Biostats

Question 1

Sensitivity

Answer 1

Sensitivity=TP/(TP+FN)
Looks at ALL the people who DO have the disease.
“True is on Top”

• A high sensitivity test is one in which the FN account for a small percentage of patients who DO have the disease.
• Rules OUT disease well. (SnOut)
• Good for a screening test.

-To determine if a test effectively rules OUT disease, we must look at the test results of the people who DO have the disease (sounds kind of backwards, but if the test is good at identifying people who DO have the disease, then we can trust that a negative result really means that the person does NOT have the disease).

Question 2

Specificity

Answer 2

Specificity=TN/(TN+FP)
Looks at All the people who DO NOT have the disease.
“True is on Top”

-A high specificity test is one in which the FP account for a small percentage of patients without disease.
(Tip: 1-Specificity=Probability of a False Positive)
-Rules IN disease well. (SpIn)
-Good for a confirmatory test.

-To determine if a test effectively rules IN disease, we must look at the test results of the people who Do NOT have the disease (sounds kind of backwards).

Question 3

Positive predictive value

Answer 3

PPV=TP/(TP+FP)
-Looks at All the positive results. “True is on Top”

• How likely is it that a positive result actually means the patient has the disease?
• Proportional to prevalence.
• The more prevalent a disease, the more likely a positive test result actually means disease is present.

Question 4

Negative predictive value

Answer 4

NPV=TN/(TN+FN)

• Looks at all negative results. “True is on Top”
• How likely is it that a negative test result actually means a patient does not have the disease.
• Inversely proportional to prevalence (high prevalence decreases the NPV of the test).

Question 5

Relative Risk

Answer 5

aka Risk Ratio
RR=[Exposed with disease/Total exposed] / [Unexposed with disease/Total unexposed]
-Ratio of risk of disease in exposed vs unexposed individuals
-Used for cohort studies (in which exposure status is known at study onset and risk of developing the bad outcome is followed over time)

Question 6

Odds Ratio

Answer 6

OR=[Exposed cases/Unexposed cases] / [Exposed controls/Unexposed controls]

• Ratio of odds of exposure in cases vs controls
• Used for case-control studies (in which disease status is known at study onset and odds of exposure is looked at retrospectively)
• If disease is very rare, then OR~=RR