Hi ,
When Hungarian doctor Ignaz Semmelweis started working at Vienna General Hospital's maternity clinic in 1846, it's unlikely he expected to find himself face-to-face with the perfect set-up for a natural experiment.
Yet that's exactly what he found.
The clinic was divided into two wards - one staffed by doctors and medical students, and the other by midwives - with the maternal mortality rate almost five times higher in the doctors' ward.
By comparing the conditions between the two wards and testing a series of hypotheses, Semmelweis was ultimately able to identify the true cause of the mortality difference - the doctors were performing autopsies before delivering babies while the midwives weren't.
And this led to the discovery of the importance of hand washing in controlling the spread of
disease.
Here's the thing...
A/B tests are the gold standard for causal analytics. But they're not always practical, or even ethical, to implement.
Had Semmelweis randomly divided the patients in the hospital into two groups - those who would have their babies delivered by doctors or midwives who did observe hand hygiene and those whose babies would be delivered by doctors or midwives who did not. That would have been very wrong.
However, by operationalising the
randomness that existed in his environment, and using that to devise a series of natural experiments, Semmelweis was able to identify a causal relationship that we all still benefit from understanding today.
Just because conducting an A/B test might not be possible doesn't mean causal analytics isn't something you can use.
Natural experiments are all around us. You just need to know where to look.
Talk again soon,
Dr Genevieve Hayes.
p.s. I recently had the opportunity to discuss causal analytics and natural experiments with Joanne Rodrigues, author of Product Analytics. You can listen to our entire conversation here: Episode 47: Leveraging Causal Inference to Drive Business Value in Data
Science
