Causation is impossible to prove in empirical studies. Yet Scientists emphasize causality in scientific studies. Only through necessary proofs can things be known to exist in a cause and effect relationship. This claim to knowledge is also evident in epidemiology where necessary causes are looked for. This bias has resulted in difficultly in finding good reasons to say that a factor and a disease are related. This confusion is shown by the use of the black box analogy to provide support for causation. The idea of the black box is most often used when physical determinants for disease cannot be necessarily ascertained. We are used to thinking of physical bodies as acting according to the laws of physics and therefore being determinant in nature. But when we look at other factors it becomes clear that the casual factors of disease can not be known to exist necessarily and therefore have a necessary effect.
When looking at disease factors one must look at not only the physical factors but also in some cases the chemical, biological, ecological, environmental, and sociological factors. When talking about factors that lead to disease multiple factors may be involved and more often than not these factors are not all physical.
But according to science in order for science to be useful it must establish a cause effect relationship. Hume define a necessary and sufficient condition as being that in which if the cause does not exist then the effect will not exist either (Olsen, 2003). This criterion is rarely if ever demonstrated.
The philosopher Hume pointed out that what we know is based on experience and therefore we cannot know the necessary antecedents that result in consequential determinants or even if this determinant nature exists (Karhausen, 2000).
What is called counterfactuals increases this confusion. While we can know that in a hypothetical argument if the antecedent is true and the consequent is false that the claim is false. Yet material claims are different from counterfactual claims. If I am in Mexico then I am in North America. This is true when I am indeed in fact in Mexico. But this conditional claim is also true if I happen to be in Africa and this fact has no bearing on the truth value of the claim that if I am in Mexico I am also in North America. Yet also the material claim is still true. If the claim that one is in Mexico is false (e.g., me being in Africa) and the consequent that therefore I am in North America is still a true claim! This is the nature of hypothetical statements which are borne out in beginning logic truth tables. So we cannot make a claim that if there is or is not a causal factor that the disease will not invariably occur.
Proving some factor is causal has been the “holy grail” in science. We can’t know by observation whether a factor has a causal effect on something else. All we know is through observation which in itself is limited by the possible number of observations a certain probability that something might occur. Because this falls short of the necessary and sufficient conditions desired in hard science an attempt is made to adopt what may seems to be somewhat arbitrary criteria to firm up this supposed causal relationship. These criteria are production, necessary causes, sufficient– component causes, probabilistic causes, and counterfactuals.
Production is simply referred to a cause that “produces” an effect contrary to something that does not and is therefore not causal, but the nature of production and even how it is different from causation is unclear and is therefore an elusive concept.
Necessary causes are incorporated in the concept of sufficient causes. One reason the necessary and sufficient criterion is seen as flawed is because most diseases are not the result of a specific (e.g., sufficient) cause. This may vary not only because of the irregularity of causal factors, but also temporal consideration as well as non-biological factors.
Sufficient-Component causes attempts to address the fact that most disease does not have a singularly specific “genetic” factor. While this may provide framework for a variety of necessary causes the problem still exists asking which factor is most prominent, (e.g., including dose response factors), and do the individual factors rely on other independent factors? “A model of causation that describes causes in terms of sufficient causes and their component causes illuminates important principles such as multicausality, the dependence of the strength of component causes on the prevalence of complementary component causes, and interaction between component causes” (Rothman, 2001). There is no doubt that there are a lot of variables to keep in mind and because of this the value of this idea may be drawn into question.
Probability factors may seem to fit in best with a statistical model but while necessary and sufficient conditions may be referred to in a probabilistic fashion, probabilistic measures cannot be talked about in a necessary manner. A probabilistic way of talking about physical phenomena is especially problematic for scientific disciplines. Like relativity theory and quantum mechanics, nary the twain shall meet. One main shortcoming is that if the criterion is probability then one cannot truly predict if a causal factor will result in an outcome thus contradicting the spirit of medical research.
Finally there are counterfactuals. Counterfactuals make a distinction between causation and mere correlation which is essential in any study but does not by itself result in a definition of causation. While the counterfactual is not inconsistent with necessary causes it does not provide a rationale for arriving at a necessary cause (Parascandola, 2001). Whether something is true (ontology) or whether something can be known to be true (epistemology) are both dependent on the counterfactual argument in the particular conditional statement to arrive at useful information and this distinction is often a source of confusion. The ceteris paribus condition is essential. While one may not know if “the contaminated Broad Street well is a cause of the cholera epidemic; when it is the case that if the Broad Street pump was shut down, the cholera incidence would decrease”. Yet problems can result with this too. For if malaria is thought to be the result of swamp gas and we drain the swamp and therefore eliminate malaria, it turns out the conclusion was false. In order to use the counterfactual validly one must be sure that the consequent is true when the antecedent is true otherwise no useful knowledge may be obtained.
Establishing causation for chronic diseases is difficult. Many chronic diseases have many factors and the relationship between these factors cannot always be known. A sufficient-component approach may be useful in much chronic disease especially if the risk factors can be easily identified. Nevertheless a probabilistic approach may be more efficacious because the weight of the individual risk factors may not be known.
More effective for infectious disease may be looking at necessary and sufficient conditions (i.e., the infective agent producing tuberculosis). Production as well could be very useful if it can be shown that for example when someone has AIDS they invariably have the HIV virus although even this does not show proof of direct causation. As also has been shown, the counterfactual can be used most efficiently shown from the example of shutting off the Broad Street pump by denying the true consequent and therefore righting the Broad Street wrong.
References:
Karhausen, L.R. (2000). The Elusive Grail of Epidemiology. Medicine, Health Care and Philosophy, 3:59-67. Retrieved January 3, 2008, from http://proquest.umi.com/pqdweb
Olsen, J. (2003). What Characterizes a Useful Concept of Causation in Epidemiology? Journal of Epidemiology and Community Health, 57(2). Retrieved January 3, 2008, from http://proquest.umi.com/pqdweb
Parascandola, M., Weed D.L. (2001). Causation in Epidemiology. Journal of Epidemiology and Community Health, 55(12). Retrieved January 3, 2008, from http://proquest.umi.com/pqdweb
Rothman, K.J., Greenland, S. (2001). Causation and Causal Inference in Epidemiology. American Journal of Public Health, 95(S1). Retrieved January 3, 2008, from http://proquest.umi.com/pqdweb