Clinical Epidemiology

Question 1

What is risk difference?

Answer 1

Risk in population 1 - Risk in population 2. If smokers have a 50% chance of getting lung cancer and nonsmokers have a 4% chance of getting lung cancer, risk difference is 46%

Question 2

What is relative risk?

Answer 2

Also called risk ratio. Relative risk is a comparison of risk between two populations. For smokers (50% chance lung cancer) and nonsmokers (4%) you’d have 50/4 = 12.5. So the risk of getting lung cancer is 12.5xbaseline in smokers

Question 3

What is absolute risk reduction?

Answer 3

Absolute risk reduction is amount of risk removed by a treatment, or added by exposure. Same as risk difference.

Question 4

What is absolute risk?

Answer 4

Total risk of getting a condition.

Question 5

What is relative risk reduction?

Answer 5

Relative risk reduction is the percentage of baseline risk that’s removed by a treatment. (risk of control - risk of treatment) / risk of control

Question 6

What is a null hypothesis?

Answer 6

Null hypothesis is the argument that there will be no difference between two groups in a study.

Question 7

What is type 1 error?

Answer 7

Type 1 error = alpha = risk of false positives. The % chance that your data, or data more extreme than yours, actually comes from the null population even when you said it doesn’t.

Question 8

What is type 2 error?

Answer 8

Type 2 error = beta = risk of false negatives. The % chance that your data, or data more extreme than yours was actually from a different population when you said it was the same population.

Question 9

What is standard error?

Answer 9

A sort of sample-wide version of standard deviation. It tells us the degree of uncertainty associated with our calculated means. SE = SD/sqrt(n)

Question 10

What is standard deviation?

Answer 10

A measure of variability in individual measurements within a study. SD = sqrt(variance)

Question 11

What is the hazard ratio?

Answer 11

Ratio of hazard among people exposed to pathogen vs. hazard among control population. Similar to risk ratio

Question 12

What is the number needed to treat?

Answer 12

1/ARR. The number of people you would need to treat to prevent one condition from occurring (this doesn’t mean everyone else gets it, it just means you have to treat this many people to keep one person who could have randomly gotten the disease from getting the disease)

Question 13

What is a t-test’s 95% confidence interval?

Answer 13

mean + or - 2*standard deviation/sqrt(n). It tells you the interval in which you have a 95% of containing the actual difference between the two groups in a study.

Question 14

What is a major limitation of the p-value?

Answer 14

P values are impacted by study-wide variance, number of people in study, and size of an effect. You could have a really small effect (clinically insignificant) but still have significant results if you had a huge sample size or not much variability.

Question 15

What is statistical power?

Answer 15

1-beta. The likelihood that your insignificant findings were actually insignificant.

Question 16

What is the alternate hypothesis?

Answer 16

The argument that two samples come from different populations.

Question 17

What is specificity?

Answer 17

% of people without disease who are recognized by the test as not having the disease. (ability to find negatives)

Question 18

What is sensitivity?

Answer 18

% of people with disease who are recognized by the test as having the disease. (ability to find positives)

Question 19

Draw a 2X2 table. What pieces of information can you get from a 2X2 table?

Answer 19

upper left quadrant is true positives. upper right quadrant is false positives. lower left quadrant is false negatives. lower right quadrant is true negatives. Sensitivity = tp/fn + tp
Specificity = tn/fp + tn
PPV = tp / tp + fp
NPV = tn / tn + fn

Question 20

What is negative predictive value?

Answer 20

The probability that you actually do not have the disease if you’re tested negative.

Question 21

What is positive predictive value?

Answer 21

The probability that you have the disease if you were tested positive.

Question 22

What is pre-test probability?

Answer 22

The risk that someone in your “population” would have the disease. Pre-test probability can change from the general population to something really specific like someone with the BRCA1 gene. But pre-test probability then would be the likelihood that someone with the BRCA1 gene would have breast cancer.

Question 23

What is post-test probability?

Answer 23

Post test probability = Positive predictive value if the test was positive and 1-negative predictive value if the test was negative.

Question 24

What variables in 2x2 tables are affected by pre-test probability?

Answer 24

NPV and PPV. Actually everything except for sensitivity and specifictiy.

Question 25

What is a receiver operator characteristic curve?

Answer 25

It shows you have specificity and sensitivity of a test change as you adjust the threshold for a positive or negative test for a given drug. You want a test to have a curve that is convex (curves outward)

Question 26

Which part of a 2x2 table tells you false positives? False negatives? True positives? True negatives?

Answer 26

False positives is upper right quadrant (assuming disease + - is on top and test + - is on side). False negatives is lower left quadrant. True positives are upper left, and true negatives are lower right.

Question 27

What is incidence?

Answer 27

Incidence is the number of new cases of a disease in a given population within a time period over number of people in population.

Question 28

What is prevalence?

Answer 28

Prevalence is total number of people with a disease.

Question 29

What is incidence density?

Answer 29

number of cases that develop over total population time at risk. People who get sick do not count any longer towards the amount of time in the denominator. So if 5 people out of 100 get a disease immediately at the beginning of a year-long study, your incidence density would be 5/95 people years.

Question 30

What is cumulative incidence?

Answer 30

of people who get a disease/# of people followed over a given time period

Question 31

What is case fatality rate?

Answer 31

Number of people who get die/number of people who get sick.

Question 32

In what situation could prevalence of a disease increase while incidence decreases?

Answer 32

If you have a vaccine or prophylactic treatment that reduces number of people getting disease but you also have treatments that increase how long people can survive.

Question 33

How do you estimate pre-test probability?

Answer 33

You can use “gut feeling” (bad) or literature values or medical reports that provide guidelines for approximating risk.

Question 34

What is test-treat thresholding?

Answer 34

“I will test if the person has a 20% chance of having the disease, and I will treat if the person has a 90% chance of having the disease”

Question 35

What factors influence test-treat thresholding?

Answer 35

Severity of disease, severity of treatment.

Question 36

What would your test-treat threshold look like for a serious disease with a very dangerous treatment?

Answer 36

Low test threshold, high treat threshold

Question 37

What would your test-treat threshold look like for a non-serious disease with a non-serious treatment?

Answer 37

low test threshold, low treat threshold

Question 38

Why do you need to know pretest probability to use a diagnostic test?

Answer 38

You don’t know what the likelihood the person actually has the disease is unless you use the pre-test probability.

Question 39

How are pre-test probability and post-test probability used with regards to a test-treat threshold?

Answer 39

Pre-test is used to determine whether you do nothing, test, or just straight to treatment. If you’re above test threshold and you test, and your new post-test probability is above the treat threshold, you would treat.