Education and debate

Complexity science
The challenge of complexity in health care
Paul E Plsek, Trisha Greenhalgh

Across all disciplines, at all levels, and throughout the This is the first
world, health care is becoming more complex. Just 30 Summary points in a series of
years ago the typical general practitioner in the United four articles
Kingdom practised from privately owned premises The science of complex adaptive systems provides
with a minimum of support staff, subscribed to a single Paul E Plsek &
important concepts and tools for responding to Associates Inc, 1005
journal, phoned up a specialist whenever he or she the challenges of health care in the 21st century Allenbrook Lane,
needed advice, and did around an hour’s paperwork Roswell, GA 30075,
USA
per week. The specialist worked in a hospital, focused Clinical practice, organisation, information Paul Plsek
explicitly on a particular system of the body, was undis- management, research, education, and director
puted leader of his or her “firm,” and generally left professional development are interdependent and University College
London, London
administration to the administrators. These individuals built around multiple self adjusting and N19 3UA
often worked long hours, but most of their problems interacting systems Trisha Greenhalgh
could be described in biomedical terms and tackled professor of primary
In complex systems, unpredictability and paradox health care
using the knowledge and skills they had acquired at Correspondence to:
medical school. are ever present, and some things will remain P E Plsek
unknowable paulplsek@
You used to go to the doctor when you felt ill, to directedcreativity.com
find out what was wrong with you and get some medi- Series editors:
New conceptual frameworks that incorporate a Trisha Greenhalgh
cine that would make you better. These days you are as dynamic, emergent, creative, and intuitive view of and Paul Plsek
likely to be there because the doctor (or the nurse, the the world must replace traditional “reduce and
care coordinator, or even the computer) has sent for resolve” approaches to clinical care and service BMJ 2001;323:625–8
you. Your treatment will now be dictated by the organisation
evidence—but this may well be imprecise, equivocal, or
conflicting. Your declared values and preferences may
be used, formally or informally, in a shared
management decision about your illness. The solution
Complex adaptive systems: some basic
to your problem is unlikely to come in a bottle and may concepts
well involve a multidisciplinary team. Definitions and examples
Not so long ago public health was the science of A complex adaptive system is a collection of individual
controlling infectious diseases by identifying the agents with freedom to act in ways that are not always
“cause” (an alien organism) and taking steps to remove totally predictable, and whose actions are intercon-
or contain it. Today’s epidemics have fuzzier bounda- nected so that one agent’s actions changes the context
ries (one is even known as “syndrome X”1): they are the for other agents. Examples include the immune
result of the interplay of genetic predisposition, system,4 a colony of termites,5 the financial market,6
environmental context, and lifestyle choices. and just about any collection of humans (for example,
The experience of escalating complexity on a prac- a family, a committee, or a primary healthcare team).
tical and personal level can lead to frustration and dis- Fuzzy, rather than rigid, boundaries
illusionment. This may be because there is genuine In mechanical systems boundaries are fixed and well
cause for alarm, but it may simply be that traditional defined; for example, knowing what is and is not a part
ways of “getting our heads round the problem” are no of a car is no problem. Complex systems typically have
longer appropriate. Newton’s “clockwork universe,” in fuzzy boundaries. Membership can change, and agents
which big problems can be broken down into smaller can simultaneously be members of several systems.
ones, analysed, and solved by rational deduction, has This can complicate problem solving and lead to unex-
strongly influenced both the practice of medicine and pected actions in response to change. For example, Dr
the leadership of organisations. For example, images Simon (box) cannot understand why staff are so resist-
such as the heart as a pump frame medical thinking, ant to a small extension of surgery opening hours. Per-
and conventional management thinking assumes that haps it is the fact that the apparently simple adjustment
work and organisations can be thoroughly planned, to working arrangements will play havoc with their
broken down into units, and optimised.2 own lunchtime inivolvlements with other social
But the machine metaphor lets us down badly systems—be these meeting a child from school, attend-
when no part of the equation is constant, independent, ing a meeting or study class, or making contact with
or predictable. The new science of complex adaptive others who themselves have fixed lunch hours.
systems may provide new metaphors that can help us Agents’ actions are based on internalised rules
to deal with these issues better.3 In this series of articles In a complex adaptive system, agents respond to their
we shall explore new approaches to issues in clinical environment by using internalised rule sets that drive
practice, organisational leadership, and education. In action. In a biochemical system, the “rules” are a series
this introductory article, we lay out some basic of chemical reactions. At a human level, the rules can
principles for understanding complex systems. be expressed as instincts, constructs, and mental mod-

BMJ VOLUME 323 15 SEPTEMBER 2001 bmj.com 625

Education and debate

Henderson have together evolved a system of

Complexity in the life of an ordinary GP behaviour; they have both contributed to the pattern of
frequent visits we now observe. The health centre is
Dr Fiona Simon is a part time partner in a large health also embedded within a locality and the wider society,
centre and the clinical governance lead for her
and these also play a part in Mr Henderson’s
primary care trust. After a busy morning surgery she
goes on to chair a multidisciplinary educational behaviour. A subsequent article in this series will
meeting on a local initiative to establish local asthma explore how medical care for people with diabetes is
guidelines at which an academic expert gives a talk on embedded in wider social and other systems.11 Our
evidence. She emerges from the meeting somewhat efforts to improve the formal system of medical care
irritated that the world presented by the academic is so can be aided or thwarted by these other more informal
black and white. She was surprised to hear herself
“shadow systems.”12 Since each agent and each system
described by a colleague as an “opinion leader and
advocate of evidence based medicine.” In fact, she is nested within other systems, all evolving together
reflects, she found herself agreeing with a group of and interacting, we cannot fully understand any of the
nurses in the audience, who protested that “patients agents or systems without reference to the others.
very rarely fit the textbook case or the evidence based
medicine guidelines.” Tension and paradox are natural phenomena, not
Later, during an overbooked afternoon surgery, she necessarily to be resolved
sees Mr Henderson, a 71 year old widower who has The fact that complex systems interact with other com-
diabetes and little in the way of social support. He has plex systems leads to tension and paradox that can
no new physical problems and Dr Simon notes that never be fully resolved. In complex social systems, the
the patient was told last time to see her in six months’
seemingly opposing forces of competition and
time—but once again he has returned after less than
two weeks. She gives him five minutes and writes “Gen. cooperation often work together in positive ways—
chat” in his record. fierce competition within an industry can improve the
In the evening, there is a practice staff meeting to collective performance of all participants.13
discuss a proposal that the surgery should stay open an Many will sympathise with Dr Simon’s uneasiness
additional 30 minutes over lunch to accommodate about evidence based medicine. There is an insoluble
patients who can only leave work in their lunch breaks. paradox between the need for consistent and evidence
Dr Simon has sent round a memo suggesting that a
different duty team of doctor, nurse, and receptionist
based standards of care and the unique predicament,
could run the service each day. The meeting was context, priorities, and choices of the individual
scheduled to last 20 minutes but goes on for over an patient. Whereas conventional reductionist scientific
hour, and the issue is not resolved; two of the five thinking assumes that we shall eventually figure it all
partners are vehemently opposed and did not even stay out and resolve all the unresolved issues, complexity
for the meeting. “Opening over lunch worked fine in my theory is comfortable with and even values such inher-
brother’s practice,” thinks Dr Simon on her way home.
ent tension between different parts of the system.
“Why the furore among the staff and my partners?”
Interaction leads to continually emerging, novel
behaviour
els. “Explore the patient’s ideas, concerns, and expecta- The behaviour of a complex system emerges from the
tions” is an example of an internalised rule that might interaction among the agents. The observable out-
drive a doctor’s actions. comes are more than merely the sum of the parts—the
These internal rules need not be shared, explicit, or properties of hydrogen and oxygen atoms cannot be
even logical when viewed by another agent.7 For exam- simply combined to account for the noise or shimmer
ple, another doctor might act according to the of a babbling brook.14 The next article in this series
internalised rule “Patients come to the doctor for a sci- considers the application of complexity thinking in
entific diagnosis.” In the example in the box Dr Simon’s healthcare organisations; it will describe how the
partners and staff probably do not share her implicit productive interaction of individuals can lead to novel
behaviour rule—“Try to accommodate patients’ desire approaches to issues.15 The inability to account for sur-
to be seen outside standard surgery hours.” prise, creativity, and emergent phenomena is the major
The mental models and rules within which shortcoming of reductionist thinking.
independent agents operate are not fixed. The fourth
Inherent non-linearity
article in this series—on complexity and education—
The behaviour of a complex system is often non-linear.
will explore this point in more detail.8
For example, in weather forecasting the fundamental
The agents and the system are adaptive laws governing gases contain non-linear terms that
Because the agents within it can change, a complex sys- lead to what complexity scientists have called “sensitive
tem can adapt its behaviour over time.9 At a biochemical dependence on initial conditions,” such that a small
level, adaptive micro-organisms frequently develop anti- difference in the initial variables leads to huge
biotic resistance. At the level of human behaviour, Mr differences in outcomes.16
Henderson (see box) seems to have learnt that the This property of non-linearity appears in all
surgery is somewhere he can come for a friendly chat. As complex systems. Dr Simon, for example, was
this example illustrates, adaptation within the system can surprised by the uproar over her suggestion of a seem-
be for better or for worse, depending on whose point of ingly small change—to remain open an additional 30
view is being considered. minutes during the lunch hour.
Systems are embedded within other systems and Inherent unpredictability
co-evolve Because the elements are changeable, the relationships
The evolution of one system influences and is non-linear, and the behaviour emergent and sensitive
influenced by that of other systems.10 Dr Simon and Mr to small changes, the detailed behaviour of any

626 BMJ VOLUME 323 15 SEPTEMBER 2001 bmj.com

 Education and debate

complex system is fundamentally unpredictable over

time.16 Ultimately, the only way to know exactly what a Low

Degree of agreement
complex system will do is to observe it: it is not a ques- Chaotic
tion of better understanding of the agents, of better
models, or of more analysis.
Co
mp
Inherent pattern lex

Despite the lack of detailed predictability, it is often

possible to make generally true and practically useful Simple
statements about the behaviour of a complex system.
There is often an overall pattern.17 For example, Mr High
High Low
Henderson will turn up periodically in Dr Simon’s sur-
gery until something is done to alter his behaviour. We Degree of certainty

cannot predict the exact timing of his appointments or The certainty-agreement diagram (based on Stacey23)
his chief complaint—nor is this detailed information
necessary to deal with the problem. structures on the planet relative to the size of the build-
Attractor behaviour ers.5 Yet there is no chief executive termite, no architect
Complexity science notes a specific type of pattern termite, and no blueprint. Each individual termite acts
called an attractor. Attractor patterns provide com- locally, seemingly following only a few simple shared
paratively simple understanding of what at first seems rules of behaviour, within a context of other termites
to be extremely complex behaviour. For example, in also acting locally. The termite mound emerges from a
psychotherapy, clients are more likely to accept a process of self organisation.
counsellor’s advice when it is framed in ways that In everyday life many complex behaviours emerge
enhance their core sense of autonomy, integrity, and from relatively simple rules in such things as driving in
ideals.18 These are underlying attractors within the traffic or interacting in meetings. While no one directs
complex and ever changing system of a person’s our detailed actions in such situations, we all know how
detailed behaviour. Relatively simple attractor patterns to behave adaptively and end up getting to where we
have been shown in share prices in a financial market,6 want to go. We shall explore this concept further in the
biological systems (such as beat to beat variation in forthcoming article on management and leadership in
heart rate19), human behaviour (such as Mr Hender- healthcare organisations.21
son’s frequent consulting), and social systems (such as
nurses’ staffing patterns on a hospital ward20). The zone of complexity
Doctors’ behaviour is notoriously difficult to
influence, but, as we shall illustrate in the article on Langton has termed the set of circumstances that call
organisational applications in this series,21 attractor for adaptive behaviours “the edge of chaos.”22 This
metaphors can be used to identify potentially fruitful zone (the middle area in the figure) has insufficient
areas for work. agreement and certainty to make the choice of the next
step obvious (as it is in simple linear systems), but not
Inherent self organisation through simple locally so much disagreement and uncertainty that the system
applied rules is thrown into chaos (figure).23 The development and
Order, innovation, and progress can emerge naturally application of clinical guidelines, the care of a patient
from the interactions within a complex system; they do with multiple clinical and social needs, and the coordi-
not need to be imposed centrally or from outside. For nation of educational and development initiatives
example, termite colonies construct the highest throughout a practice or department are all issues that
lie in the zone of complexity.
Our learnt instinct with such issues, based on
reductionist thinking, is to troubleshoot and fix
things—in essence to break down the ambiguity,
resolve any paradox, achieve more certainty and
agreement, and move into the simple system zone. But
complexity science suggests that it is often better to try
multiple approaches and let direction arise by
gradually shifting time and attention towards those
things that seem to be working best.24 Schön’s reflective
practitioner,25 Kolb’s experiential learning model,26 and
the plan-do-study-act cycle of quality improvement27
are examples of activities that explore new possibilities
through experimentation, autonomy, and working at
the edge of knowledge and experience.
Not all problems lie in the zone of complexity.
Where there is a high level of certainty about what is
required and agreement among agents (for example,
the actions of a surgical theatre team in a routine
operation) it is appropriate for individuals to think in
IIANE PAYNE

somewhat mechanistic terms and to fall into their pre-

agreed role. In such situations the individuals

BMJ VOLUME 323 15 SEPTEMBER 2001 bmj.com 627

Education and debate

relinquish some autonomy in order to accomplish a 7 Stich SP. Rationality. In: Osherson DN, Smith EE, eds. An invitation to cog-
nitive science: thinking. Vol 3. Cambridge, MA: MIT Press, 1990.
common and undisputed goal; the system displays less 8 Fraser S, Greenhalgh T. Coping with complexity: educating for capability.
emergent behaviour but the job gets done efficiently. BMJ (in press).
9 Holland JH. Hidden order: how adaptation builds complexity. Reading, MA:
Few situations in modern health care, however, have Addison-Wesley, 1995.
such a high degree of certainty and agreement, and 10 Hurst D, Zimmerman BJ. From life cycle to ecocycle: a new perspective
on the growth, maturity, destruction, and renewal of complex systems. J
rigid protocols are often rightly abandoned.
Manage Inquiry 1994:3;339-54.
11 Wilson T, Holt T, Greenhalgh T. Complexity and clinical care. BMJ (in
press).
12 Stacey RD. Strategic management and organizational dynamics. London:
Conclusion Pitman Publishing, 1996.
13 Axelrod RM. The complexity of cooperation. Princeton: Princeton University
This introductory article has acknowledged the Press, 1997.
14 Gell-Mann M. The quark and the jaguar: adventures in the simple and complex.
complex nature of health care in the 21st century, and New York: Freeman, 1995.
emphasised the limitations of reductionist thinking 15 Plsek PE, Wilson T. Complexity, leadership, and management in
healthcare organisations. BMJ (in press).
and the “clockwork universe” metaphor for solving 16 Lorenz E. The essence of chaos. Seattle: University of Washington Press,
clinical and organisational problems. To cope with 1993.
17 Briggs J. Fractals: the patterns of chaos. New York: Simon & Schuster, 1992.
escalating complexity in health care we must abandon 18 Schafer R. A new language for psychoanalysis. New Haven, CT: Yale Univer-
linear models, accept unpredictability, respect (and uti- sity Press, 1976.
lise) autonomy and creativity, and respond flexibly to 19 Goldberger AL. Nonlinear dynamics for clinicians: chaos theory, fractals,
and complexity at the bedside. Lancet 1996;347:1312-4.
emerging patterns and opportunities. 20 Sharp LF, Piesmeyer HR. Chaos theory: a primer for health care. Quality
management in healthcare 1995;3(4):71-86.
Competing interests: None declared. 21 Plsek P, Wilson T. Complexity, leadership, and management in healthcare
organisations. BMJ (in press).
22 Langton CG. Artificial life. Proceedings of the Santa Fe Institute. Studies in the
1 Hansen BC. The metabolic syndrome X. Ann N Y Acad Sci 1999;892:1-24. sciences of complexity. Vol 6. Redwood City, CA: Addison-Wesley, 1989.
2 Morgan G. Images of organization. 2nd ed. Thousand Oaks, CA: Sage, 23 Stacey RD. Strategic management and organizational dynamics. London: Pit-
1997. mann Publishing, 1996.
3 Waldrop MW. Complexity: the emerging science at the edge of order and chaos. 24 Zimmerman BJ, Lindberg C, Plsek PE. Edgeware: complexity resources for
New York: Simon and Schuster, 1992. healthcare leaders. Irving, TX: VHA Publishing, 1998.
4 Varela F, Coutinho A. Second generation immune networks. Immunol 25 Schon, DA. The reflective practitioner. New York: Basic Books, 1983.
Today 1991;12(5):159-66. 26 Kolb DA. Experiential learning. Experience as the source of learning and devel-
5 Wilson EO. The insect societies. Cambridge, MA: Harvard University Press, opment. Englewood Cliffs, NJ: Prentice Hall, 1984.
1971. 27 Berwick DM. Developing and testing changes in delivery of care. Ann
6 Mandelbrot B. A fractal walk on Wall Street. Sci Am 1999;280(2)70-3. Intern Med 1998;128:651-6.

A patient who changed my practice

Fifteen years before looking at job

When we recently looked at a survey on chronic This case illustrates two important points: (a) only if
uraemia that included patients for whom there was no doctors ask their patients about exposure may a causal
occupational history, we felt that not enough attention association be established; and (b) only if doctors know
was paid to this factor. We went back to the story of a the potential risk implications of an occupation on
38 year old man who was admitted to our hospital in health may a correlation between symptoms and
September 1976 complaining of severe abdominal exposure be recognised.
pain. His clinical history was unremarkable up to 1961, Although some doctors think occupational diseases
when he first experienced abdominal pain and was are a thing of the past, occupational risks do still exist,
admitted to another hospital. As laboratory tests as do environmental and lifestyle exposure risks (from
showed no abnormalities and his clinical picture pollution, hobbies, habits) such as lead toxicity in
spontaneously improved, he was discharged, but he children and elderly people. Occupational diseases
relapsed with colic-like severe abdominal pain and lack “glamour” for potential clinical investigators, but
hypertension, and he was admitted to the same they are important because they can be prevented.
hospital once a year for the next 15 years. Diagnostic They also serve as models (few workers exposed to
hypotheses were pancreatitis, liver disease, diverticulitis, high concentrations of toxins over short periods) that
and pancreatic cancer, without any confirmation. help in the understanding of environmental diseases
On his admission to our hospital, a suspicion of (large populations exposed to low concentrations over
plumbism was formed. Body lead burden on EDTA
long periods).
mobilisation tests was 1650 ìg (normal value 150 ìg,
So, please ask patients about job experience and
toxic levels > 1000 ìg). He had been working since
other possible occupational or environmental
1952 in a ceramic industry, making hand prepared
exposure, as we have done routinely since looking after
enamel. When somebody asked him why he never
this patient.
talked about his job, he said: “ I did, but the physicians
told me I have not ‘black tooth,’ so I could not be lead Piero Stratta clinician
intoxicated, and I was most probably suffering from Caterina Canavese clinician, Nephrology, Dialysis, and
psychosomatic symptoms or was even a drug addict.” A Transplantation, University of Torino, Italy
renal biopsy showed ischaemic glomeruli and
arteriolosclerosis compatible with lead-nephropathy. We welcome articles up to 600 words on topics such as
Chelation therapy was done for 20 years (body lead A memorable patient, A paper that changed my practice, My
burden was still > 600 ìg in 1992) and halted only in most unfortunate mistake, or any other piece conveying
1998, as normal lead excretion was obtained. There instruction, pathos, or humour. If possible the article
was no relapse of abdominal pain, and he is now should be supplied on a disk. Permission is needed
enjoying good general health, with normal renal from the patient or a relative if an identifiable patient is
function and good blood pressure control . referred to.

628 BMJ VOLUME 323 15 SEPTEMBER 2001 bmj.com