Mediation analysis as a statistical approach in palliative care research


Written by Dr Kelly Jones, Research Fellow, Research Centre for Palliative Care, Death, and Dying

Public health research relies on many disciplinary frameworks: health economics, sociology, demography, public policy, health promotion, and epidemiology. Mixed method approaches lie at the heart of public health research, socio-political research helps to elucidate the why and the epidemiology gives us confidence of the representativeness of the ‘what’. The goal of most epidemiologic inquiry is the elaboration of the causes of ill health by means of statistical examination. (1) Epidemiological research and analysis contributes to public health assessments of population level burden of health problems and prioritisation of public health activities.

Unlike the more ‘straight forward’ methodologies used in clinical epidemiology; randomised controlled trials or case control studies, social epidemiology moves away from a focus on clinical effects to a broader lens of community development, social research, bioethics, and health economics. (2) Social epidemiology, which assumes that the ‘distribution of advantages and disadvantages in a society reflects the distribution of health and disease’ attempts to adjust for the presence of confounding, mediating or interactive factors and relies on complex statistical analyses which account for causal inferences and requires attention to scientific, socio-political, and ethical reasoning. (3) These methods are central to the search for causal determinants of health/disease and to improve public health and provide an exciting and innovative new approach for public health palliative care research.

Of the 60,000 people who die in Australia each year, many do not receive care that reflects their choices or meets their needs. (4) The needs of under-served populations and efforts to improve access for all remains a gap between the vision of high-quality palliative care and the reality that many people cannot access appropriate end-of-life care. Palliative and end of life research has traditionally focused on medical, clinical, and health care approaches. There is limited Australian research on non-clinical mediators in understanding palliative care risks and limited approaches exploring the mediated effects of the potential socio-political or eco-political risk factors on outcomes.

Mediation analysis can assist to empirically explore the relative causal contribution of any given effect on an outcome. Mediation analysis assesses the relative magnitude of different pathways and mechanisms by which an ‘exposure’ or an intervention (eg: a healthcare program) may affect an outcome and estimate both direct and indirect effects of such policy or approach. Statistical methods and theoretical approaches are now available to assess both mediation and interaction simultaneously to better understand the impact of any intervention on a mediator when it is not possible to alter an exposure and trying to understand why any given intervention succeeded or failed. For example, ‘what might be the role and magnitude of the effect of care giver burden in palliative care programs?’ Or ‘what is the effect of carer health literacy on better EOL outcomes?’ or “what is the actual effect on well-being of palliation in the presence of other determining factors’? Qualitative research can highlight the ‘problematic’, but decisions as to whether altering an health program or intervention will have the intended effect can be answered using these kinds of advanced statistical methods and also provide estimates of the relative effect of altering or improving programs and interventions and allow economic assessments to be made around efficiencies. (5)

The growing ageing Australian population and varying State, Federal or localised policy interventions may contribute to elevated rates of professional or caregiver burden and may impact on the quality of life of those they care for. The trend in growth of outpatient palliative care programs intending to relieve unnecessary pain and suffering has not been augmented with research which explores health professional, carer or family characteristics as relevant factors, potentially affecting the effectiveness of palliative care programs or to determine the extent of caregiver burden as a possible mediator on the effectiveness of home-based palliative care programs. Additionally, exploring the effects of non-patient related modifiable risk factors may highlight previously unexplored effects of program outcomes which contribute to well-being and better experiences of palliative patients. Public health research in palliative care that combines a diverse set of research methodologies will contribute to better understanding of the complexity of circumstances of dying and provision of palliative care in populations, particularly in the climate of rising chronic disease. There are some exciting opportunities for new methods in population based palliative care research to generate a more detailed understanding of the relative effects of social determinants and health policies on better quality of life outcomes.

1. Rothman KJ, Greenland S, Lash TL . Modern Epidemiology. Lippincott, Williams and Wilkins, Philadelphia, 3rd edition, 2012.
2. Berkman LF, Kawachi I. A historical framework for social epidemiology. In: Berkman LF, Kawachi I, eds. Social epidemiology. New York: Oxford university press, 2000: 3–12.
3. Honjo K. Social epidemiology: Definition, history, and research examples. Environ Health Prev Med. 2004;9(5):193‐199. doi:10.1007/BF02898100
4. Australian Healthcare Association. Exploratory Analysis of Barriers to Palliative Care. Australian Government Department of Health. Summary Policy Paper
5. Weed DL Methods in epidemiology and public health: does practice match theory? Journal of Epidemiology & Community Health 2001;55:104-110.

Posted in

Leave a Reply