Coulda-Woulda-Shouldas: Counterfactuals in Epidemiology

What if you hadn't been a cheapskate and had just bought the damn bike lock? Suppose you'd never said hi to that girl you thought was way out of your league? Whenever something eventful occurs, we tend to wonder how different things would be if only ‘x’ had not transpired..

We do the same thing all the time in epidemiology. We pick an event or 'exposure' and wonder whether your result or 'outcome' would have differed significantly had you been unexposed. In other words, was x a causal factor in determining the subsequent series of events? For example, an individual might wonder whether he would have still developed lung cancer had he never smoked and, on a larger scale, the epidemiologist might wonder whether the percentage of lung cancer cases in the population might have differed significantly had all those smokers never smoked at all. Essentially, we consider the implications of the counterfactuals.

In our current, factual universe, we have a population with its unique characteristics (demographic features, size, lifestyles etc.) and a percentage of this population tends to come down with the illness under study. We may hypothesize that a given exposure, smoking, is associated with lung cancer. We may then look at the frequency of smoking in the population and see whether lung cancer is more common among smokers than among non-smokers. However, even if we find that the statistics support our hypothesized association, we cannot say with certainty that this association is necessarily causal. For this, we would need a counterfactual world where, counter-to-the-fact, there had been no smoking. (In this parallel counterfactual universe, everything is the same except there is no smoking.) If there is significant contrast between what happens in the real, smoking world and what happens in this hypothetical, non-smoking world, then we have a stronger foundation for claiming that there is a causal link between smoking and lung cancer.

We estimate odds and risks and other hypotheticals with these two universes in mind, but the results must be interpreted with care. Even if you've eliminated smoking, you may be able to reduce smoking-related lung cancer but the prevalence of lung cancer may not change appreciably: if people who would otherwise have smoked were already genetically predisposed to develop lung cancer, then they might still get it… albeit some time (a few days?--several decades?) later.

So we weigh the odds and try to figure out our best bet: Maybe if you’d had the bike lock you'd have been out $20 but you'd still have your $200 bike. But maybe, just maybe, another thief would've come armed with boltcutters and stolen your bike anyway and you'd have been out $220. Maybe if you'd never said hi to that girl she never would've become your one true love… or maybe because you said hi you missed your chance to say hi to your real true love who was right behind her that day. It's easy to drive yourself crazy with the coulda-shoulda-wouldas, and unfortunately, it's harder to admit that it might have gone either way regardless.


Melecia Wright is a PhD student in Nutrition Epidemiology at the University at North Carolina, Chapel Hill. She graduated from Princeton in 2011 with a degree in Molecular Biology and a certificate in Environmental Studies. Melecia loves food and enjoys engaging in discussions pertaining to agriculture, nutrition biochemistry and the epidemiology of food-related lifestyle diseases.


comments powered by Disqus

Get the best

Get monthly email updates with the best from The Concepts Project. No spam, ever.

Contact us

Get in touch, we'd love to hear from you: theconceptsproject@gmail.com

Greatest Hits


Thinking At The Margin: what to do when you drop your piggy bank in the middle of the forest.

Strategy and Backward Induction: how to win a week of lunches from your unsuspecting colleagues.

What is Multiple Imputation?: when statisticians turn into detectives.

On Shuttle Drivers, Chocolate and NP Completeness: a deliciously difficult problem in computer science.

Rest and Digest vs Fight or Flight: how your body (and medications) help with fighting tigers.

Sites we like


William Shaw, writing about Politics, Theatre, Sci-fi… Mainly Sci-fi.

Better Explained, for maths explanations that click.

Science Non Fiction, a graduate student perspective on science in the news and in our lives.

Clearer Thinking, learn to think more clearly and make better decisions.

EconScribe.org, working to improve the quality of research communications.

Jess Whittlestone, a blog about decision making and behavioural science.