Evidence based medicine Flashcards
3 contexts in which diagnoses sought
- screening
- case finding
- people come in with prob
what is test for PE
D-dimer test - product of breakdown clot
interpretation of D-dimer
500 pos. for thrombosis
what is type 1 error
false +ve
- abnormal person
- has another disease
what is type 2 error
false -ve
- person’s disease give normal value
- tech problem with test
def. sensitivity
TP/TP+FN
def. spec.
TN/TN+FP
what do we want to rule out
highly sensitive test (few false neg) (SnOut)
what do we want to rule in
highly specific test (few false pos) (SpIn)
what is trade off
as move cutoff value, you trade spec. for sens.
what is ROC curve
refers to the tradeoff in sens and spec when moving values
what is test vs. gold standard
test gives you an indication of when you should do the gold standard
what steps in predictive values
- start with Pt that has a pretest probability
- administer the test
- given results, now how likely to have disease
where does pre-test probability come from (4)
- experience (gut)
- pop. prevalence
- published data on patients with similar issues
- clinical prediction rules
what are clinical prediction rules
decision making rules containing variables from Hx, physcial and/or simple diagnostic tests
what is pos. predictive value
TP/(TP+FP)
what is neg. predictive value
TN/(TN+FN)
what is prob. of disease if test is negative
100-NPV
how does pre-test probability influence the test
even with same sens. and spec. , a low pre-test prob will mean a lower-posttest probability
how to determine where to put the treatment threshold for post-test probability
based on out comes
when to make threshold low (2)
- disease serious if untreated
2. treatment is effective and inexpensive
when to make threshold high
treatment risky or expensive
def. likelihood ratio
prob. of that result in person who has disease/prob of that result in person who doesn’t have that disease
what is LR used for
to effectively calculate post-test prob. of disease at multiple levels of a test
formula for LR pos. test
sens/(1-spec)
formula for LR neg. test
(1-sens.)/spec
4 adv. of using LR
- easy to determine post-test prob. using a nomogram
- can use at multiple levels of test
- with sequential tests can multiply LRs
- at a glance can see how helpful findings might be
2 LR magnitudes that are helpful
- LR > 5-10 - helps for ruling disease in
2 LR
