EBM Flashcards
What do LR+ and LR negative stand for
ALWAYS, diseased/nondiseased.
LR+ = likelihood of pos test in diseased/pos test in non diseased
aka true positive rate/true negative rate
= sensitivity/ 1-specificity
***LR negative
= likelihood of neg test in diseased/neg test in nondiseased
= 1-sensitivity/specificity
Attending says he wants to do a test to be close to 100% certain a patient does not have a specific disease. What does this imply>
That this test has a high sensitivity to have a high negative predictive value. This is only the case if the condition is really serious or the treatment is very effective.
What kind of studies require a placebo arm?
If the outcome is graded and subjective such as pain.
For what kind of question is RCT best, prospective cohort study, case control
in terms of whether to do a diagnostic test
A randomized controlled trial (the bigger the better) is the best design if your question is what course of action will lead to the best results for your patient – e.g. Will doing test A lead to better outcomes than doing an alternative test? For determining test characteristics, a prospective cohort study in which all tested persons (“consecutive”) get both the test under consideration and a gold-standard test is optimal. A retrospective cohort study examining patients whose doctors happened to order both tests is likely to be biased – e.g. Do sicker patients tend to get both tests rather than one? A case-control study is optimal only when no other study is practicable, for example when looking at rare outcomes of medications or procedures.
What is ascertainment bias?
This bias occurs when the results of your study are skewed due to factors you didn’t account for like the researcher’s knowledge of which patients are getting which treatments in clinical trials or poor data collection methods which lead to non-representative samples.
What is verification bias?
Verification bias is when the index test (the diagnostic test in question) determines whether a gold standard reference test is performed.
In reality all patients should get both the experimental and gold standard test
What are the values for area under the ROC in terms of perfect - worthless test
.9-1 super good .8-.9 good .7-.8 fair .6-.7 poor D .5-.6 fail (you are just as likely to get a true positive as you are to get a false positive)
How is beta used?
It is a concern in type 2 error. A lower beta means more power because power = 1- beta.
What is relative risk?
Relative risk is how likely an event will occur in the treatment group relative to the CONTROL group
An RR > 1 means treatment increased risk of outcome
An RR < 1 means treatment decreases risk of outcome.
1- relative risk is RRR.
RRR also = ARR/Risk