the Quality Chasm.)

8) Ditto for patients. This model would say that patient teams need their own Personal Health Record as part of a real-time feedback self-management model, that the doctors or clinical staff are only a very small remote second-order part of, for chronic disease management that involves life-style changes.

9) And, ultimately, this model points ever upwards. It says that people cannot be healthy unless their peer-group is healthy, and that cannot be healthy unless it's peer group is healthy, and, ultimately, all this depends on the national culture and planetary population being healthy.

So, yes, not only are you your brother's keeper, but your brother is, in many real ways, your keeper.

10) The "public" that "public health" must be concerned with (among others) is actually a fractal, nested, hierarchical part of the hierarchy of life. This cannot be made to "go away."
We need to "go to the mountain." Predictions as to the value of interventions in the behavior of a part of that hierarchy, on some level, whether cellular drugs or pancreas care of health system regulations, have to take into account that the parts are connected and will determine each other's behavior through feedback responses to interventional pressures.

It doesn't make sense to say "we put in a good system but the culture rejected it." The word "good" needs to be defined with respect to the whole hierarchy of life including culture. If the system is "good' in that metric, then the culture will, almost by definition, not reject it.

Well, that's pretty pedantic, and maybe you have a different view or some contrary evidence. I'd love to hear it. Let's have a good debate! See that "comment box" down there? Please use it and tell me whether you think I'm right, wrong, or need to increase my meds! Or email me. My email is in my profile.

Wade

Patient team empowered diabetes care in LA

Much of the model I suggest in my Capstone presentation is in use in Los Angeles - including teams of caregivers amplified by teams of local residents in a cascading help-your-neighbor approach.

Here's excerpts from an article Diabetics Need Much More Than a Shot from today's LA Times.

Managing the disease requires constant support and substantial investment, but would pay off in the end for thousands of people.

By Susan Brink
[Los Angleles] Times Staff Writer

May 7, 2007

Diabetes is afflicting more people, at younger and younger ages, sending doctors, insurers and public health officials into a tizzy as the epidemic threatens to overwhelm the healthcare delivery system. The annual cost of healthcare for an adult with diabetes is more than $13,000, and rates of Type 2 have risen sharply in the wake of the upsurge in obesity in this country.

A bold experiment is unfolding in Los Angeles County that may serve as a lesson for the nation as it battles the epidemic.

Experts know that the cost of care could be much lower if patients could take simple measures to control their disease and avoid complications: nerve damage, amputations, heart disease, blindness, even death. But surveys show that many, even those with adequate health insurance, do not get that care, which is costly and labor intensive, demanding daily attention from patients and timely responsiveness from doctors.

Poverty creates additional obstacles, such as finding fresh vegetables or a safe place to exercise. Study after study shows that low-income people have less access to healthcare and a greater risk of getting sick and dying prematurely.

But in an odd twist to the usual healthcare disparity story, more than 1,000 L.A. residents in low-income areas, most of them uninsured or on MediCal, are getting the gold standard of aggressive diabetes management — better, even, than many with insurance who live in ritzy ZIP codes.

The care is taking place at clinics in East L.A. and South L.A., two communities with the highest rates of diabetes in the county, as well as three other outlying clinics. A team of L.A. doctors is participating in the experiment, training nurse practitioners, pharmacists, social workers and community educators to intervene in a way that doctors cannot do.

They're offering frequent patient checkups to monitor the disease, and teaching patients to track blood sugar, get out and take a walk, cut out the doughnuts, all the things they need to do to keep complications at bay.

They are reaching people with uncontrolled disease in some of the county's poorest pockets.

"The county patients [in the program] receive care that is as good, and probably better, as anywhere in the country," says Dr. Mayer B. Davidson, endocrinology professor at Charles Drew Medical Center and UCLA.

There are signs that it's working. Studies so far show that patients in the program have improved blood sugar and have had fewer emergency room visits and hospitalizations.

Local pioneers

The intensive program is a response from local academic and public health experts to the crisis ahead. One of those experts, Dr. Anne Peters, professor at USC's Keck School of Medicine and an endocrinologist who specializes in diabetes, works both sides of the disease's socioeconomic divide.

Peters has a Beverly Hills practice ... But she also has a second job: supervising teams of workers on the same intensive model at five clinics including the Edward Roybal Health Center in East L.A., and the Hubert Humphrey Health Center in South L.A.

Reducing complications is key to protecting patients and controlling costs. It's not rocket science: lose weight, watch your diet, exercise, monitor blood sugar, blood pressure and cholesterol, take your medications, have regular eye and foot exams.

But often, these simple things aren't done.

Doctors needed

Peters' practice is an exception. She's a private practice doctor who spends as much time with her patients as they need. Her practice follows a team model, with a staff of nutritionists, educators, nurses and nurse practitioners to advise and prod patients via face-to-face discussions, phone calls and e-mails.

"I personally believe that anyone can take good care of their diabetes, no matter who they are or what their level of education," she says. "But they need a team, or at least a guide."

So successful has Peters' method been, that six years ago, she launched a pilot program for the county, supervising a team of professionals trained to educate, monitor and, when necessary, nag. The program took off in four other county centers in 2005.

The treatment team members make phone calls, hold classes, help patients change their diets, prescriptions or medication doses. They will even visit patients' homes to keep treatment on track.

Community members who speak the same language and share the same culture and who have successfully controlled their diabetes are recruited to teach classes and help coordinate care for newcomers to diabetes management. The idea is to allow specialist physicians to become consultants to community-savvy teams who offer up a steady drumbeat of medical attention and lifestyle education.

"It makes more sense to reach out to people where they live 24/7, and not think that a visit every three months to a provider is where all the care happens," says Dr. Jeffrey Guterman, medical director of the L.A. County Department of Health Services.

Before being accepted, patients in the county program sign an agreement that they will keep appointments and follow medical instructions. If they fail to comply, they're out. [editor note - it's hard to evaluate the success rate if failures are excluded. The point of my Capstone is to try to figure out what determines the drop-out rate and address that.]

Because funding is limited, patients can stay with the program only 6 to 9 months, but the expectation is that they can learn to control their disease in that time, then go back to a primary care physician. Those county doctors are ready, having been trained in how to manage diabetes patients.

Early studies suggest that the program works. A report in the April 2006 American Journal of Managed Care looked at how a key blood sugar test, called A1c, was controlled in 367 patients in the L.A. County program the year before and the year after they entered the program.

The A1c guideline was met by only 28% of participants when they were under traditional medical care. After a year in the program, 60% of patients met the blood sugar level goal.

[Effective Diabetes Care by a Registered Nurse Following Treatment Algorithms in a Minority Population -

Conclusion: A nurse making clinical decisions based on detailed treatment algorithms did a better job of achieving ADA-recommended process and outcome measures than physicians providing usual care.

(Am J Manag Care. 2006;12:226-232)

A second study published in February in the journal Diabetes Care found that diabetic patients in the program reduced their use of emergency rooms by half and cut down on hospitalizations. Total hospital charges dropped that year for the 331 patients studied to $24,630, from $129,176 the year before.

[ Effect of a Nurse-Directed Diabetes Disease Management Program on Urgent Care/Emergency Room Visits and Hospitalizations in a Minority Population

Mayer B. Davidson, MD¹, Adeela Ansari, MD¹ and Vicki J. Karlan, MPH

Diabetes Care 30:224-227, 2007]



Peters, ... keeps her private practice going only through donations from grateful, wealthy diabetic patients to a foundation she has started.

The health insurance benefits that her Westside patients have cover visits — generally about every three months. They don't cover extra visits to change medications or search out the reason blood sugar has fallen out of control. They don't cover time spent analyzing complex blood glucose printouts, insulin pump data or teaching patients to adjust insulin and drug doses.

Nor do they cover time for long discussions, phone calls, e-mails or the nutritionist, educators and nurses she employs to advise and prod patients....

---

susan.brink@latimes.com

capstone slide 2

My Quantitative Biomedical home-page.

My web logs:
Perspectives in Public Health


Systems Thinking in Public Health

Other links:
Intelligent Agent Infrastructures For Supporting Collaborative Work (Sen, Durfee, and Schuette, 1995 - Computer Science and Engineering graduate project, EECS department, University of Michigan)

Evaluation of Blogger. Ching-I Chang. Narayan Kansal. Younah Kang. Wade Schuette. SI 689 (Computer-supported Coooperative Work, UM School of Information graduate program Group Project. December 13, 2005. )

Evaluation of Blogger - powerpoint presentation.

Biographic:

I inherited my interest in computing from my uncle, Roger Schuette, who is shown in slide 6 in a publicity photo from 1952 (roughly), which shows the computer his team had just designed and built at the Barber-Coleman Company in Rockford Illinois. Unfortunately, Howard Coleman's genius at invention wasn't matched with his insight into business, and the company decided these "computers" had too many bugs to ever amount to much, and sold their patents to other companies, such as, I think, IBM.

In any case, I built my own first analog computer, from a kit, in 1956 - it played Tic-Tac-Toe and would always either beat you or tie the game, depending on who went first. I was trained in the language "1401 Autocoder" at IBM in Cleveland, Ohio, in1965 while working for the Thompson Ramo Woldridge company, which became today's TRW.

In 1976, I got my MBA and joined a team at the New York State College of Veterinary Medicine and the NYS Diagnostic Laboratory that copied the Electronic Medical Record system developed by G. Octo Barnett at Massachusetts General Hospital, written in a new language called MUMPS, and converted it to handle multiple species. The work was led by John Lewkowicz, (The Complete MUMPS: An Introduction and Reference Manual for the MUMPS Programming Language, John Lewkowicz) , and was part of what led ultimately to the current largest medical records system in the USA, the Veterans Administration system VistA. (Veterans Health Information Systems and Information Architecture, with a name that precedes Microsoft's use of the name for their own operating system, no relation.)

We had the animal hospital up with sub-second response time, 80 functions - admissions, discharge, billing, histopathology slide indexing, decision-support for medication orders, etc. - fully implemented in 1976.

In 1976, we all thought that human hospitals were going to be just a few years behind us in putting in Electronic Medical Records systems. Given 30 years perspective, I think that was optimistic.

My major lesson, however, is that "There is no such thing as a technical problem." The technology to build entire EMR systems has been available for 30 years. The designs are freely available from the VA system, or from the state-run national health service in the Netherlands, to name two. The impedance, reluctance, resistance to implementation of such systems is not due to money, because we did the whole thing in 2.5 years with a team of 5 people, technically.

The issues hospitals have are psychosocial issues, often perceived as "political" issues, or discovered with shock and awe by yet another technical team as "implementation" or "acceptance" issues, which were mistakenly thought to be "minor issues" or "bumps in the road to be dealt with as they arose, at the end of the project."

After 30 years watching this field, I'd go the other way and say these are psychosocial issues and the technology is the trivial part. A standard laptop today has more computing power than we used to run an entire hospital system in 1976, or than the Netherlands uses to run a gigantic 2,500 bed hospital with sub-second response time. (in 1989 at SCAMC in San Francisco I had lunch with their chief developer - they were running a hospital on one "MicroVAX", with a second one as a hot-spare, and power left over. Of course, they had to rewrite the operating system to do it...)

Of course, no technology group wants to "hear" the message that the shoals they are crashing on are social in nature and that their whole concept needs to be rethought. The good news is that there is a growing body of expertise, in places such as the School of Information at the University of Michigan, in "social computing" - now an official graduate major at UM, which has a 30 year background in "Technology-Mediated Collaboration".

The design features of collaborative software are so different from those of single-user software, such as a spread sheet or word processor, that the old insights about software design and evaluation are worse than valueless - they actually lead you down the wrong pathways. Software that looks great when one person tests it in isolation, and has a good "human interface" (for 1 human) can still have a wretched "multi-human interface" behavior.

The national CCHIT approval process for medical record systems doesn't even begin to assess this level of this multi-level problem, but you can be sure that the hospital staff will experience that level and respond to it. You can also be confident that, if this level wasn't consciously and explicity well-designed, that it will be somewhere between poorly-designed and pathologically designed.

And, it's rather hard to design such a system without substantial interaction with and feedback from the entire contemplated user community.

The odds that an off-the-shelf system can be simply dropped into an unprepared hospital setting and "take" are low, regardless how strongly this is desired or mandated from above, or promised by the vendor. In fact, there may be an inverse relationship between how much the system is seen as imposed from above ("take it or leave") and social acceptance of the corresonding cultural change that is required to readjust to that technology. As Public Health has learned repeatedly, outside interventions that are not culturally-sensitive, dropped from a speeding helicopter in local villages, tend to be barely tolerated with false smiles during implementation, then die a rapid death as soon as the implementation team leaves.

We'll have a sense that this concept is finally understood when we see EMR development teams start with the idea of social acceptance of this new paradigm (electronic collaboration), and when the planning team includes social psychologists, cultural anthropologists, and people from the Information Sciences. If the problem is perceived as simply "electronic records", that is, as one related to databases and messaging tasks, and human beings interacting are not prominantely featured on any of the architecture diagrams, then the odds are against success of the project. There will be large-scale social "tissue rejection" of the kind that Public Health has encountered routinely so much for decades, in response to which Public Health has developed the ecological model, "PRECEDE/PROCEED", etc. (See Health Program Planning - An Educational and Ecological Approach ed., by Lawrence W. Green and Marshall W. , 4thKreuter, McGraw-Hill, (c) 2005 - 1st ed (c) 1961.)

So, it's not that the solution to such problems are unknown - they are just not part of the "Information Technolgy" literature, but are instead over in the "Public Health" literature, and the two have very little cross-talk. This is where there is a pressing need for Public Heath Informatics to step in and take a lead getting these disparate groups to talk to each other.

Later I'll also recommend that Public Health Informatics may be required to cross the bridge between the "feedback control problem" that public health keeps crashing into, and the "feedback control solutions" that Control System Engineering has mastered, off in a different universe that again has no cross-talk in the literature.