EBM and Statistics

Question 1

What is sensitivity?

Answer 1

True positive test / All with disease

disease-centric

Question 2

What is specificity?

Answer 2

True negative test / All without disease

disease-centric

Question 3

Which of the following scenarios is sensitivity, specificity, PPV and NPV useful for:

1. Individual vs. Populations
2. Screening test vs. Diagnostic test

Answer 3

sensitivity/specificity are more suited to populations.
PPV/NPV are more suited for individuals

Screening test = high sensitivity
Diagnostic test = high specificity

Question 4

What is positive predictive value (PPV)?

Answer 4

Positive test / All positive test

test-centric

Question 5

What is negative predictive value (NPV)?

Answer 5

Negative test / All negative test

test-centric

Question 6

Draw a table for a diagnostic test and include arbitrary calculations for:

1. sensitivity
2. specificity
3. PPV
4. NPV
5. Accuracy

Answer 6

Description of table:
Top across (label): D+ / D-
Side down (label): T+ / T-
Top row (L to R): a, b
Bottom row (L to R): c, d

Row calculations:

1. sensitivity = a / (a+c)
2. specificity = d / (b+d)

Column calculations

3. PPV = a / (a+b)
4. NPV = d / (c+d)

Accuracy = (a+d) / (a+b+c+d)

Question 7

How is the accuracy of a test calculated?

Answer 7

Accuracy = (true positive + true negative) / all subjects

Question 8

What is positive likelihood ratio (LR+)?

Answer 8

The likelihood (probability) that a given test result would be expected in a patient WITH the target disorder

LR+ = sensitivity / [1 - specificity]

‘always choose sense over specifics’
LR+ is sensitive

Question 9

What is negative likelihood ration (LR-)?

Answer 9

The likelihood (probability) that this same result would be expected in a patient WITHOUT the target disorder

LR- = [1 - sensitivity] / specificity

‘always choose sense over specifics’
LR- is specific

Question 10

Using LR+ and LR-, calculate the post-test probability of positive or negative result.

Answer 10

Positive result: Post-test probability (odds) of disease = Pre-test probability (odds) x LR+

Negative result: Post-test probability (odds) of disease = Pre-test probability (odds) x LR-

Question 11

T/F: LRs are independent of the prevalence of the disease in the population, unlike predictive values.

Answer 11

True.

Question 12

Describe the difference in the different levels of RECOMMENDATIONS for investigations and therapy in EBM i.e. I, IIa, IIb and III.

Answer 12

Class I: benefit much great than risk, should be performed

Class IIa: benefit greater then risk, reasonable to perform

Class IIb: benefit greater than or equal to risk, may be considered

Class III: risk greater than benefit, harmful therefore do NOT perform

Question 13

Describe the difference in the different levels of EVIDENCE for investigations and therapy in EBM i.e. A, B and C.

Answer 13

Level A: multiple populations studied, RCTs and meta-analyses

Level B: limited populations studied, single RCT or non-randomised studies

Level C: very limited populations studied, consensus opinion, case studies or standards of care

Question 14

How is NNT calculated?

Answer 14

NNT = 1 / ARR

Question 15

For a population with normal distribution, what % of the population lies within:

1. +/- 1x standard deviation of the mean
2. +/- 2x standard deviations of the mean
3. +/- 3x standard deviations of the mean

Answer 15

1. 68%
2. 95%
3. 99.7%

Question 16

Case-control study is prone to what type of bias?

Answer 16

Recall bias