Entangled complexity: Why complex interventions are just not complicated enough

Introduction: recognising complexity

In 2000 the (UK) Medical Research Council (MRC) produced a document ¹ which exemplified and pioneered what has been called the ‘appropriation of complexity’ ² within health care. The Council's framework was amongst the first to frame contemporary health interventions as inherently complex, involving overlapping modes of operation, and hence challenging any straightforward measurement or evaluation of their impact. Over the next few years, the framework was widely-cited, yet also at times aroused critical comment. Subsequently, in 2008 it was replaced by ‘New Guidance’, ³ which explicitly addressed a range of criticisms, and has since become a central document for those concerned with designing and testing health interventions.

But what does this mobilisation of complexity signify? Both the 2000 and 2008 documents largely avoid directly confronting the specific question of what constitutes ‘the complex’. The potential importance of this omission has been raised in debates in this journal and elsewhere, where questions of the ‘essential’ nature of the concept of complexity have begun to surface.^2,4,5,6,7 For example, whilst Paley approaches the notion in terms of a narrow and austere ‘explanatory category’, requiring rigorous policing, ² Greenhalgh and colleagues treat complexity more as a ‘world view’ with regard to the nature of change.^4,6

Our own discussion similarly draws on broader approaches to complexity within and beyond the natural sciences. Yet, rather than attempting to define what complexity is or is not in the abstract, we wish to draw attention to the consequences of the ‘fit’, or lack thereof, between the rhetoric of ‘the complex’ and current research cultures of public health and public policy more generally ⁸ – specifically, the claims of evidence embedded within the Randomised Controlled Trial. Overall, our argument is quite simple: a richer appreciation of complexity and the commitment to the RCT as the ‘gold standard’ of evidence (to which the MRC Framework is ultimately directed) are ultimately incompatible.

After briefly reviewing the MRC's guidance, we highlight two interconnected issues that emerge from a genuine engagement with complexity. The first concerns confronting its ‘ecological’ character, while the second shows how this inherently challenges the conventional standards of reproducibility and fidelity that govern the development and testing of health interventions. We argue that both these issues must be taken seriously if the ‘challenge of complexity in health care’ ⁴ can be engaged with meaningfully. Two possible solutions arise from this. Either such interventions should abandon the use of the term ‘complexity’ to more accurately reflect how studies pragmatically reduce interactions to a restricted number of variables; or, more radically, if there is to be meaningful engagement with complexity, alternative means to test, evaluate and represent interventions and their effects need to be developed.

The emergence of complex interventions

The 2000 MRC publication was ground-breaking in its acknowledgement that many novel health interventions did not consist of singular elements. Consequently, researchers were no longer being asked to isolate and assess the efficacy of one component over another, as in the case of a classic drug trial. Rather, they were increasingly being forced to evaluate components across a range of different domains – bio-medical, organisational, psychological and social – all at work simultaneously. This, the Framework argued, presented an analytical problem requiring fresh thought from the research community. To make sense of the implications of these interconnected components, it proposed a ‘stepwise approach’ to intervention development, improvement and testing, and suggested that there should be space for a mixture of research methods. It was asserted that qualitative research was particularly useful for refining the nature of an intervention and predicting its possible impact so that insights generated could be used in the stages that followed. The implicit assumption was that by adopting a mixed methods approach, ‘the complex’ could be adequately captured and understood to enable the final design of the intervention and trial.

There can be no doubt that the MRC's framework has been influential and has contributed to the generation of a wide range of innovative research approaches. Subsequent studies conceptualised complex interventions in a wide variety of ways. For example, a study which followed the guidelines to develop a standardised pathway for the treatment of heart failure conceptualised complexity as a property arising from the multiple components and trajectories of established clinical path-ways. ⁹ In contrast, another initiative piloting diabetes education groups posited that the complexity emerged from an array of factors acting in combination, and that this variability should be accommodated rather than be subject to any kind of standardisation. ¹⁰ In response to this variability, the 2008 revision of the MRC document was designated ‘guidance’ rather than a framework and dropped the titular reference to RCTs. It also incorporated other important changes. In particular, it described less linear models of design and data extraction in order to allow for greater input and feedback at various stages. However, the revised text largely continued to avoid the specific question of what constituted a ‘complex’ intervention. Our general point, therefore, is that while researchers increasingly agreed that health interventions targeted on practices and behaviours were inescapably complex, no consensus on what ‘complex’ meant was forthcoming. ¹¹

Reconceptualising complexity: from mechanical to ecological

The 2008 MRC guidance reproduced the assumption that complex interventions are those that are comprised of ‘several interacting components’. It is also implied that such multiplicity can effectively be captured via a process of identification and enumeration. Thus, whilst this approach acknowledges the significance of the interactions between elements, the term ‘components’, coupled with an emphasis on measuring them as discrete elements (whether behaviours, variables, or outcomes etc.), produces a depiction of complexity that is essentially mechanical. Such an approach simply cannot accommodate the idea that together such elements form a dynamic and integrated system.

In contrast, within a wide range of scientific fields such as chaos theory and systems biology, it is now commonly emphasised how phenomena cannot be reduced to constituent variables. Rather, key properties of the interactions are emergent and contingent upon one another, while there are also always significant elements that remain uncertain and unknown.^12,13 Similarly, social theorists have explored complexity in terms of links between the individual and society, or the behaviour of groups versus individuals, arguing that such relationships are inherently dynamic and dialogical, and, as a consequence, ultimately irreducible to their parts.^14,15 More recently, these enquiries have expanded further to include the parts played by material, nonhuman actors within networks of complexity. ¹⁶

The notion of ecology encapsulates much of this current thinking on complexity in both the sciences and social sciences. Integral to this perspective is the recognition that alteration in one part provokes change throughout the system, and that the ‘system’ can never be isolated from its ‘environment’. A classic example was provided long ago with Charles Darwin's celebrated ‘entangled bank’ account of the hedgerow. ¹⁷ Darwin describes how all of the life-forms that inhabit such a setting rely on both ‘internal’ factors, such as the other living things around them, and on ‘external’ factors, for instance, climatic conditions or farming practices. The introduction of a new element inevitably establishes itself as part of the whole in multiple and varied ways. The general point here is that engaging seriously with complexity from this ecological standpoint is neither simply a matter of attempting to develop more ‘sensitive’ or ‘accurate’ strategies to capture more and more data, nor is it, as has sometimes been suggested in the context of designing health interventions, to do with pursuing ever more sophisticated causal models to identify those ‘active ingredients’ which might ultimately drive a complex problem. ¹⁸ Rather, this appreciation of complexity begins with recognition of the sheer range of relevant factors and the significance of the variability of local conditions.

Capturing ecological complexity

The ecological approach we describe conceptualises complexity not as a closed system of identifiable constituent parts, but rather as a dynamic and constantly emerging set of processes and objects that not only interact with each other, but come to be defined by those interactions. In contrast, research toolkits such as the MRC's Guidelines, ostensibly designed to guide researchers to deliver and assess an intervention ideally through a RCT, require the isolation of measurable parts and hence the sacrifice of any genuine commitment to complexity. If we genuinely want to acknowledge complexity rather than merely note, in a mechanistic way, that health-related issues are very complicated, we must find a way of engaging with its dynamic variability. In other words, the challenge is how to go about studying complexity without fully un-ravelling it.

We want to suggest that social scientific approaches, beyond simply asserting the value of ‘qualitative methods’, can aid the development and implementation of intervention research that offers alternative, but potentially complementary, forms of evidence and scientific standards. In accordance with the notion of ecological complexity, knowledge generated by the social sciences concerns itself with more than just the production of detailed and accurate accounts of the social world. With a commitment to holism and interconnectedness, the key strength derived through research methods such as ethnography is the attention paid to tracking and identifying the processes and relationships through which ‘particular events, practices and things’ ¹⁹ become meaningful and important. Thus, these approaches can assist in the identification of experiences and practices that are otherwise liable to be left out of accounts, including those associated with the research process itself.

Social scientists such as Lock acknowledge the importance of this holism in the study of health and medicine. She adopts the term ‘local biologies’ to refer to ‘the way in which biological and social processes are inseparably entangled’. ²⁰ Such particularistic visions are not just an anthropological preserve: they are also increasingly becoming standard in more traditional biomedical circles. ²¹ These approaches accept that studies of particular entanglements provide a partial account of how elements emerge as meaningful and relevant in relation to each other. But they nevertheless do some justice to the complexity of complexity, and recognise the many ways in which a complex world can at once be encountered and intervened upon. ²²

But crucially this also implies that a commitment to the generalisablility and reproducibility of things that are inherently complex is naïve and misplaced. Complex health interventions will always be highly variable in both character and form due to the dynamic nature of their constituent parts and the inevitable adaptations that emerge from their implementation in local contexts.

Conclusion

The ‘turn to the complex’ has clearly been a necessary and productive response to the changing nature of common health problems. However, drawing on theoretical approaches from beyond medicine, and in particular from the social sciences, we have argued that genuinely acknowledging and addressing complexity requires more than the simple adoption of an ever-expanding number of variables or array of statistical tests. Rather, we have suggested that the notion of ecological complexity perhaps best captures the specific dynamics of complexity in the domain of health and illness. Such an approach emphasises not only how different elements come together to produce a system, but also how such elements become meaningful and change as they travel through and interact in particular contexts.

This view of complexity, however, also challenges the extent to which RCT thinking and standards can aid in the development of high quality, appropriate and effective health interventions. We have argued that rethinking complexity in these ways offers an alternative and complementary account of how causal relationships emerge and are made meaningful, but such an account demands new methods through which they can be assessed. Only when such evaluative frameworks have been developed can the complex nature of interventions fully emerge and be put to work to improve health.