Causation Flashcards
Which factors should be considered before establishing a cause and effect relationship?
chance, bias and confounding variables
How is chance measured?
Using P values and p<0.05 means less than 5% probability results are due to chance
What are confidence intervals?
Interval where the true value lies (usually 95%)
What does it mean if the confidence interval is 95%?
If a study was repeated many times, 95% of them would lie in that interval
What is bias?
A systematic error resulting in a wrong value being obtained
What can lead to bias?
The study design or the execution of the study
What cannot reduce bias?
Sample size or analysis of results
What are the two main types of bias?
- selection - no response, healthy entrant effect, loss of follow up
measurement - recall bias
What is a confounding variable?
Affects the dependent and independent variable e.g. factors that have a causal relationship with disease
What are examples of confounding variables?
age, sex,geography and socio-economic status
What are the Bradford Hill Criteria?
- Strength of association
- Consistency
- Specificity
- Temporal relationship
- Dose response relationship
- Plausibility
- Experimental evidence
- Coherence
- Analogy
What two areas must be addressed to ensure causality?
That the association between exposure and outcome is valid, and the evidence from many sources support causality
What is strength of association and how is it measured?
By size of relative risk. Strong increases likelihood of causation but weak doesn’t mean exclusion
How is consistency related to causation (BHC)?
More likely to be causal if similar results obtained as unlikely same errors would have occurred. But no consistency doesn’t mean no causation.
How is specificity related to causation (BHC)?
If one factors increases the risk of one particular disease then it is good evidence but doesn’t mean that without it causation isn’t there.
