The Appropriation of Complexity Theory in Health Care

During recent years, complexity theory has proved popular in the health care literature, with complex systems being referred to in a wide range of contexts, including inter-professional education, medical education, clinical governance, evidence-based practice, consultations in general practice, public health nursing, achieving change in health service organizations, shared care for patients with long-term mental illness, research on maternity care, services in elderly care, decisionmaking in nurse management, and problem-based learning in dentistry and occupational therapy among many others. Almost all the authors who write on this topic cite one or more of a series of articles in the BMJ,^1–4 a series which draws on Plsek's contribution to Crossing the Quality Chasm: A New Health System for the 21st Century, ⁵ and which has effectively sanctioned what might be called the ‘received interpretation’ of complexity in health care.

In view of the influence of this series, it is unfortunate that the BMJ series is so misleading: complexity embodies three fundamental principles, and the series fails to recognize two of them. It may well be true that complexity theory can be applied successfully to both health care and health care organizations; however, if it is to be useful in this context, it is necessary to be clear what it entails.

It is crucial, in the first place, to recognize that complexity is an explanatory concept. To invoke complexity when referring to complex systems is to proffer a form of explanation. It is to explain how a particular structure or pattern of behaviour has arisen. This is absolutely clear in the scientific sources^6,7 as well as in the philosophical and popular literature. How do slime mould cells aggregate to form a single organism? How do ant colonies develop? How do termites create air-conditioned nests? How do flocks of birds manage to fly in echelon? Why does closing streets or adding traffic lights sometimes improve traffic flow? Why are revolutions more likely to happen when long-term oppressive rule is relaxed? These questions all refer to examples of patterns or structures for which complexity explanations have been proposed.

Second, a complexity account always takes the same form. The global pattern/structure is to be explained by the behaviour, locally, of individual units interacting with their immediate neighbours and with their environment. To put this more formally: successive states of the system, globally defined, are determined by previous states, locally defined. The function which links the state at t₁ to the state at t₂ is defined as a set of stimulus–response rules (‘If… then…’) applying to individual units (whether cells, ants, termites, birds or drivers) whose behaviour conforms to this function. Normally, future global states are unpredictable, given only initial conditions and the state transition function; however, the system's states are still completely explained by the starting condition and the rules governing local behaviour. ⁷

The third principle is a corollary of the second. Complexity explanations account for global order by specifying the local behaviour of units which have no awareness of the order thereby being produced, and which have no intention to produce it (in many cases, they lack the ability to formulate intentions: termites do not design their nests, birds do not intend to fly in formation). This is to sever the default connection between order and design. Complex systems are examples of a kind of order which is not the result of plans, intentions, goals or values. This is what ‘self-organization’, an easily misunderstood term, means. The system ‘self-organizes’ in the sense that no intentional force or design is required. Order just happens as a consequence of individuals conforming to local stimulus–response rules. Complex systems mimic design, they are not instances of it.

The BMJ writers partially understand the second of these principles, but seem not to have grasped the other two. Consider self-organization – the absence of any link between order and design. The general problem in applying complexity explanations to the social world is that people are intentional beings who are frequently, though not always, aware of a connection between their individual behaviour and collective outcomes. This creates a tension between one necessary feature of complexity, self-organization and the new context to which it is being applied, social action. It becomes very tempting, therefore, to re-interpret ‘self-organization’ and make it mean something else – namely, the way in which people are capable of organizing themselves without the direction of a leader, manager or external authority. Those who succumb to this temptation (and many health care writers do, following Plsek and Wilson) ³ apply ‘self-organization’ to the process whereby people get together, discuss their ideas, agree goals and plan a strategy. However, this is clearly not self-organization, because it restores the link between order and design which complexity thinking disrupts. To suggest that a group of people working collaboratively, without hierarchical supervision, fulfils the requirements of complexity is to miss the point. In a self-organizing system, there is no plan, and no collaborative working. Individuals act unilaterally, without reference to the global order that they (might) create. They do not share goals. They do not have meetings. That is the core of complexity. Intentionality, the restoration of the order–design connection, is the antithesis of self-organization.

Another technical term whose meaning was changed in the BMJ series is ‘attractor’. In complexity thinking (more usually, dynamical systems theory), an ‘attractor’ is a purely mathematical idea, defining the end-state of a system in a space which represents all the mathematically possible states of the system. A ‘point attractor’ is a unique end-state (for example, a pendulum coming to rest); a ‘limit cycle attractor’ refers to a system which oscillates between two or more specific values; and a ‘strange attractor’ characterizes a system whose end-state consists of an indeterminate range of values which never repeat themselves, but which remain within a certain range. This mathematical concept is very different from ‘intrinsic motivation’, ³ a ‘sense of purpose, direction or vision’, ⁸ and the ‘values or behaviour people are drawn towards’. ⁹ We should avoid confusing a technical term, ‘attractor’, with the colloquial sense of ‘attract’. It is not unlike supposing that π has something to do with baking.

More fundamentally, however, there is no acknowledgement in the BMJ series that complexity provides a type of explanation. Instead, stimulus–response rules are replaced by ‘minimum specifications’, which are designed to bring something about by providing ‘direction pointing, boundaries, resources, permissions’. ³ As illustration, they cite the 10 ‘simple rules for the design of the 21st century healthcare system in the United States’. ⁵ An example is Rule 2: ‘Customization based on patient needs and values. The system of care should be designed to meet the most common types of needs, but have the capability to respond to individual patient choices and preferences.’ Such ‘rules’ are no doubt to be welcomed, but they are clearly not stimulus–response rules of the kind associated with complexity explanations. ¹⁰ Indeed, they are not even explanatory; instead, they are normative principles, or general guidelines, not something capable of explaining a known outcome. Guidelines such as these, designed to bring about an unspecified, utopian change, cannot be assimilated to explanatory rules.

The BMJ authors' failure to recognize the explanatory function of complexity is associated with, and may in part be derived from, their (apparent) belief that complex systems cannot be explained. They rightly observe that the behaviour of complex systems is unpredictable, but mistakenly infer from this that it cannot be ‘understood’. ‘Ultimately, the only way to know exactly what a complex system will do is to observe it: it is not a question of better understanding of the agents, or better models, or more analysis.'’ ¹ But that is exactly what it is a question of. It is true that ‘the only way to know exactly what a complex system will do is to observe it’; but it does not follow from this that we cannot explain its global structure by studying individual units, analysing the stimulus–response rules they follow and constructing a more accurate model of local interaction. ¹¹

It is possible that underlying this mistake is a conflation of complex systems with dynamical systems. Plsek and Greenhalgh ¹ note that the behaviour of a complex system is often non-linear and that, consequently, they have ‘“sensitive dependence on initial conditions”, such that a small difference in the initial variables leads to huge differences in outcomes'’. This characteristic, however, is more usually associated with dynamical systems (which can also be non-linear). The main difference is: whereas the state of a complex system at t₂ is determined by its state at t₁, locally defined, the state of a dynamical system at t₂ is determined by its state at t₁, globally defined. Some dynamical systems are therefore highly sensitive to initial conditions – depending on the mathematical function that governs their behaviour – in a way that complex systems typically are not. ¹²

Like other contributors to the literature,^9,13–15 the BMJ authors suggest that complexity theory is a source of fruitful metaphors; and it might be that they would argue for a metaphorical understanding of complexity. However, it is difficult to see what is left of complexity, even metaphorically, if its main function – to provide explanations – is overlooked, if the default link between order and design is restored, if mathematical concepts such as ‘attractor’ are psychologised, and if complex systems are conflated with dynamical systems.

The ‘received interpretation’ of complexity in health care, derived from the BMJ series, needs radical amendment, since it is founded on several misconceptions. I have no doubt that it is possible to apply complexity to health services in useful and interesting ways; ¹⁶ however, successful application will need to focus on proposed explanations of known outcomes and adopt a non-intentional approach to the understanding of organizational behaviour.