1B Stats Flashcards
Gold standard is
most reliable and accurate test (compare other tests to this test)
degree to which instrument measures what it says it does
validity
degree to which measurement is free from error
reliability
ability to detect change over time in measured construct
responsiveness
contingency table for diagnostic test: a
a =true positive
b =
false positive
c =
false negative
d =
true negative
given that individual has condition, probability that test is +
sensitivity
given that individual does not have condition, probability that test will be negative
specificity
true + rate
sensitivity
true - rate
specificity
sensitivity equation
a/a+c
true positives/true pos + false neg (ALL COND POS)
specificity equation
d/(b+d)
true neg/ false pos+true neg (ALL COND NEG)
given that test result is positive, probability that individual has condition
+ predictive value
given that test result is negative, probability that individual DOES NOT have condition
- predictive value
+ predictive value equation
a/a+b
true positive/ALL TEST POS
-pred value equation
d/c+d
true neg/false neg+true neg (ALL TEST NEG)
condition with low prevalence will have
lower +PV
higher -PV
likelihood ratios combine
sensitivity and specificity values
positive LR
sensitivity/(1-specificity)
negative LR
1-sensitivity/specificity
higher positive LR
more certain that patient HAS condition if test is positive!
ratio of true + to false +
lower negative LR
odds that individual has condition is SMALLER
ratio of false neg to true neg
If LR is over 10
If -LR is smaller than 0.1
large, conclusive shifts in probability!
moderate shifts in probability
LR: 5-10
-LR: 0.1-0.2
small likelihood ratio
+LR: 1-2 (close to 1)
-LR: 0.5-1 (close to 1)
increased ______(reliability/validity) of a test decreases the ____value in that pop
RELIABILITY
MDC
What is the most powerful tool for quantifying importance of a test?
Likelihood ratio!
importance of positive and negative test results can be examined indiependently!
Reliability measures of change
MDC and MCID
MDC: amount of change needed to exceed measurement error of test
MCID: amount of change that is significant to patient’s improvement