In my last post, I reflected on the big picture of theory in demography. I thought about questions like: What is the difference between models and theories? (None, if one subscribes to the model-based view of science), How “big” does the theory/model need to be? (As small or as big as you want it to be; it is okay to model unique events), and How does one begin to build social science theory in the era of data science? (With a combination of narrative explanation and empirical prediction). In this post, I start reviewing demographic theory relevant to mortality convergence in the enlarged European Union by summarising Abdel R. Omran’s theory of the epidemiologic transition and its development over time. One caveat before we start: my interest is in the high-income countries, so the review that follows mostly glosses over the work done related to other contexts.

There are several demographic theories that aim to explain and predict long term trends in mortality in an international comparative perspective, even if they do not mention mortality convergence. A good place to start is the well-known theory of the epidemiologic transition (Omran, 1971). Epidemiologic transition (ET) theory was developed in response to the more general theory of demographic transition, which aims to explain how the modern pattern of low mortality and low fertility arose from the high mortality and high fertility regimes of the past (Caldwell 2007). Omran’s summarises the basic point of ET theory in his 1971 paper as follows:

Conceptually, the theory of epidemiologic transition focuses on the complex change in patterns of health and disease and on the interactions between these patterns and their demographic, economic and sociologic determinants and consequences. An epidemiologic transition has paralleled the demographic and technologic transitions in the now developed countries of the world and is still underway in less-developed societies. Ample evidence may be cited to document this transition in which degenerative and man-made diseases displace pandemics of infection as the primary causes of morbidity and mortality.

I find that a key strength of ET theory is its compelling narrative. We start in the pre-modern past, where extreme fluctuations in the mortality rate were common. There were very high peaks during pandemics, famine, and war, interspersed with high mortality plateaus, where death occured due to chronic malnutrition, endemic disease, and complications during delivery. This puts mortality in the drivers’ seat when considering overall population dynamics as the large (several-fold) variation in mortality rates can easily overwhelm the more narrow variation in fertility rates.

Omran describes two changes in mortality patterns that bridge pre-modern and modern societies. The first is a reduction in the frequency and range of variation in mortality rates. This corresponds to less frequent and less intense pandemics, famines, and wars. The second is a reduction in the mortality rate during the plateaus, from 30 per 1000 to about 10 per 1000; this reduction was non-linear: slow at first, then fast, then slow again. These changes in mortality occured in three colourfully named stages:

  1. the Age of pestilence and famine, with high and fluctuating mortality rates due to infectious disease, malnutrition, war, and maternal mortality,
  2. the Age of receding pandemics, when the mortality plateaus become lower and mortality peaks grow less frequent, and
  3. the Age of degenerative and man-made diseases, when mortality stabilises at a new low steady state and most deaths are caused by cardiovascular disease and cancers.

Why did these changes in mortality occur? Omran proposes three drivers that underlie any populations’ journey through the stages of the transition:

  1. ecobiologic determinants, which are related to the epidemiologic triad (agent, host, environment) and could represent a decrease in vector populations,
  2. socioeconomic, political, and cultural determinants, which include improved standards of living and health behaviours (including personal hygene and nutrition), and
  3. medical and public health determinants, which include specific preventive and curative treatments such as public sanitation, immunisation, and antimicrobials.

The importance of these drivers differs between countries and periods, as does their precise nature. I would add that the drivers overlap and interact (perhaps synergistically), so trying to identify the “one key driver” to explain a particular transition makes little sense. For example, Omran points out that the epidemiologic transition in western countries began in the 1700s, before many of the medical and public health interventions now available were developed. Perhaps this means that there is a measure of redundancy between the different drivers, and populations could move between stages on the merits of one or two alone.

Omran highlights that the decrease in mortality is particularly pronounced in children and women of reproductive age. He speculates that these individuals were perhaps more physiologically susceptible to infectious disease and malnutrition, so as the environmental changes made these less frequent, women and children benefited most. This assertion is difficult to support with evidence at the individual level as both groups are historically underrepresented in medical research.

The differences in subpopulation gains altered the composition of populations, which in turn influenced reproductive behaviour and socioeconomic status of societies. Interestingly, the argument related to reproductive behaviour is supported by two simulations, the REPSIM model (Ridley et al. 1967) and Heer’s 1966 simulation of births necessary to assure the survival of sons, both of which are interesting examples of the use of simulation use to generate theory. As for the socioeconomic impacts of ET, Omran notes that lower mortality and infectious disease burdens lead to higher effectiveness of labor and increase economic productivity. This is mostly due to the fact that an individual’s production tends to exceed their consumption roughly between the ages of 20 to 65 (by 1970s standards).

Finally, Omran recognises that the way the epidemiologic transition unfolds varied between places in terms of pattern, pace, determinants, and its societal consequences. He identifies three overarching models to encapsulate this variation:

  • the classical/western model,
  • the accelerated model (e.g., Japan), and
  • the contemporary/delayed model (the rest of the world).

Omran updated the ET theory twice. The first “preliminary update” in 1983 begins to hint at a “maturation” of the third stage in western contexts. The “maturation” corresponds to further increases in life expectancy due to decreased cardiovascular mortality that occured in the 1970s and 1980s in the west. These developments are likel associated with a decrease in the risk factors, earlier diagnosis, improved treatment, and intensive care. However, Omran notes that the net mortality decrease may have been tempered by the rising incidence of other chronic diseases (cancer), as well as lingering or resurgent communicable diseases (e.g., influenza, Legionnaire’s disease).

The second major update was published posthumously in 1998. It admits the possibility of transition reversals, incorporating the experience of life expectancy decline in Russia and Ukraine during the collapse of the USSR. It also provides a much more thorough treatment of developing countries compared to the 1971 formulation. For western countries, it formalises the “maturation” of the third transition stage as a new fourth stage, the Age of declining cardiovascular mortality, ageing, lifestyles modification, emerging and resurgent diseases and introduces a fifth stage, the Age of aspired quality of life, with paradoxical longevity and persistent inequities. This fifth stage is particularly interesting. Omran predicts that it will occur in the mid-21st century and be characterised by paradoxical longevity (longer lives in poorer health), increased socio-economic polarization, and emergence of novel diseases. I would argue that Omran underestimated the pace of change in population health in high-income settings over the last 20 years, and that the fifth transition is a phenomonon that started already in the 2000s and 2010s. For example, we are already experiencing emerging diseases in the form of rising obesity and deaths of despair, both of which are associated with the socio-economic characteristics of individuals and the places they inhabit. Similarly, the interest in health expectancies and long-term care reform reflect the presence of paradoxical longevity in the EU.

Where does ET theory help with thinking about mortality convergence? I think it helps in at least three ways.

  1. It identifies broader societal and economic developments (for example norms and living conditions that jointly determine health behaviours) as key determinants of prevailing mortality conditions and rejects focusing solely on prevailing medical technology. A rich understanding of politics, economic development, social norms, and prevailing technology are all required to understand how and why mortality changes over time.
  2. ET theory reminds us that what we often imagine as a linear progression from social determinants to population health outcomes is better understood as a feedback loop. The population structure that emerges from one transition stage influences the determinants that lead to the next stage. This makes the usual modelling approaches less useful, especially if one is interested in how mortality developed over decades. The way forward is through analysis of shorter periods and chaining these together (sort of sub-stages), or using an modeling approach that can handle feedback, like lagged terms or simulation.
  3. In its admission of failure to identify a generic set of stages and processes to describe ET worldwide, the theory highlights the contingent nature of the epidemiologic transition. The message for me is that studying the development of mortality in well-defined contexts may be more productive than trying to construct a universal narrative. This well-defined context, however, may still include several countries with a shared social and economic history.

There are also elements missing from ET theory that are very important for studying mortality convergence. Mortality convergence is fundamentaly a relational phenomenon that occurs between populations. However, ET theory includes no substantial treatment of the relationships between populations and how these relationships influence how the epidemiologic transition unfolds. To be fair, there are two exceptions: the observation of the existence of large-scale public health programmes in developing settings in the original paper (although the transnational origin of these programmes remains unattributed), and a paragraph on the differences in ET between colonial and colonizing nations in the final paper. By and large, however, ET theory seems to assume that national populations develop largely independently from one another. This is an important omission and we can easily imagine that this is not the general case, especially after the second era of globalisation that followed the Second World War. Countries today routinely influence each others’ mortality trajectories through benchmarking and policy learning in international and supranational organisations (EU, UN system, World Bank, OECD) and economic integration, including trade, investment, and common market arrangements. These relationships exist both between countries within high- and low-middle-income groups, as well as between them.