Calculations in EBM

Question 1

Therapy: CER is what?

Answer 1

CER = number of events in control group/number of subjects in control group

Question 2

EER is what?

Answer 2

number of events in intervention group/number of subjects in intervention group

Question 3

RR is what? RRR is what?

Answer 3

EER/CER;

1 - (EER/CER)

Question 4

ARD is what? NNT is what?

Answer 4

CER - EER;

1/ARD

Question 5

For the equations applicable to therapy, what variables can they be applied to?

Answer 5

Dichotomous outcomes ONLY!! NOT continuous data like means and medians, etc.

Question 6

What is the LNE and when would it be 0? 1? If a study’s CI crosses these lines, what does that mean?

Answer 6

0: absolute stats like ARR or mean difference
1: relative statistics like RR, OR;
It means the study would NOT be statistically significant

Question 7

Kaplan-Meier curve can help track people who ____ at different times; if people are removed once they’ve had an outcome, what is this called?

Answer 7

enter/leave;

CENSORING

Question 8

Diagnostics: Sens is what? Spec is what?

Answer 8

Sens = a/(a+c) (true positives/all disease positives);
Spec = d/(d+b) (true negatives/all disease negatives)

Question 9

Pos predictive value is what? Neg predictive value?

Answer 9

PPV = a/(a+b) (true positives/all test positives)
NPV = d/(d+c) (true negatives/all test negatives)

Question 10

Likelihood ratio for a positive test? Negative test?

Answer 10

LR pos = [a/(a+c)/b/(b+d)] = Sens/(1 - spec);

LR neg = [c/(a+c)/d/(b+d)] = (1-Sens)/Sepc

Question 11

SpPIn means what? SnNOut means what?

Answer 11

Test with high specificity means that with positive test you can rule in the diagnosis; test with high sensitivity means that a negative test rules out the diagnosis!!

Question 12

What stat won’t change with a change in prevalence?

Answer 12

Likelihood ratio (because of sens and spec!!)

Question 13

How good is a likelihood ratio?

Answer 13

> 10, 5-10, 2-5, 1-2 (excellent, good, fair, poor); for a LR neg, just inverse the numbers (<.1 is excellent, .1-.2 good, .2-.5 fair, .5-1 poor)

Question 14

NNH calculation?

Answer 14

1/AR (AR = EER - CER)