W10_01 evidence based medicine Flashcards
define evidence based medicine
the conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients
what’s a diagnostic test?
have a presenting problem, make a hypothesis, using test to check hypothesis
recall the 4 approaches to diagnosis (Ps)
possibilistic
probabilistic
prognostic
pragmatic
what’s D-dimer?
a breakdown product of a crosslinked fibrin mesh - indicates that there was a clot
what’s the threshold for d-dimer interpretation?
_ 500mg/L means positive for thrombosis
what’s type I error?
“false positive”. Disease is absent, but test is positive.
what’s type II error?
“false negative”. Disease is present, but test is negative
what’s sensitivity?
proportion of those with the disease that test positive
what’s specificity?
proportion of those without the disease that test negative
what’s ruling out?
if the test is negative, and the test is very sensitive. Nearly everyone who has the disease tests positive
what’s ruling in?
if the test is positive, and the test is very specific. Nothing else gives this positive test result
how to maximize sensitivity?
low cutoff
how to maximize specificity?
high cutoff
what’s the receiver operating characteristic (ROC) curve?
specificity vs sensitivity
what’s the gold standard for trisomy 21?
amniocentesis w karyotyp
what’s the gold standard for ectopic pregnancy?
laparoscopy
what’s the gold standard for coronary artery disease?
coronary anteriogram
what’s the gold standard for appendicitis?
CT abdo
what’s the gold standard for alcoholism?
detailed interview
what’s the gold standard for pulmonary embolism?
pulmonary angiogram
what’s are predictive values?
given the test results, how likely are they to have the disease?
define PPV?
PPV = TP/(TP+FP)
define NPV?
NPV = TN/(TN+FN)
note: bayes’ theorem allows us to calculate the post-test probability, given the pre-test probability
ok
when are test/treatment thresholds low?
serious disease and/or easy/inexpensive test/treatment to do
what’s the likelihood ratio?
P(result|disease)/P(result|~disease)
note: pretest probability can be converted to post-test probability via the likelihood ratio
notice: numerator is sensitivity, denominator is 1-specificity
note: post-test odds = pretest odds X likelihood
ok
what likelihood ratios suggest to rule disease in?
LR > 5 to 10
what likelihood ratios suggest to rule disease out?
LR < 0.1 to 0.2
