How do social structures influence health? This question is at the heart of my research project. It is also a question of some complexity that is not amenable to a naïve, atheoretical approach. To appropriately define the exposure in this type of inquiry, a reasearcher needs to make decisions about what they mean by social structures and what social structures are relevant to health outcomes. Lucky for me, there is already a rich literature on the topic of political epidemiology of health on which I can draw. A lot of this literature is empirically rooted in the comparative analysis of the effect of welfare regimes on heath outcomes, hence the example in this post draws on this area of research.

An interesting case study to start with political epidemiology is the Espelt / Muntaner et al. versus Lundberg debate in the IJE in 2010. Espelt et al. published a cross-country comparison of European welfare regimes and analysed how these were associated with health inequalities among the older population. Critically, the authors construct the exposure on the basis of the political party that was in power for longest after 1950; the party in power was the identification mechanism for welfare regime. Lundberg takes issue with this, arguing that this may lead to misidentification since the specific welfare systems implemented by ostensibly similar parties (e.g., social democrats) vary between countries and changed over time. He supports this point by arguing that this type analysis requires an explicit mechanism through which we assume welfare regimes act on health and proposes the following formulation:

Ultimately, I would argue, it is the resources available to people that will be of importance for the levels of and inequalities in health in a country. These resources are generated within the family, in the market and also through the welfare state. Welfare state institutions will thereby contribute to people’s resources, either directly through transfers and services or indirectly through policies that affect people’s possibilities to generate resources in the market. The degree to which welfare state institutions do so and the extent to which this in turn is linked to health and health inequalities is the key issue, therefore. Consequently, it is features such as the coverage and generosity of cash transfer programme like unemployment insurance, sickness insurance, family support or pensions that should be in focus if we are to find out what it is about welfare states that are important to people’s health. Or, for that matter, the availability and quality of services provided.

The authors respond by arguing that a narrow focus on policies alone ignores important parts of the story:

Politics affects health outcomes through policies, of course. We do not disagree on this. But politics is the starting point. And our concern is that this dimension has been the least studied part of health policy research.

They also cite a body of work that supports their contention of the association between politics sensu lato and health outcomes, beyond policies alone. However, they did not explain what these additional pathways between politics and health outcomes are until Lundberg takes them to task for this omission in his rejoinder. The final response from the study authors fleshes out the various ways politics influences health outside the narrow confines of policy (emphasis mine):

Politics could affect population health via other social processes such as, for example, grassroots organizing, social movements, wars, strikes, protests and non-government organizations. […] Even if we accept that policies are the only means by which politics have an effect, one should be interested in why certain egalitarian policies cluster together in certain societies and not in others. Policies are not randomly distributed in societies; they follow political patterns. In that sense, the root causes (political and economic arrangements) of the causes (single policies) are a necessary part of a deeper explanation of social epidemiology, such as when income inequality and poverty (indicators of economic structure) might determine proximal social determinants such as crime rates, lack of social networks, lack of exercise facilities, lack of healthy food stores, alcohol stores, open illicit drug markets, lack of public transportation or lack of access to health care. […] Another benefit of the political, in addition to policy, comparative approach involves the likely inter-sectoral origins of health inequality reduction. It is likely that egalitarian policies involve the ‘synergic’ effect of egalitarian policies from different sectors (health, labour market, social services). Such inter-sectoral effects are more likely to arise in certain political traditions than others due to political-policy coherence. In addition, the evaluation of the effect of any single policies in such context might be inadequate (cannot deal with their interaction) or very difficult to conduct due to their complexity.

Moreover, they introduce the concept of political epidemiology:

Although we concur with most of Dr Lundberg’s considerations, there is a larger, important issue that underlies our exchange. Namely, the underlying discussion about ‘political epidemiology’ is one of technocracy vs social justice, an issue that pits technical supposedly value-free public health policies vs policies rooted on a set of political views (e.g. egalitarianism). As any technology such as public health deals with decisions about how to change society, it involves values regarding what constitutes right and good policy to improve population health. Thus, as opposed to ‘just policy’ approaches to social epidemiology and health policy, we conclude that it is more insightful to study politics than to keep them in the background.

Where does this leave us? I believe that the final rejoinder makes a compelling case for a holistic approach to comparative health analysis that explicitly includes the prevailing power relations in a society, even if we are “technocratically inclined” to focus on the contents of policy alone. I am convinced that a decontextualised policy analysis is almost always impossible, since the effects a specific policy will depend on the characteristics of the population it acts upon, since people embody the broader socio-economic environment, and on the other policies that act upon the same population.

For example, the effects of a health insurance policy on health outcomes will depend on whether we are working in a corporatist or universalist system, as this will determine the target population of the policy and thus shape its impact. It will also depend on whether there is access to good quality health services, since even perfectly covered healthcare means little if the care is of poor quality. Finally, even the most generous insurance coverage and world-leading healthcare providers may be undone by poor health literacy, where the population does not recognise a need for healthcare intervention in time.

Certainly, this level of context dependence may not be a problem when a study aims to inform policy in very circumscribed contexts or when we are satisfied with low generalisability of our findings. But international comparative health research usually has higher aspirations. As the authors write:

[T]he focus on national health policies that defines most the US health services research leads to a narrow set of policy alternatives ‘and to a lack of understanding of their political origin’ (e.g. Navarro). A policy approach misses why the US government still ignores ‘single payer’ options because the political trajectory of the USA does not lead to such egalitarian universal policies. How many health service researchers in Europe might be excited by the prospect of spending years studying alternative arrangements that consolidate the role of private insurers such as the new Obama reform? Very few we presume.

In the end, a focus on mechanisms and “opening the black box” of politics does not really lead us to the analysis of individual policies and programmes, no matter how well we can define them. The same policy may produce very different results in different contexts. This fact requires us to grapple with the messy complexity of societies in which the policies and programmes work. Specifically, it requires a careful selection and analysis of contexts to compare and methods that can cope with the interconnected nature of the sociotechnical systems we hope to understand and improve.