In my last post, I highlighted a previous conversation on the importance of the wider social context in health policy analysis. A part of the wider context includes politics, so we may call this endeavour political epidemiology. In this post, I will delve deeper into the emergence of this approach, provide a sense of its relationship with the broader field of social epidemiology, and reflect on the methodological challenges in political epidemiology.

I think Link very nicely articulates the motivation behind social epidemiology and why it matters today:

Ironically, perhaps, the ascendancy of social factors occurred because of the tremendous growth of biomedical knowledge and technology. To a greater and greater extent, new knowledge about risk factors and new technology to address those risks has placed the control of disease and death in human hands, and it is those human hands that determine whether new information and technology is deployed, at what rate the uptake occurs in different places and at different times, and who benefits most and least from our new-found capacities. Social factors have become more important precisely because epidemiological and biomedical knowledge has shifted the causes and consequences of disease from fate, accident, and bad luck to factors that are under some human control. When humans obtain control, it is their policies, their knowledge, and their behaviors that shape the consequences of epidemiological and biomedical accomplishments, and thereby extant patterns of disease and death.

In “Locating Politics in Social Epidemiology”, Muntaner et al. observe that social epidemiology has became depoliticised. Many social epidemiology studies today focus on the proximal social characteristics of observed individuals or groups and how they relate to health (e.g., poor neighbourhoods are associated with poor diets). However, these studies often do not ask why these associations exist and why the neighbourhoods are poor in the first place. The classic works of social epidemiology have been very different. For example, Engels’ The Condition of the Working Class in England and Virchow’s Report on the typhus epidemic in Upper Silesia were very explicit in their identification of political mechanisms (e.g, industrial capitalism), which distributed the risk factors between population groups (e.g., unsafe working conditions and poor housing) that in turn caused poor health outcomes (e.g., premature mortality).

Bambra et al. suggest how the depoliticisation unfolded. They find shifting definitions to be at root of the issue, which led to two conflations. The first conflation is of health with the healthcare system and the provision of medicine, which reduced health to the acquisition of products and services that alleviate disease and resulted in the medicalisation of health inequalities. The second is the conflation of politics with the activities of governments. This narrowed the scope of social epidemiology and health policy to a study of what the government does to influence the distribution of healthcare resources and the oversight of health professionals. Upstream determinants of health such as safe housing, secure employment, or time for active leisure, and - especially important - mechanisms that distribute these resources among individuals, thus fell out of scope, even though they exert important influence on population health.

How does politics influence health? The usual answer would be through the policies enacted and enforced, especially health policy. However, according to Muntaner et al. and in light of the above broader definitions of health and politics, this approach fails to account for four key facts:

  1. the influence of politics on health includes politics in the form of social movements (e.g., the functioning of NGOs and unions), international relations (e.g., sanctions and wars), political upheaval (e.g., revolutions or coups), and education and information exchange (e.g., digital and health literacy);
  2. the distribution health-relevant policies and institutions is not random but rather associated with political patterns;
  3. the effect of political indicators (e.g., left-wing government) remains significant even after adjusting for specific policies; and
  4. individuals are affected by several policies (or institutional arrangements) simultaneously with their effects not easily disentangled, a phenomenon Beckfield et al. call institutional imbrication or overlap.

The way political epidemiologists commonly implement their theoretical perspective is by conducting cross-country comparative studies. Most commonly, they would divide countries into clusters according to their welfare regime (see Esping-Andersen’s classification) or dominant political tradition and compare the average life expectancy between the clusters. There are also studies that exploit sudden changes in political and economic systems, such as the transition to democracy in Europe after World War II, where one can compare health outcomes before and after the change. Some recent studies use more continuous and/or precise measures of politics such as democracy indices (e.g., Freedom House Ratings), economic variables (e.g., foreign investment, debt, economic concentration), or the presence or absence of certain institutions (e.g, civil society participation) to construct comparisons between countries and over time.

However, these approaches have important limitations. Pega et al. propose a typology of political epidemiology studies, which divides them into three groups: the welfare regime approach, the politics approach, and the individual policy approach. They argue that the first two categories are not conducive to establishing causal effects for three reasons: (1) they tend to focus on cross-sectional comparisons as sudden and externally caused variation in these variables is rare, (2) they rely on static and perhaps overly general assumptions about group membership (especially the welfare regime approach), and (3) generally do not provide a mechanism of effect for the associations they study (e.g., how precisely does a transition to democracy cause improvements in life expectancy?). They argue that these limitations make these two approaches less suitable for informing policy as they fail to identify specific political changes to advocate for.

What is more, political epidemiology studies tend to struggle with path dependency in the sense that specific political regimes, institutions, and configurations are present in a limited number of countries, which tend to be similar in other ways as well, thus making statistical control of potential confounding variables unfeasible. As Muntaner et al. illustrate:

Social Democracies are Northern European countries and Late Democracies are Southern European countries. As a consequence of their different historical trajectory, they are characterized by different patterns of risk factors: less smoking in the South; protection via the traditional “Mediterranean diet;” and a less stressful lifestyle. In addition they have different historical trajectories in spite of their relative geographical proximity. In the post World War II period, Late Democracies suffered from non-democratic “right wing” or Fascist regimes, while Social Democracies enjoyed stable democracies. [C]ountries with different political and welfare state traditions are extremely difficult to compare due to the confounding effect of cultural and economic factors or historical trajectories on population heath. In this situation, comparing results (as in a formal statistical approach) is not sufficient. We need to explain different initial conditions and the same process in different contexts.

One proposed solution to these challenges is the single policy approach where the researcher focuses on understanding the population health impact of a single well-defined policy. The problem with this more pragmatic approach is that while it is conducive to causal inference, it can severely limit the scope policies considered (see previous post). This can lead to ineffectual tinkering with the provisions of existing policy (e.g., Obamacare) instead of considering radically different options that could address the underlying issues (e.g., Medicare for all). So how do we overcome the impasse when one approach promises to tackle the root of the issue yet is insufficiently precise, while the other one is precise but tends to ignore the big picture?

Diez Roux suggests that complex systems approaches may be a useful way of thinking about the effects of upstream policies and institutional arrangements. Interactions between populations and their political contexts may be imagined as complex systems:

Complex systems typically include heterogeneous agents at various levels, contact structures between agents, adaptation, nonlinear dynamics, and stochasticity. These features lead to the emergence of patterns at various scales. The recognition that the health of individuals and populations is the manifestation of a system in which biology interacts with environments and individuals interact with each other and with environments over time is a key element of the concept of population health.

This is useful, because systems approaches both allow and force researchers to think about the broader context in which they work. In particular, systems models require that we formalise our mental models of the situation, including all assumed interactions between individuals, between individuals and their environments, feedback loops, and delays. Commonly used system modelling techniques then allow researchers to interact with a simulated version of their system, gaining a better sense of its behaviour under different circumstances, perhaps allowing for an understanding of why and how the system has managed to resistant our efforts to change it (see Sterman, or wait for my post summarising his paper).

Very interesting! In my next post, I will delve deeper into how epidemiologists have thinking and working with systems modelling.