Looking Inside The Black Box

21 November 2013 by Tania Browne, posted in Epidemiology

Epidemiology is having an identity crisis.

It started over 20 years ago, when there was a sudden flurry of papers and editorials wondering who epidemiologists were, what we were doing and what would come next. And it's clear that even 20 years later, we still haven't come to any conclusions.

You see, all of those newspaper articles you read on your commute? About how lemon pith or licorice might give you cancer but fresh coffee or learning to play the tuba might stop it again, you know the ones. They're all our fault. Admittedly the media makes them a bit more alarmist and brash, but they're still based on journal articles and press releases that epidemiologists write. As three colleagues from the Center For Disease Control wrote in a letter to The American Journal of Public Health:

"Like members of other successful institutions in our society, we have an image problem. For some, we've become the kvetch of science - the regular bearers of bad news or, even worse, the regular contradictors or modifiers of our own previous findings. In the process of trying to communicate our findings, we too often scare people, confuse them or inadvertently promote guilt."

The cancer stories are classic examples of "cause and effect", or what is sometimes termed "black box" epidemiology. We know the particular "exposures" that go into the box and we know the likely "outcomes" but we have no real idea of the mechanisms, the magic that takes place inside. We just work on lessening the exposure and hoping the outcome will be less cases. But the subtleties of our statistical analysis have become so fine that the true effects may be tiny, not anything to worry about at all.

This approach can have results, of course. You don't have to know the true mechanism to help the problem. The first great era of public health improvement was based on the misguided notion that disease was caused by "Miasma" - bad smells. Luminaries such as Edwin Chadwick and John Snow knew nothing of microbes and their effects on the body, they just thought that the poor, overcrowded streets of London needed to smell nicer. In producing the sewer system that ensured no waste contaminated drinking water, the engineer Joseph Bazalgette inadvertently improved London's health when all anyone wanted to do was get rid of the stink. Sometimes, it just happens that you get the right effect for the wrong reasoning.

But somewhere later in the Victorian period, where the pioneering notion of "public health" was overshadowed by the germs we discovered inside us, epidemiology started to lose its grand public remit and look at the individual, isolated from their circumstances. We quite literally searched inside ourselves. A rift opened up between epidemiology based on the grandiose idea of public health, and epidemiology searching for causes of disease and figuring out how to stop its progression. That rift has been widening ever since.

In the post-war era, epidemiology has been all about the transition. In the wealthiest nations we have left behind the worst eras of infectious disease and moved on to something just as worrisome - the chronic diseases attributed to wealth and ageing populations. Cancers, heart disease, diabetes, lung diseases and mental health all have their place at the epidemiologist's metaphorical lab bench now, where there used to be measles, typhus and pertrussis.

It all started so well. Big new methods of study like cohorts and case control studies, with limited blunt statistical methods - but it didn't matter, because the effects being measured were in some cases huge. When Richard Doll and Austin Bradford Hill first concluded that tobacco smoking caused lung cancer in 1956, nobody had to quibble about whether their result was statistically significant or not (well, the tobacco companies tried....) But the blunt instruments of early studies didn't stay that way for long. Faced with the "web of causation", the idea of confounding, the concepts of bias and misclassification, epidemiologists refined their study designs and introduced an arsenal of statistical subtleties... subtleties that now produce subtle results. And it's those you read on your morning commute.

And all the while, with epidemiology in the wealthiest nations concentrating on the cause and effect of wealth and long life, we have infectious diseases forgotten yet still rife in middle and low income nations. The TB, malaria, syphilis and HIV epidemics baffle us. We struggle on to eradicate polio. It's the opposite of the black box. We know the mechanisms, but have no idea how to end it, little sense of the big picture. With some diseases, the Black Box has become Pandora's Box with no clear way to close the lid.

Epidemiology has become polarised. On the one hand we're stuck in the biological fallacy - searching for exposures in individuals so that we can educate and empower them to take better care of themselves, without even a glance at their role in society. And on the other side, we're using the big data and information systems available to gather information on a global scale without ever really making that data count to real people in real situations. Both of these extremes are threatening to crowd out the centre ground of the discipline and leave nothing else there.

What we need is to re-establish the role of epidemiology in public health, not just as an academic discipline taking place in a lab. Of course epidemiologists can't change the public face of health alone. It needs political will, resistance to commercial pressures such as tobacco and food corporations, economic support and some real changes in our attitude and even the structure of society - finally accepting that an individual, even when educated, can only do so much and society needs to play its part.

In a pair of papers by Mervyn and Ezra Susser, published in The American Journal of Public Health in 1996, they proposed a view of epidemiology that could be based on the study of eco-systems. They argued that, just as biology can't be ruled by a single set of laws above a certain level because of the influence of the eco-system around any organism, epidemiology was no longer able to be constrained by simple "cause and effect". Society is (and thus epidemiology needs to be) influenced by demographics, economics, the built and natural environments, the effects of climate change and many other things. People cannot be isolated from those. Your race, class and age or gender are more than boxes to tick on a list of possible confounders. They have influenced your whole life.

Mervyn Susser used the analogy of Chinese boxes, but I rather prefer the idea of Russian Dolls. The smallest is the individual, the next up your family, next your home, next your town... And so on and so forth. All co-existing in the same space, all needing to be examined as part of your life and accounted for. This seems to me, relatively new to the discipline, to be a good analogy. So let's talk about the ecological approach. Let's remember the public health remit, the great experiments of Snow and Chadwick, the vision of Bazalgette. Let's not let the discipline die of entropy on a dusty academic lab bench.


3 Responses to “Looking Inside The Black Box”

  1. Jeff Reply | Permalink

    I think you have inadvertently hot on one of the problems many have with epidemiology and public health. There is an underlying value judgement that the evidence is being used to improve people's lives. But in many cases, the evidence is used to hector and bullying people into behaving in a way that the public health community have deemed 'the best' but may be against the preferences of individuals and society.

    In particular, the 3rd last paragraphs is an example of this. The belief is that if only people were just educated they would do the right thing. And if they don't do the right thing then we need to make them do so - the current vogue approach is 'nudge theory' for which my own discipline of economics can be blamed. What never seems to be asked is, if we educate people and they still make unhealthy choices why should we make them change? Public health in your example of John Snow was about improving the health of the public through public interventions. More recently it seems to be about improving the health of the public through changing individuals.

    So while epidemiology can be a valuable tool for identifying the problems and generating hypotheses it seems to me that there is a need to question the value judgements underlying how that information is used. (Again, true in economics, where many take efficiency to be a value-judgement free objective when it is anything but). Anyway, that's just some initial thoughts and I need to go teach economists about value judgements now!

    • Tania Browne Reply | Permalink

      Thanks! It wasn't inadvertant though, I'm going to be expanding on a lot of the stuff here over the coming months, especially around study design but a lot on what we can and can't make the public do with the information as well.
      I'm one of those people who is, the more I'm nagged, the less likely do something just because I'm contrary that way. The nudge policy certainly puts me off going to my doctor and I dislike the patronising way health reminders are often couched in the media. I think if we really wanted to put our money where our mouths are then we could do a lot more, but it would dependably annoy a lot of big business and individuals too,through "overlegislation" and "nanny state" interventions. But hey! They work. It's a very fine line to balance the health benefit of telling people what to do with being seen as interfering with individual freedom. Changing the law on seat belts, for instance, has saved untold lives yet is a good example of people doing something reluctantly.

Leave a Reply

four × = 4