More on Causality and Introduction to Statistics and Epidemiology (part one)
We continue on our discussion on cause and effect relationships. Today, we are going to discuss another aspect of cause and effect association and upload another article on how to deduce cause and effect relationships in health care research. This article was written by the British epidemiologist [Sir Austin Bradford Hill][h].
Notes on the causality article by Sir Austin Bradford Hill
Hill (1964) identified nine issues that he wrote to be considered for any cause and effect relationship discussion. These are as follows
Strength of Association — By this, Hill meant the extent to which an exposure (or an intervention) is associated with a health outcome must be “high” enough for it to support a causal association. How high is high enough and how low is too low? What do you think?
Consistency: Hill asks, “has it been repeatedly observed by different persons,in different places, circumstances and times?” When you read this section, you encounter terms like: Age-specific death rates (essentially indicates the rate of deaths for persons in each age group)
Specificity: this refers to the observation that if an agent or exposure A is supposed to be the cause of an outcome B, then A should be associated with B only and not another condition C. If this is the case, then, according to Hill, this is a strong indicator that A is a “cause” of B. But he also states, the reverse may not be true. This is where the web of causation becomes a strong counter point.
Temporality: this refers to the fact that a cause must always precede the disease. Thus, if exposure E is considered to be a cause of disease D, then E must precede emergence of D in time scale.
Biological gradient: This is the issue around a dose response curve, meaning if the extent of exposure E increases, there will be a corresponding increment in the emergence or rate or severity of disease D as well. “the dustier the environment the greater the incidence of disease we would expect to see (if dust is supposed to be a cause of the disease)”
Plausibility: Can the association be explained based on the existing knowldege of biology (or any other relevant science)?
Coherence: This is somewhat related to the biological plausibility in that if A is to be considered as a cause of B (health outcome), then, the data should be consistent with the existing knowledge of the day with respect to natural history and biology
Experiment: This brings us back to the roots of epidemiology where applicability of epidemiological knowledge is an issue. Hill asks for example, if an exposure E is to be considered as a cause of disease D, then, if one removes E from the environment, does corresponding the rate of D subside? If the answer is yes, then consider E as a cause of D. Else think again.
Analogy: Here, for an exposure to be a “cause” of a disease outcome D, the case becomes strong if one can also demonstrate other comparable situations where an exposure similar to E (say F) produces diseases that are similar to D.
These are the nine “viewpoints’,if you will, that Hill has described in this article.
Please read this article and prepare to discuss the importance of this article and try to support or refute his points of argument. Think of setting up or asking your own questions and see how far you can use Hill’s viewpoints to support your own examples. I am interested to hear from you. I shall discuss my own takes on the issue later.
Bradford, A. (1964). The Environment and Disease: Association or Causation? Proceedings of the Royal Society of Medicine (pp. 295-300). London.