Inferring Causation Flashcards
What are the strengths and weaknesses of a “review” article?
Strengths - good discussion from experts in the field
Weak - subject to bias of author or journal
Why are systematic reviews better than “reviews”?
They follow a prescribed protocol
Where do you find systematic reviews? (3)
Peer-reviewed journals, databases, Cochrane Library
Our research question is defined using PICO. What does each letter mean?
P_ - people of interest (patient); is the group or person similar to the patient?
I - Intervention; treatment, exposure, other agent
C - compare different treatments
O - outcome, is the outcome relevant to our patient? Is this what the patient wants to happen
What are the two main goals someone has when writing a systematic review, to put together and present to the reader?
Increase sensitivity (detect an article when it exists) and reduce author bias.
Publication bias
Bias towards publishing only significant outcomes. What if not significant? Want to include those too.
Which experiments tend to have better effect sizes? (More significant results) weaker or stronger?
Weaker
How are results of a systematic review summarized?
Forest plot form.
Difference btwn meta-analysis and systematic review
Studies are combined in meta-analysis and analyzed. SR - just a compilation and presentation.
Do causal processes involve intermediate steps?
Almost always
What do we use to counter the persistent single-cause model?
Web of causation.
For impact on web of causation, how many strands do you need to cut for impact?
Only at least 1
Causes are necessary and/or sufficient. What does necessary mean?
Necessary - without the cause, the effect would not exist. Need more factors for effect to happen.
What does sufficient mean?
If the cause is there, the effect will result, whether other causes are there or not.
Name two examples of statistical numbers we use to assess association, or risk of outcome given exposure?
Risk ratio, odds ratio
B-H criteria - what is the only required criteria?
Temporality,
Strong associations from a well designed study are more or less likely to be caused by chance or bias?
Less
To infer causality, what is required of the dose-response curve?
Strong association (change one, change the other), doesn’t need to be linear
What is the B-H criteria of reversibility?
Take the cause away, effect goes away.
What is the B-H criteria of consistency?
The association (or cause and effect) are seen across multiple groups - they haev high external validity.
What is the B-H hill criteria of biological plausibility to determine cause & effect?
The association should be possible and explainable.
What is the B-H criteria of specificity for inferring causation?
One exposure, one outcome.
What is the analogy criteria for the B-H criteria for inferring causality?
The cause and effect relationship are similar to those that already exist.
Is the B-H criteria a checklist? How does one establish causality
No - the more criteria, the stronger the case
Ecologic studies
Study a sample population for associations in question
Ecologic fallacy
Applying the results of a group to an individual
What is the best use of an ecologic study?
Generating a hypothesis from an idea and moving forward from there.
