Evidence based medicine

Question 1

What do case-control studies look at?

Answer 1

compare risk factor frequency
-useful to study individuals w/ a disease compared to individuals without the disease
-evaluates risk factors and outcomes for the disease

Question 2

What do retrospective cohort studies look at?

Answer 2

compare disease incidences
-follows a population of patients with a disease over time
-likely to yield important information about the natural history of a disease

Question 3

What do prospective cohort studies look at ?

Answer 3

compare disease incidences

Question 4

Which tests can be used for a single sample metric test?

Answer 4

-t test
-Z test

Question 5

Which tests can be used for multiple, independent samples in a metric test?

Answer 5

-two group t-test
-Z test

Question 6

Which test can be used for paired samples in a metric test?

Answer 6

paired t-test

Question 7

What test can be used for single sample nonmetric tests?

Answer 7

-chi-square
-K-S
-runs
-binomial

Question 8

Which tests can be used for multiple, independent samples in a nonmetric test?

Answer 8

-chi-square
-Mann-Whitney
-median
-K-S

Question 9

Which tests can be used for paired samples in a nonmetric test?

Answer 9

-sign
-Wilcoxon
-McNemar
-chi-square

Question 10

What are examples of metric (parametric) tests?

Answer 10

-interval = interval between observations in terms of fixed until of measurement (temperature)
-ratio = scale has a fundamental zero point (age, income)

Question 11

What are examples of nonmetric (nonparametric) tests?

Answer 11

-nominal = define an attribute (gender, marital status)
-ordinal = rank or orders the observation as scores or categories from low to high in terms of “more or less” (education, attitude, opinions)

Question 12

What is a type 1 error?

Answer 12

-alpha error
-incorrectly rejects null hypothesis
-says there’s a difference when there really isn’t
-alpha is the probability of making a type 1 error

Question 13

What is a type 2 error?

Answer 13

-beta error
-incorrectly accepting the null hypothesis
-says there’s no difference when there is
-beta is the likelihood of making a type 2 error
-results in there are too few subjects

Question 14

What is p-value?

Answer 14

risk of a false positive result due to chance (describes the strength of the result)
-if < 0.01 highly statistically significant
-if </= 0.05 statistically significant
-if > 0.05 statistically insignificant

Question 15

What is statistical power?

Answer 15

likelihood a study correctly disproves the null hypothesis
-likelihood of finding a difference between two groups

Question 16

How do you calculate power?

Answer 16

power = 1-beta

Question 17

What does the power of a study depend on?

Answer 17

-effect size
-level of significance desired (p value)
-sample size

Question 18

What is a confidence interval?

Answer 18

range w/in which magnitude of effect lies

Question 19

What is the confidence interval if alpha = 0.05?

Answer 19

95%

Question 20

When is a confidence interval statistically significant?

Answer 20

-CI does not include or span the “no effect point” of one or zero for risk ratios like relative risk and odds ratio
-numbers do no overlap

Question 21

Define sensitivity.

Answer 21

proportion of patients with a disease who test positive

Question 22

Define specificity.

Answer 22

proportion of patients without a disease who test negative

Question 23

How do you calculate PPV?

Answer 23

PPV = A/A+B (true positive/total number of positives)

Question 24

How do you calculate NPV?

Answer 24

NPV = D/D+C (true negative/total number of negatives)

Question 25

One the 2x2 grid what category is on top (x-axis) and which is on the side (y-axis)?

Answer 25

-x = disease
-y = test

Question 26

How do you calculate sensitivity?

Answer 26

A/A+C

Question 27

How do you calculate specificity?

Answer 27

D/B+D

Question 28

How do you calculate relative risk reduction?

Answer 28

RRR = (1-RR)/RR

Question 29

How do you calculate number needed to treat?

Answer 29

100/absolute risk reduction