A Clinical Problem as Genetic Algorithm Game:
Reducing C-sections - the first "cut" in the story

Beginnings of an effort to use computational tools of complexity for clinical improvement

Told by: Brenda Zimmerman and Curt Lindberg
Reflections by: Brenda Zimmerman and Curt Lindberg

Illustration of:

• genetic algorithms
• pattern recognition skills
• modeling complexity
• non-linearity
• interdependencies
  
  

At a meeting to discuss caesarean section rates in their community, Stephen Larned, Vice President for Medical Affairs at Maine Medical Center, began to see links between the topic of the meeting and the presentation he had heard by John Holland at a VHA conference the week before on genetic algorithms and modeling.

"What if we frame the caesarean section problem as a game for genetic algorithms? The objective of the game is to have the lowest caesarean section rate. The pieces of the puzzle are the expectant mother and her attitudes, maybe those of her family, thirty different obstetricians, five different ways of managing pain, two or three different ways of managing induction, the stage of the nurse’s shift, etc. You multiply all those variables together and you get 45,000 different ways of managing one patient."

Larned looked for a partner in this endeavor and found a doctor who had developed software using concepts from neural networks and genetic algorithms to address clinical care options in a mental health hospital. Larned invited his colleagues to meet with this doctor to discuss the software’s potential application to the caesarean section problem.

The software required a definable input, a definable output and a large supply of experience. For most of the management challenges at Maine Medical Center the software would not be useful because these requirements are not met. However, Larned argued, the requirements were met for some of the clinical challenges including reducing the number of unnecessary caesarean sections. "The argument is that there are non-linear links between some of the input variables that will not be obvious to us." The software is intended to help one see the interdependencies or non-linear links. Larned saw this as an opportunity to intervene at the system at the point of greatest leverage.

Larned had not had a chance to implement the software to the caesarean section problem yet. He was excited by its potential for a clinical improvement but also for learning more about complexity.

For the last several months, Larned had been facilitating monthly meetings with colleagues interested in complexity. Much of their discussions had been on the management side of health care. Over time, there developed a core group of people who were reading and thinking about complexity science. The people who attended the genetic algorithm presentation were not members of this core group. Larned saw this as an opportunity to attract new people to the complexity perspective.

"One of the things that’s so exciting is this would be a quantitative hard core clinical application. That would be extremely gratifying. The core issue is how to provide care in a better way."

Reflection: Larned’s story is about two themes: accessing different members of his organization using different methods or "hooks" to complexity and using the mathematical modeling of complexity to improve their capacity to recognize patterns and improve patient care. It is this second theme which is unique among our inventory of complexity initiatives.

The mathematical lessons from complexity are well suited to understanding many clinical interventions. The focus is on the interdependencies which are not intuitively obvious. Treating wellness in a holistic sense is not new. It is indeed ancient wisdom. Yet the modeling process of genetic algorithm and neural networks may provide some insights into why and how the ancient wisdom works. This has the potential to increase the leverage from health care interventions - or in some cases the understanding when not to intervene. Genetic algorithms and similar modeling techniques from complexity have some untapped lessons for health care.

We captured Larned at the beginning of this exploration. The outcome of the story has yet to emerge.

Next | Previous | Return to top

Copyright © 2001, Brenda Jane Zimmerman and Curt Lindberg. Permission
to copy for Educational purposes only. All other rights reserved. Excerpt
from "Stories of the Emergence of Complexity Science in US Health Care" -
paper to be published in a book edited by Eve Mitleton-Kelly of the London
School of Economics.