Research report no.8 2002
Children's health and development: new research directions for Australia
edited by Ann Sanson
1. New ways of causal pathways thinking for public health by Fiona Stanley, Ann Sanson and Tony McMichael
'States of health exist in people; people form societies; any study of the attributes of people in relation to health outcomes is also one of the manifestations of the form, structure and processes of social forces.' - Susser 1973: 6
'Modern epidemiologists still do studies in populations, but they do so in order to study decontextualised individual risk factors, rather than to study population factors in their social and historical context.' - Pearce 1996: 680
This paper is written as a background to the concept of establishing a national research collaboration to enable better and more constructive research to underpin public policies which aim to improve family, child and adolescent health and wellbeing. It was initiated from an epidemiological perspective with input from a behavioural scientist interested in developmental psychopathology. It concludes that multi-disciplinary approaches are required to properly address complex pathways to a variety of outcomes, only some of which are specifically health related.
There is currently a debate (mostly in the United States literature) about causal thinking which is yet to have an impact on much of mainstream epidemiology and public health and certainly on government departments who may be setting policy. And certainly it is far away from the view of the media who continue to love simple answers and single risk factors. To paraphrase H.L. Mencken: 'To every complex problem there is always a simple answer - and it is always wrong.'
The paper is also written to encourage broader thinking around what kind of study or studies are needed to advance best a national research and policy agenda for families in Australia.
Multi-disciplinary approaches for complex pathways
Epidemiology is one of the main scientific disciplines underpinning public health. Modern epidemiology arose in response to the epidemics in developed (economically rich) countries of so-called chronic diseases of the 20th century such as cardiovascular disease, lung cancer and peptic ulcer. It employs a fairly rigid paradigm of analyses of risk factors in individuals, using cohort and casecontrol studies, without considering how and in what contexts such risk factors might have developed. There tends to be an underlying naive assumption that there are true universal individual level risk factors that transcend culture, class and context (McMichael 2000).
Many papers have described the associations of disease outcomes with social or economic circumstances, and have then proceeded to control for them in analyses which focus on more proximal risk factors in the causal pathway.
Indeed, some of these risk factors could be thought of as early signs of the disease. There has been intense methodological development of criteria to judge whether a risk factor may be causal, and sophisticated mathematical techniques to sort out the independent effects of proximal risk factors - all essentially in isolation from a thoughtful causal pathway model.
Modern epidemiologists seem to have forgotten that to understand the causes of disease in a population it is important to study its historical and social context and how diversity or cultural norms influence pathways. Mathematical models concentrating on proximal risk factors (that is, those close to the individual and/or close in time to the onset of the disease) actually preclude causal pathway analyses that involve more distal risk factors (that is, those more distant from the individual, such as social or cultural factors, and/or occurring earlier in time). This is because in a multi-variate model proximal risk factors are usually so powerful (strongly associated) that the more distal risk factors do not appear significant (Stanley, Blair et al. 2000). By 'controlling for confounding' we not only ignore the earlier part of the causal path, we also often ignore the causal sequences and interactions which may be so important.
As Susser (1998) has so forcefully pointed out, it seems we have a myopic overemphasis on proximal causes of disease and a concomitant neglect of upstream or distal factors that culminate in the final causal chain.
This situation does not apply only to physical health outcomes. The significance of distal influences in causal pathways leading to behavioural and emotional maladjustment is equally evident and (at least until recently) equally neglected. The complexity of interactions between factors at different levels of analysis poses methodological and statistical challenges to attempts to understand the pathways. An example from the Australian Temperament Project of pathways to antisocial behaviour (Figure 1.1) illustrates some of the complexities (Smart et al. 1999).
In the Figure, 'proximal' and 'distal' refer to a time dimension - here, from Grades 2 to 10 (bottom of Figure), as well as type of influence (left of Figure). An example of a proximal influence is oppositional behaviour at Grade 8. An example of a more distal influence is family/social context at Grade 2. The arrows indicate the direction of effects. The first numbers on the arrows show the standardised beta weights, indicating the strength of the direct relationship between the two variables. Those in italics show the univariate correlations, indicating the total between-variable relationships (including both direct effects and effects exerted through indirect pathways). Only significant paths are shown.
Attention to only the proximal influences on antisocial behaviour, namely prior behaviour problems, school difficulties, parental monitoring, and deviant peer associations at Grade 8, ignores the complex interactions between these and other factors at this relatively late stage. Furthermore, the pathways by which these characteristics have become entrenched in earlier years, and which involve a variety of factors at all four levels of influence, are ignored.
This particular analysis ignores a great deal: first, biological components, and influences in the very early childhood years; and second, the fact that all these factors are embedded in a rapidly changing socio-cultural context. These broader contextual factors (such as high levels of family stress and insecurity, increased unemployment and poverty, a rise in racist sentiments, and so on) undoubtedly contribute directly and/or indirectly to antisocial outcomes. An even more complex diagram than Figure 1.1, and a research paradigm of challenging complexity, would be required to elucidate this more complete picture.
Individual risk factor epidemiology ignores how significant societal changes have created ecologically adverse environments which are having, and will continue to have, profound effects on global health and human development. Thus, modern epidemiology has failed to address the complex and everchanging social, behavioural and ecological pathways. Only recently has it started to realise that molecular and genetic markers could be used to advantage to help elucidate these pathways, rather than being rejected as basic biomedicine and nothing to do with 'proper' epidemiology or public health.
At a conceptual level, developmental psychology has undergone a similar paradigm shift to that being promoted here for epidemiology. Up to 1970s, there were two apparently incompatible paradigms, a mechanistic (behaviourist/environmental) one which saw an individual's development as being shaped by the environment ('nurture' being all important and the individual being viewed as basically passive); and an organismic (biological/trait) model which saw development as the unfolding of genetic/biological makeup ('nature' is dominant and the environment is passive).
In the last 30 years an integration and expansion of these has occurred which mirrors newly emerging epidemiological paradigms, and builds on the earlier opposing views. The new paradigm, most commonly referred to as a transactional model, posits ongoing interactions between influences intrinsic and extrinsic to the child across time as the driver of development. The newly emerging field of developmental psychopathology adopts this model as its theoretical foundation in seeking to understand the complex pathways to psychological adjustment and maladjustment. However, it remains methodologically and analytically poorly equipped to account for the influence of the broad political, social, cultural and ecological environment, and thus continues to shed little light on the influence of powerful distal factors such as poverty and exclusion.
Towards a new research paradigm
These criticisms have profound implications for both research and prevention agendas. Modern epidemiologists would argue that there have been many successes with the individual risk factor approaches in prevention and treatments (vaccination, smoking and lung cancer in males, incidence of heart disease, peptic ulcer management, folic acid and spina bifida, cot death and sleeping position).
However, the success of risk factor epidemiology has been more limited than many claim. For example, on a global basis the achievement of the public health movement has been to shift the problems (such as smoking-related illnesses) from rich to poor countries, or from the rich to the poor within countries. This is a direct effect of the limitations of an epidemiology based on individual factors (for example, tobacco smoking) rather than on population factors (for example, tobacco production, advertising and distribution, and socio-economic influences on consumption) (Pearce 1996). As another example, concentrating all of our efforts on reducing the use of illegal drugs by individuals to prevent suicide and mental ill-health in young people, while ignoring the complex chains of population influences, will ensure that drug taking and other poor outcomes will continue to rise, particularly in the most disadvantaged groups in our societies.
Thus we are suggesting that modern epidemiology as currently practised is unable to solve the socially based public health problems. Since much of public health is socially influenced, current epidemiological research is failing public health.
The elucidation of the code for the human genome and the proliferation of genetic and molecular biomedical research makes this situation of more concern (McMichael 2000). Not only is there a huge international push to explain most diseases and risk factors as 'genetic' - 'the gene for achieving social support!!' - but if such new research is done and applied in complete ignorance of the global societal-cultural impacts it could have, and of the marked economic inefficiency of molecular genetic screening as a population strategy (Vineis et al. 2001), then we are more likely to be hindered by this new knowledge than helped by it. Biological determinism always appears to be popular in eras of political conservatism.
As our own studies in Western Australian child health and the Australian Temperament Project have shown, the social, community, family, biological and economic influences identified for complex disease pathways appear similarly important for a range of other educational, psycho-social, behavioural and criminal outcomes (Zubrick et al. 2000; Prior et al. 2000). Causal pathways thinking opens up the concept of multiple outcomes from single pathways (multifinality) and multiple pathways to single outcomes (equifinality). The fact that these influences mostly act in complex cumulative and interactive ways poses enormous challenges for both research and policy.
This situation is becoming more urgent as the social, biological and ecological environments in which we live have been changing rapidly over the last two to three decades. While some of these changes have obviously been beneficial, we have observed increases in various major disease categories and disabilities (except for infectious diseases other than HIV, cancers and some causes of intellectual disability) (Stanley 2001). These include mental health disorders, asthma and allergy, diabetes and neurological problems such as the cerebral palsies. The increases have been so substantial that the levels of morbidity can only really be tackled by preventive strategies since the health and psycho-social care system cannot meet the demand for treatments and services, and, in any case, for many of these diseases we do not have many really effective treatments.
The great changes which have occurred in the social and demographic structures of communities and populations and in the environment add further complexity and a sense of urgency to our work. Changes in work patterns, family structure, and the social security system, the consequences of globalisation and environmental degradation, the rise of racist sentiments: all these and more have potential to adversely affect developmental health and wellbeing. Simplistic research paradigms which ignore the multilevel and complex nature of pathways, and policies and interventions which focus on a limited number of 'risk factors' at some point along these pathways, are not going to serve us well. Hence we need to move rapidly, but carefully.
It is evident that an era of truly collaborative inter-disciplinary research paradigms is urgently needed. The United States National Institutes of Health have called for 'new research that is multi-disciplinary, longitudinal, and that embraces an ecological approach to the study of child development' (their workshop participants included developmental psychologists, economists, educational researchers, evolutionary biologists, medical epidemiologists and sociologists) (NIH RFP 2000: 5).
We envisage a large research agenda. For example, for social issues to be researched, we need to tackle conceptual issues (what are the relevant attributes of groups?) and practical ones (how do we create data sets with sufficient level specific detail or power?); the theoretical (what is the appropriate unit(s) of analysis? how best to analyse causal pathways?) and the methodological (how can lag times between exposure and health status be incorporated into the analysis, while at the same time allowing for changes over time in status or circumstance?). The difficulties in accurately measuring such things as societal changes and other distal risk factors may explain why they have been excluded from aetiological models.
We need to be challenged to tackle problems at higher levels of aggregation. This issue is crucial as we think about how best to use existing data bases and embark upon new cohort studies. So instead of dividing our disciplines into narrower, competing ones (to use the example of epidemiology, into social, infectious disease, molecular and genetic epidemiology, and so on), we need to bring in new approaches that serve the endeavours of all the collaborating disciplines.
The implications here are considerable. We need to look at this with the eyes of those with whom we want and need to collaborate - psychologists, sociologists, economists, environmental scientists, neuroscientists, geneticists, educational researchers. This represents a chance, not only for epidemiologists, but for all of us, to begin a dialogue and the process of learning to 'see through others' eyes' in the cause of improving child health and wellbeing.
These issues are also of profound importance to the development of policy and where best to put efforts and resources to make the greatest improvements. Causal pathway thinking opens up many more possibilities for prevention, many of which may be more effective and cheaper than more proximal solutions which may be too close to the disease outcome to influence it significantly. For example, the most effective preventive strategies for improving low birth weight and infant death rates in Indigenous mothers may be to tackle the disempowerment, despair, discrimination and dislocation of Indigenous communities, rather than going into such communities with an anti-smoking or nutrition program targeted at women, who may feel further victimised and undermined.
Reigel (1990) describes the process of development as occurring through an ongoing dialectical interaction between four levels of analysis - the innerbiological, the individual-psychological, the outer-physical, and the sociocultural. At present, each discipline adopts one of these levels as its province and, at best, attempts cursory analysis of the other levels. Yet, with our current state of knowledge and in our own particular 'socio-cultural context', it is apparent that we cannot well serve public health on our own. Coming together will not only enable each discipline to better serve public health, but working together will ensure that the social and medical sciences are used appropriately to influence a social policy agenda that includes health, wellbeing and education.
Conclusion
In summary, when we take a causal pathways perspective, it becomes clear that we need to establish a national partnership on developmental health and wellbeing for a variety of reasons.
We now know that pathways to many health, educational and psycho-social outcomes are similar to each other, and similar in their complexity; and that most poor health, educational and psycho-social outcomes are strongly associated with measures of social distress.
It is also clear that effective, enduring solutions will only be developed with the sorts of paradigms which have been proposed here, in which different disciplines bring the expertise needed to develop the most important questions and suggest the most appropriate methodologies to answer them. Further, existing Australian databases and studies (such as record-linked population data, cohort studies, surveys) which have been established by the different discipline groups are at the moment under-used, and could be made available to researchers across all disciplines to inform new research. By working collectively and adopting a 'consortium' approach, researchers in a national partnership could increase the capacity for research and government funding.
Finally, a national partnership has the potential to bring together the best minds nationally to underpin the planning for studies of Australian children with broader and better research agendas than those presently available. A partnership of this sort holds the promise for contributing to better outcomes for Australian children in the future.
References
McMichael, A.J. (2000), 'Germs, genes and greenhouse gasses: The changing landscape of population health', Queen Elizabeth, The Queen Mother, Lecture, Presented to Faculty of Public Health Medicine Annual Scientific Conference, Scarborough UK.
NIH (2000), 'The Science and Ecology of Early Development (Seed)', Program Announcement, National Institutes of Health, Bethesda, Md.
Pearce, N. (1996), 'Traditional epidemiology, modern epidemiology, and public health', American Journal of Public Health, vol. 86, pp. 678-683.
Prior, M., Sanson, A., Smart, D. & Oberklaid, F. (2000), Pathways from Infancy to Adolescence: Australian Temperament Project 1983-2000, Australian Institute of Family Studies, Melbourne.
Riegel, K. (1976), 'The dialectics of human development', American Psychologist, vol. 31, pp. 689-700.
Smart, D., Sanson, A., Toumbourou, J., Prior, M. & Oberklaid, F. (1999), 'Longitudinal pathways to adolescent antisocial behaviour and depression', Paper presented at the Life History Research Society Conference, Hawaii, September 23-25.
Stanley, F.J. (2001), 'Centenary article: Child health since Federation', in 2001 Year Book Australia, ABS Catalogue No. 1301.0, Australian Bureau of Statistics, Canberra.
Stanley, F.J., Blair, E. et al. (2000), The Cerebral Palsies: Epidemiology and Causal Pathways, Blackwells, London.
Susser, M. (1973), Causal Thinking in the Health Sciences: Concepts and Strategies of Epidemiology, Oxford University Press, New York.
Susser, M. (1998), 'Does risk factor epidemiology put epidemiology at risk? Peering into the future', Journal of Epidemiol Community Health, vol. 52, pp. 608-611.
Vineis, P., Schulte, P., McMichael, A.J. (2001), 'Misconceptions about the use of genetic tests in populations', Lancet, vol. 357, pp. 709-712.
Zubrick, S.R., Silburn, S.R. et al. (2000), 'Mental health disorders in children and young people: Scope, cause and prevention', Australian and New Zealand Journal of Psychiatry, vol. 34, pp. 570-578.