B5.073 Colon Cancer Screening/B1.019 Screening and Test Performance Flashcards
how do you maximize sensitivity/specificity?
plot 1-spec vs. sense
area under curve as close to 1 as possible
what is the formula for NNT?
calculate absolute risk reduction
1/ARR = NNT
aims of cancer screening
reduce cancer related mortality and morbidity
detect precursors or early disease
risks of cancer screening
- test itself: exposures, complications
- false positives
- overdiagnosis: unnecessary treatment
- cost
benefits of cancer screening
reduced morbidity reduced mortality psychological reassurance health maintenance quality of life
what is lead time bias
identification of disease earlier in natural course, but screening has no impact on disease course
die at same age despite earlier detection
what is length bias
long preclinical phase is earlier to detect
screened individuals are more likely to have milder or slower developing diseases
what is overdiagnosis
diagnosed cancers that would not go on to cause symptoms/death given their natural course
essential features of a disease been screened for
disease common or of unusual severity
known natural history and lag phase between detection and irreversible outcomes (detectable pre-clinical phase)
disease is treatable
essential features of a screening test
acceptable to public risk/benefit profile favorable economically and technically feasible accurate and repeatable acceptable levels of sens and spec
what is sensitivity
ability of a test to correctly identify those who have disease
- worry about false positives
- few false negatives
SnNOUT
when sensitivity is high, a negative result can rule out disease
most negatives are true negatives
what is specificity
ability of a test to correctly identify those who do not have a disease
- worry about false negatives
- few false positives
SpIN
when specificity is high, a positive result can rule in disease
most positives true positives
how to calculate sensitivity
probability of a test being positive given that disease is present
SN = TP/ (TP+FN)
first column in 2x2 table
how to calculate specificity
probability of a test being negative given the disease is not present
SP = TN/ (TN+FP)
second column in 2x2 table
perfect sensitivity
detects all cases
no false negatives
perfect specificity
rules out all non-cases
no false positives
what is positive predictive value
probability that disease is present given test is positive
PPV = TP/(TP+FP)
first row on 2x2 table
what is negative predictive value
probability that disease is not present given that the test is negative
NPV = TN/(TN+FN)
second row on 2x2 table
sens/spec vs. ppv/npv
sens/spec: how likely is a test to rule in or rule out disease, given whether or not disease is present
ppv/npv: how likely is a person to have or not have disease given the test result
how is prevalence related to predictive value
screening more efficient id directed at patients with high prevalence
PPV and NPV depend on prevalence of disease in population unlike sens and spec
accuracy
ability of a test to identify who have disease and who does not
-sense and spec
precision
results of a test are consistent each time the rest is conducted
prevalence
pretest probability
probability of disease before test result is known
predictive value
posttest probability
probability of disease after test result known
likelihood ratios
LR = likelihood of a test result when disease is present/ likelihood of a test result when disease is absent
