Association & Causation Flashcards
Association
statistical dependence between 2 variables
Degree to which the rate of disease in persons with a specific exposure is either higher or lower than the rate of disease without that exposure.
What to consider when evaluating statistical association
Chance
Bias
Confounding
Cause
Why assess chance?
Inference may be made from samples rather than whole populations
Use stat. tests, find p values and confidence intervals
What is bias?
A systemic error leading to an incorrect estimate of effect of an exposure on development of a disease or outcome of interest
Eliminating bias?
bias = consequence of defects in design or execution of a study.
Can’t be controlled in analysis of study.
Can’t be eliminated by increasing sample size.
What are the 2 broad types of bias?
Selection bias
Measurement bias
What is selection bias?
A systematic difference between characteristics of people selected & those who weren’t.
What is measurement bias?
When measurements/ classifications of disease/ exposure are inaccurate
Give an example of selection bias
Healthy entrant effect
Give an example of measurement bias
Recall bias
Confounding
any factor believed to have a real effect on risk of disease under investigation & is also related to the risk factor under investigation
Confounding includes factors
that have a direct causal link with the disease (e.g. smoking & lung cancer)
that are good proxy measures of more direct unknown causes (e.g. age & social class).
Common confounders
Age
Sex
Socio-economic class
Geography
Why is socio-economic class a confounder?
–Poorer people have higher rate of almost all disease.
–Higher risk of early death in poor people.
Cause?
Judgement of a cause-effect relationship
based on 2 main areas:
1. Observed association between an exposure & a disease is valid
2. Totality of evidence taken from a number of sources supports a judgement of causality
Bradford Hill Criteria for establishing causation
Temporal relationship Plausibility Consistency with other investigations Strength of association Dose-response relationship Specificity Experimental evidence Coherence Analogy
What should you also consider when establishing causation
Reversibility
Strength in BHC
Strong association more likely causal than weak association
Consistency in BHC
Similar results found in different studies
Unlikely they were all subject to the same type of errors.
Specificity in BHC
A particular exposure increasing the risk of a certain disease but not the risk of other diseases
Temporal relationship in BHC
For a putative risk factor to be the cause of a disease it has to precede the disease.
Which criterion of BHC is essential
Temporal relationship
Dose-response relationship in BHC
Further evidence of a causal relationship if increased levels of exposure lead to increased risk of disease.
Some causal associations show a single jump (threshold) rather than a monotonic trend.
Plausibility in BHC
Association is more likely to be causal if consistent with other knowledge e.g. animal experiments
Experimental evidence
Experimental evidence on humans (rare) or animals.
Coherence in BHC
Implies a cause & effect interpretation that doesn’t conflict with what is known of the natural history
Analogy in BHC
provides a source of more elaborate hypotheses about association in question