O'Neill may have been formally educated as an economist, but his zeal for systems-wide efficiency in the U.S. healthcare industry — his current passion — was music to the ears of his audience of operations researchers.

O'Neill has a history of speaking his mind and shaking things up in both the corporate and political worlds. As chairman and CEO of Alcoa from 1987-1999, O'Neill set a goal of zero lost workdays per year. Despite some resistance, Alcoa dedicated itself to safety perfection and just about achieved it. O'Neill's experience transforming an old economy firm into a new economy success has been chronicled as a study by the Harvard Business School and studied in business schools across the nation.

O'Neill shook up the political world with the publication of Ron Suskind's 2004 bestseller, "The Price of Loyalty." O'Neill was the primary source of information for the book, which depicted the Bush White House, according to an Amazon.com reviewer, as a "world out of kilter" where "policy decisions are determined not by careful weighing of an issue's complexities; rather, they're dictated by a cabal of ideologues and political advisors operating outside the view of top cabinet officials."

Today, as CEO and co-founder of Value Capture Policy Institute, O'Neill is focused on bringing the pursuit of perfection of complex systems that he pioneered at Alcoa — what he calls the "theoretical limit" — to America's healthcare system, a system he maintains is needlessly hemorrhaging red ink. As you can see from excerpts of his plenary speech that follow, O'Neill drove his point home with perfection. — Peter Horner

I like provoking people, especially if it causes them to do something useful, so I thought I'd begin with one little bit of biographical information that Mark [Mark Kamlet, provost of Carnegie Mellon, who introduced the plenary speaker] didn't give you. In 1960 and '61, I was at the Claremont Graduate School in California in a highly mathematical economics program. In those days, they believed at Claremont that it wasn't proper to get a doctorate in less than five years. After I'd been there for a year with my wife and two children, I decided we couldn't live on $1,200 dollars a year, so I accepted an appointment into what was then the Management Intern Program with the U.S. government. It was competitive process where 300,000 people took the exam, they interviewed maybe 3,000 and offered 300 of them middle management positions with the government.

I ended up accepting a job offer at the Veterans Administration, not having been a veteran but having grown up in a military family. The attraction to me was the promise of entering service with the U.S. government at the Veterans Administration and getting a year-and-a-half worth of additional schooling, basically in O.R. and systems analysis and learning how to do computer programming which looked to me like it was going to be a necessary and promising discipline to have under your hat in 1961. So I went to the VA, got some additional training and ended up with a project responsibility to apply the ideas of linear programming to the question of feeding 210,000 veterans across the country every day. It was, in effect, the Stigler [Least Cost] Problem, but on an unbelievable scale. I found it fascinating and really a lot of fun to work with because it required a reformulation, since there weren't many veterans who wanted to eat five pounds of flour. So it was necessary to restructure the problem so that it was a composition of things that could actually be fed to people and, in today's lingo, backing up into the supply chain and doing behavioral analysis to find out how often and with what frequency and at what preference level people wanted to consume different things . . . and accounting for reasonable variations.

Those of you who are computational will appreciate this: In those days, the matrix was so big that it took two-and-a-half hours on a 70/94 computer to do one iteration. Today, we could do it instantaneously, but in those days it was really fun to go to the computer room and watch the magnetic tapes running. If you were really on top of your technology, when the tape stopped, you instantaneously knew what was wrong with the code. Right? So you'd go back and rewrite your code, schedule more computer time — usually at 3 o'clock in the morning — and make another pass. That's what it was like trying to do practical O.R. 40 years ago.

I'm telling you all this just to give you a sense that I'm not up to date with you. I don't read your journals every day, but for the past 40 years in all of the different things I've done, I would say I've been doing practical systems analysis and operations research, sometimes to pretty good effect. So I want to start with a question. This is probably the largest assembly of rationalists that you could find in the country today. You all believe in rational thought and logic and optimization and improvement and all of those things that we share together, so here's my question for you: How many of you have ever accepted a prescription from a doctor that you couldn't read? Right. What does that say about rationality?

Enormous Opportunity


I believe that there is enormous opportunity for improving the condition of society — not just in our country but on a worldwide basis to a level that most people can't imagine — if we begin systematically deploying the ideas of analysis and operations research, but maybe in a somewhat different way than we've traditionally done it. I want to talk to you about that with illustrations from the health/medical care business because it's an area that I spend a lot of my own time on today. If we had the time, I could also talk to you about primary and secondary education where I think there is a yawing gulf that cries for the attention of the analytic community, but today I want to focus on health and medical care.

I believe — and if you would like to quarrel with me that's OK — that if today we simply practiced all of the things that we already know about medical care interventions and did it with excellence every day, we could have a huge improvement in the health condition of our population and in outcomes of medical care interventions and simultaneously reduce the cost by 50 percent.

Now, for those of you who don't follow the healthcare field, let me give you some numbers. Today, arguably we are spending $2-plus trillion dollars a year on health and medical care in the United States. Imagine what we could do with a trillion dollars? It's staggering. The running rate in the second quarter of this year of our GDP was $13.3 trillion dollars. Imagine what we could do with another trillion dollars and have better medical outcomes to boot? Some of you are probably saying, "Hmm, tilt." I know because I looked at the papers that were submitted for this conference so I know that many of you are working on health and medical things. I applaud you for the work that you are doing, but I don't think together we're looking at this problem in the right way. So let me begin with a structuring idea that I found really useful in the work that I've done over the last 40 years in a variety of places.

I believe in an idea called the theoretical limit. In preparation for this talk, I was fascinated to read the definition of "operations research" as defined by Webster's: "the application of science and mathematics, especially mathematical models, to study and analyze complex problems." I thought that was really interesting because it doesn't say anything about O.R. actually improving anything; it's just about studying and analysis. That didn't seem really right to me, so I hope you'll join me in trying to get Webster to change the definition.

But back to my proposition of the theoretical limit. I believe from a long time spent learning that how you structure the problem preordains the outcome. And so, when I take you back to the question of how many of you accept prescriptions that are illegible or incomplete, and then I think about the medication pathway, legibility is a problem. But if you think about medication and the pathway of medication as analysts, I think you would join me in saying, "Well, the starting point for this is the training that a doctor receives that causes some triggers to go off in his mind that say when I see this set of problems, it occurs to me that this medication intervention would be therapeutic and useful to you, and that's what you should have That's step one. Step two is actually creating a communication that begins the process of seeing that you get that medication. How many other places in your life do you know where the typical response to a non-binary communication is laughter? Especially when it can kill you. You know, it's not a trivial question of the New Yorker cartoon where one woman says to the other, "Look at this thing that my son did. You can't read it. He's going to be a doctor."

Cost Without Value


It is baked into our culture that this non-binary communication is something we should anticipate and accept. It's crazy on the face of it, but if you can get by that, and you go do a little bit of analysis work, you find this kind of thing. In a typical retail pharmacy, they spend 20 percent of their time or maybe even 30 percent playing telephone tag with doctors trying to find out what the intent was. In my theoretical limit world, that's a cost without value; it's all waste. It's something not only to be regretted but to be fixed. Then if you go and look in a typical place, what you find after the person who is responsible for filling this prescription determines to their own satisfaction they know what they are supposed to do, there are a whole lot of potentials for error even in filling the prescription. We saw it in a really dramatic, awful way just a few weeks ago when five premature babies in Indianapolis were killed by the administration of an adult dose of Heparin that the children's bodies couldn't tolerate. So five children died, just like that.

Here's an interesting test: Some of you may be on hospital boards or you do your own work in a medical care setting. It would be really interesting to ask people in hospital communities around the country if they have acted on that issue. Have you changed your system to prevent the possibility of adult doses of Heparin being in a preemie ICU? At most places the answer is no. And even if it is yes, the problem and the solution haven't been generalized to all kinds of medications that are intolerant for weight differences.

Let's get back on our thought track. We get our prescription filled and — in a lot of places, this is still true — after the order is filled, it goes into a batch process. This is directly applicable to the work O.R. people have done for a long time, with supply chains and logistics and all the rest. This is a real place — thank God it's not as typical as it used to be — but in this real place, the carts were filled to send to the ward on Monday, Wednesday and Friday. If you examined what happens in that process of distributing drugs for patient application and you look especially at the returns that come back on Monday, you find there's a huge volume of drugs that come back on Monday afternoon that weren't actually consumed on the wards between Friday and Monday. Why not? Patients' conditions changed. The doctor decided on a different intervention. So when the drugs come back, one technician spends the rest of the day restocking the shelves with things that came back. Another technician takes 40 percent of the intravenous materials that were prepared on Friday morning and pours them down the drain. Some of these things cost $2,500 a cocktail.

This is the real world; I'm not making this up. What you find too often is the administration of the drug wasn't done on the agreed schedule because the nurses were busy doing something else. If the time sequencing is important, the patient is disadvantaged by the malfunction of the administration system. And this part of the loop is missing almost everywhere — observation of the consequence of the medication intervention and an alert to the research community if the results weren't consistent with previous applications with patients with the same the set of circumstances. So if you just look along that one pathway that we all kind of assume is OK, it's not OK.

Most of the O.R. work I see going on in health and medical care tends to be very micro. For example, it will work on the logistics path from the pharmacy to the wards, and figure out a better way to schedule when medications are sent to the wards. Too often there are what I would characterize as behavioral issues — let me go back to illegible and incomplete prescriptions. Too often I think our disciple — let me include myself in your discipline — is agnostic about behavioral problems. If we are really serious about finding higher levels of optimization in things like medical care, I don't think we can be agnostic about the need for behavioral change.

Solving Half the Problem


Some of you who are really up on this are probably thinking you know the answer to the problem — we need computer order-entry systems. We just need to apply technology. Incidentally, it cost about $10 million for a medium-sized hospital for a computer order-entry system. When you introduce computer order-entry, it solves about half the problems, and the remaining half are more complicated and less tractable than the ones you started with. One of the things I've discovered — not just in working in health/medical care but in my industrial career — is that once you adopt a technical solution or a mechanical solution, it tends to freeze a lot of important things that are better left unfrozen.

I'll give you an example from traveling around the world. I get invited to a lot of places. I was invited to go to this high-technology center to see this gee-whiz piece of equipment that had been invented for inventory control and management for automobile parts, in this case, side panels. It involved automatic indexing and labeling and all the other things that O.R. people bring to the table along with great mechanical equipment. The equipment put these things on a vertical conveyor so that when you needed a part, you just punched three buttons on the keyboard and it automatically brought you the part and unloaded it.

After praising the people to the high heavens for the great work they had done, I said to them, "Of course, you know that in an ideal system, there isn't any inventory." Really. If you look at the inventory flow at a place like Dell Computer, it's four hours worth. They don't have to manage inventory because they don't have any. But I would submit to you, you don't get to those kinds of questions if you're optimizing a mess. For me, a better way to think about this — especially health/medical care where the opportunity is so huge — is that analytic people get involved in asking the famous Toyota production system "five whys." Not stopping with two or three or getting an approximate optimization, but pressing across an institution so that you can begin thinning about a whole house optimum instead of a pathway optimum

I'll give you a different illustration of this, again from a real hospital setting — a 600-bed hospital where they have the national average level of patient falls. That is to say, when people come into the institution they are weak and feeble and sometimes they have dementia and Alzheimer's and all kinds of other problems, and so they fall. Now when I first started working this subject in some detail I found an interesting thing: Almost every patient who fell while in an acute care hospital had been identified when they were admitted as someone who was likely to fall.

When you think about design principles, this is almost a perfect system: We've designed it for people to fall and they do. And we can predict the rate. It's about one person per year for every bed in an acute care hospital system. So if you say, well, there's something wrong with this, let's make observations and let's collect data because it will talk to us. What you find out is that a lot of patient fall occur after 11 p.m. and before 6 a.m. Why is that? Is there something about the moon? Not really. There is something about patients' needs, particularly in the middle of the night if they are in bed in a dark and unfamiliar place and reach down for the call button, you press it and you can't really hear it go off.

After five minutes and no one comes, the patient feels the urge to go the restroom. The patient gets up and tries to go to the restroom. So a huge percentage of the people who fall in acute care facilities fall in the middle of the night because either they couldn't find the call bell or the call bell was not answered very fast. So while attempting to get out of bed to relieve themselves they fall, and then, too frequently they fracture a leg or a foot or a thighbone, and they are immobilized and they get pneumonia. The cause of death is pneumonia, when actually the real cause of death is a fall. But because we don't keep the information so that it talks to us, we have all of these conceits about how life works.

If you push back and you say what could we do — open your mind up to the possibility blue-sky ways to solve this problem one way to solve this problem is to have all beds on the floor so if you fall it's only six inches. That's not completely facetious. There are beds that are actually on the floor. Unfortunately, those beds cost $20,000 apiece by the time you get them arrayed properly with all the equipment.

But there's something else you can do that's just an idea. When the patients are admitted, and you identify them as potential fall victims, the other thing you can do by looking at their medical condition is determine how often they should be assisted to the restroom so they don't have an inclination to use the restroom when no one is there to help them. So in this place, where they had been having the average level number of falls for forever, they introduce this idea that's a pretty simple idea. It takes a notation on a patient record that follows the patient. It's hung on the end of the bed, and nurses are pretty religious about following through on this. They went from a rate of 600 falls a year to 100 falls a year.

The good news is they're dedicated to the theoretical limit, which is zero. They're not finished doing observation and inspection and systematic experimentation to find higher levels of optimization, and I must tell you the work that we're doing with hospital centers and a lot of different places around the country is not being done to create solutions. It's being done to help people learn the ideas of continuous learning and continuous improvement so that they can blow these ideas up across their institution and apply them to everything.

An Absence of Leaders


One of the things we found most difficult in doing this work is finding leaders. Actually, there are a lot of places where people believe that they are leaders. By my measure they are not really leaders. For me, leadership starts with ideas. What is a leader? A leader is someone who believes that they are personally accountable and responsible for everything that goes on in their institution, especially for everything that goes wrong. So if you're the head of a hospital and there's a patient fall, if you don't want to know about it until the next day or the next week or two months from now where they have some sort of a staff meeting to talk about things that haven't been going right, by me you're not much of a leader. So that's a precondition. A leader really needs to believe that they are personally accountable and responsible for everything and then act on it.

This is really a problem in health/medical care because in most institutions a lot of the doctors view themselves as independent contractors. I'll tell you a little Pittsburg war story where I worked for a few years trying to create the idea that Pittsburg should be the world's leader in patient safety and perfect medical care outcomes. After making quite a bit of progress in some dimensions of what we were doing, I said we weren't moving fast enough and so I'm going to go to work on something else unless we can agree that we are going to say out loud, in public, in a recognizable way we are going to eliminate medication errors in southwestern Pennsylvania beginning with illegible and incomplete errors.

The Problem with Priorities


The major medical care system leaders said we're not going to let you set our priorities; we are not going to let you do that. That's really a problem. If you are really going to make progress on a complicated problem, I don't think you can permit obvious issues to be on somebody's priority list. In a great organization, the really important things are preconditions; they are not priorities.

It's another important thing I've learned that's a little different from what I learned beginning in 1958 when I was first introduced to the ideas of operations research. I think priorities are a not bad idea, but for things that are important ... Let me tell you what's really important. You don't put breathing on a priority list as if sometimes it's OK not to do it. Right. In the same sense, I believe things that are really important things in organizational life should never be a priority because the term suggests that the order could change. And for really important things there shouldn't be any order; they should be like breathing in and out. So a precondition for beginning to fix the medication pathway is to have an agreement among the customers and suppliers that we are going to eliminate this first-order problem we have with clarity of intent.

Again, it is a piece of what is a trillion dollar opportunity. I'll give you another example. One in every 14 people who goes into a medical care system in this country gets an infection they didn't bring with them. It's really not much of a gift. And there are consequences. It can kill you, and it kills a lot of people every year. If it doesn't kill you, it adds tremendously to the cost of your care. It imposes a burden on your physical mechanism that your body would just as soon not have in addition to whatever other complicated problems you've already got. So it adds to your burden. You're supposed to be there to get rid of a problem and they give you another one. It adds to the cost, and again, it's a cost without value.

Here's the important part. I'll give you some data. We worked on this problem here at Alleghany General Hospital. There was one doctor in particular who began to learn what we taught him as the "Alcoa Production System," but which in truth is the scientific method applied to medicine. But he really got the fever about operations research ideas and wanted to apply them in the three intensive care units that he had direct responsibility for.

You are all numerate folks so you won't mind some numbers. Alleghany is a typical hospital, and its experience up until we started working with this doctor was the average for American medical care. In these three ICUs in the base-year period, there were 1,757 patients that flowed through these intensive care units. And typical to this kind of a unit, in that base-year period, 37 people got what's called a central line infection. That's a line they usually put subclavian when they are going to give you nutrition or blood or whatever. It's only 2 percent, right. Only 2 percent of the patients got an infection and a little over 1 percent of them died. That's what it looks like in American medical care. The first wall of resistance you run into when you start talking about perfect care is people say that 98 percent is really terrific, and it's not possible to get better than 98 percent.

This doctor and his team, though, decided they were going to assign themselves the theoretical limit. So they began by making observations, and they quickly discovered that they had all been trained at different nursing schools and different medical schools, and they all did the preparation and the insertion of lines in a different way, so there was no stability in the laying on of hands. It was, whoever happens to be on assignment today and you happen to be the patient, you got treated based on where they went to school and however they do it. No order.

In the right kind of environment, there's a pre-packaged kit. This is beginning to happen more and more now in American medical care. Three years ago when we began this you could not find this, a pre-packaged kit. When you're going to do a central line insertion, you need some things. You need the line, you need a needle, you need gloves, a gown and a mask. Until we started doing this work, those things in this hospital were wherever they happened to be — around the room or down the hallway or whatever. So the first order of business they agreed that was that there was a medically indicated way to prepare a site, what kind of antibacterial to use and whether to swab it one way or scrub. It turns out scrubbing moves the bacteria around while swabbing takes it off if you do it right.

So they agreed, and this is important, to train all the people on all the shifts who had access to patients to do it the same way. They bought everybody a Blackberry with the understanding that as soon as anyone on the clinical team identified what looked like the beginning of an infection, they a pushed a button and everyone came to observe and try to determine where they had a break in the process that could have caused what appears to be a developing infection. In a year's time — the second year after base year — they had 1,810 patients and six infections. For four of them, they were pretty sure they knew where there was a break in protocol when people didn't follow the procedure. One person died. In the second year, they had about 1,900 patients, three infections and no fatalities.

It's a real, live demonstration of what's possible with the kind of analytic tools that many of you use every day, but in a real-life setting. And then if you branch out a little bit and look at what's happening in American medical care — general proposition — 50 percent of a nurse's time is spent doing non-value added work. It's what we would call repair work. It's unbelievable how busy they are racing around doing things that have to be done because the design of the system is no good. For me, there is no bigger smorgasbord of potential opportunity for the application of your talents and insights then in the practice of American medical care.

Behavioral Issues


We haven't touched on any behavioral issues for human beings where they eat too much or they eat the wrong things or all the rest of that. This is just about using tools and techniques you've already deployed in other settings to the practice of medical care, but with this additional piece — the insistence that where there's a behavioral reason why progress isn't being made that you make it part of your agenda and not make it exogenous to your work. The notion that we're not in charge of behavioral here and we only work with neat stuff that can go into linear programming models is not where the important work is, folks.

The important work engages and intersects human behavior that needs to be improved and learned. I'll tell you a really interesting thing that I wish I had a better answer to and it's this. Once it's been demonstrated clearly that there is a way that will produce unbelievably higher levels of performance however you measure it, the dilemma is how we can deploy and get traction on other people stealing these ideas and putting them to use. To me one of the real dilemmas having been engaged in the automobile industry for the last 30 years or so and watching the non-theft of ideas that clearly produce better products with lower cost and higher reliability, why anyone would resist those ideas is just beyond me, but it is true that they do.

I hope I have excited you to the possibility that there are enormous areas for the application of your time and talent. If any of you happen to be on hospital board I hope you don't let them amuse you about data with length of stay as though that was somehow an important indicator of whether or not they are getting high value out of the resources they are consuming. I hope you wouldn't buy the idea that 98 percent is OK. I hope you wouldn't buy the idea that it's unfortunate but necessary that 1 in 14 people get an infection they didn't deserve and didn't bring with them. I think I will quit with that, but believe me, I just touched on the edges of the topology of what's important and what's possible in this field. I hope you will take me up on engaging your own talent working on these issues.

Table of Contents

OR/MS Today Home Page


OR/MS Today copyright © 2007 by the Institute for Operations Research and the Management Sciences. All rights reserved.


Lionheart Publishing, Inc.
506 Roswell Rd., Suite 220, Marietta, GA 30060 USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com
URL: http://www.lionhrtpub.com


Web Site © Copyright 2007 by Lionheart Publishing, Inc. All rights reserved.