Establishing Causality

In the language developed earlier in the section, you can think of the people inthe S&V houses as the treatment group, and those in the Lambeth houses at thecontrol group. A crucial element in Snow’s analysis was that the people in thetwo groups were comparable to each other, apart from the treatment.

In order to establish whether it was the water supply that was causing cholera,Snow had to compare two groups that were similar to each other in all but oneaspect–their water supply. Only then would he be able to ascribe the differencesin their outcomes to the water supply. If the two groups had been different insome other way as well, it would have been difficult to point the finger at thewater supply as the source of the disease. For example, if the treatment groupconsisted of factory workers and the control group did not, then differencesbetween the outcomes in the two groups could have been due to the water supply,or to factory work, or both, or to any other characteristic that made the groupsdifferent from each other. The final picture would have been much more fuzzy.

Snow’s brilliance lay in identifying two groups that would make his comparisonclear. He had set out to establish a causal relation between contaminated waterand cholera infection, and to a great extent he succeeded, even though themiasmatists ignored and even ridiculed him. Of course, Snow did not understandthe detailed mechanism by which humans contract cholera. That discovery was madein 1883, when the German scientist Robert Koch isolated the Vibrio cholerae,the bacterium that enters the human small intestine and causes cholera.

In fact the Vibrio cholerae had been identified in 1854 by Filippo Pacini inItaly, just about when Snow was analyzing his data in London. Because of thedominance of the miasmatists in Italy, Pacini’s discovery languished unknown.But by the end of the 1800’s, the miasma brigade was in retreat. Subsequenthistory has vindicated Pacini and John Snow. Snow’s methods led to thedevelopment of the field of epidemiology, which is the study of the spread ofdiseases.

Confounding

Let us now return to more modern times, armed with an important lesson that wehave learned along the way:

In an observational study, if the treatment and control groups differ in waysother than the treatment, it is difficult to make conclusions about causality.

An underlying difference between the two groups (other than the treatment) iscalled a confounding factor, because it might confound you (that is, mess youup) when you try to reach a conclusion.

Example: Coffee and lung cancer. Studies in the 1960’s showed that coffeedrinkers had higher rates of lung cancer than those who did not drink coffee.Because of this, some people identified coffee as a cause of lung cancer. Butcoffee does not cause lung cancer. The analysis contained a confounding factor –smoking. In those days, coffee drinkers were also likely to have been smokers,and smoking does cause lung cancer. Coffee drinking was associated with lungcancer, but it did not cause the disease.

Confounding factors are common in observational studies. Good studies take greatcare to reduce confounding.

This page was created by The Jupyter Book Community