Section Links

Causation and Association

Associations are first identified, with causation being shown second. Associations are observed, while causation is inferred. Proving causality can be done with Hill's criteria.

Associations

Associations, or relationships, are statistical dependence between two or more events, characteristics, or other variables. However, it does not imply causation.

After designing a study to determine whether an association exists, work needs to be done to test what sort of relationship exists.

There are three different types of associations:

Spurious: a false association due to chance or bias
Non-causal: two factors of interest are both caused by another (potentially a confounder)
Causal: change in the exposure leads to change in outcome

Measuring Associations

Associations can be measured in two ways.

Relative measures are presented as:

ratios of exposed / unexposed
relative risks and odds ratios

Absolute Measures are presented as:

differences of exposed - unexposed
attributable risk and number needed to treat (NNT)

return to top

Causation

Causes are complex and almost always act in concert with other factors that have various strengths and directions.

Indirect causes can include risk factors, but not all risk factors are causal.

Four types of causal factors

Necessary and sufficient

without a given factor, disease does not develop
ie HIV and AIDS

Necessary but not sufficient

multiple factors, including the one of interest, required
ie HPV infection and cervical cancer

Sufficient but not necessary

factor can produce disease, but so can other independent factors
ie radiation and benzene can cause leukemia, but neither requires the other

Neither sufficient nor necessary

complex models of disease etiology
ie high fat diet, hypertension, diabetes, etc all are related to heart disease

return to top

Proving Causality - Hill's Criteria for Causation

temporal relationship: the only absolutely neccessary criterion
strength of the association
dose-response relationship
consistency
experiemental
biolgicial plausibility
coherence
specificity

return to top