“That's Not What I Meant”
How repeatable a test is
Reliability
Assume a population of 1,000 people, 200 of whom have a disease and 800 do not have the disease. If only 750 people test negative, what is the specificity of the screening?
705/800 x 100 = 94% (93.75%)
A new illness called “goosepox” is popping up around Chico. You find that you show some symptoms of the disease and decide to go to the doctor. The doctor decides to run two tests at the same time to confirm you have this illness. What type of testing did this doctor just run on you?
Simultaneous testing
Specificity & Sensitivity
Inverse
When calculating sensitivity for two simultaneous tests, we calculate for ___ sensitivity
Net
Ability of a test to distinguish who does and does not have a disease
Validity
Assume a population of 1,000 people, 200 of whom have a disease and 800 do not have the disease. If only 145 people test positive, what is the sensitivity of the screening?
145/200 x 100 = 73%
(72.5%)
After reviewing the results of your test, you and another pathologist have decided to calculate your overall percent agreement. Your results show that you have a 70% agreement. What does this tell you about your agreement?
You had good agreement generally (intermediate to good also works as a answer)
Are validity and reliability related independently or dependently?
Independently
Type of curve which has two distinct peaks instead of one.
Bimodal curve/distribution
Ability of a test to identify correctly those who do have a disease
Sensitivity
Assume a population of 10,000 people, 500 of whom have a disease. If a test is conducted that has a sensitivity of 85%, how many people will the test give true positives?
500 x 0.85 = 425 people
You’re a pathologist researching a new disease that is spreading in Chico called “goosepox”, you’ve started on a new test to reveal if someone is positive or negative for the disease. What type of results should you have?
Dichotomous
PPV and prevalence
As prevalence increases, PPV increases
What statistic corrects percent agreement between two observers by eliminating chance?
Kappa statistic
Ability of a test to identify correctly those who do not have a disease
Specificity
Assume a population of 10,000 people, 9% of whom have a disease. If a test is conducted that has a specificity of 70%, how many people will the test give true negatives?
9,100 x 0.70 = 6,370 people
A physician that has just conducted a new test on you for the recent disease known as “goosepox”. They are currently wondering “if the test results are positive in this patient… What is the probability that this patient has the disease?” What is the doctor worried about?
Positive predictive value
What is a specific example of a continuous variable that we have given cut off points to define health status in relation to diabetes?
Blood glucose
What is sequential (two-stage) testing (and its effects on sensitivity and specificity)?
Cheap test followed by more expensive test in pts who tested positive.
Net loss of sensitivity, gain of specificity
Proportion of patients which test positive for a disease who actually does/does not have the disease.
True positives / (true + false positives)
True negatives / (true + false negatives)
PPV or NPV
A dichotomous screening test is conducted in a population of 800, of these subjects 120 came back with a positive result. Of these 120 subjects only 80 have the disease. What is the positive predictive value of this test?
80/120 x 100 = 67%
(66.66%)
You and your team of researchers have finished creating a graph distribution curve showing the results of your test. What does your curve tell us about your reliability and validity?
Test results are reliable, but not valid
Intraobserver variation & Interobserver variation: what is the main difference between these two methods of observation.
Intra - variations made by the same observer
Inter - variations made by 2 different people
3 factors contributing to test result variations
Intrasubject, Intraobserver, Interobserver