BPT Statistics

Question 1

Relative risk

Answer 1

RR = incidence in exposed / incidence in unexposed

RR = experimental event rate (EER) / control event rate (CER)

RR = (A / A + B) // (C / C + D)

Question 2

Odds ratio

Answer 2

OR = A x D / B x C

Measure of association between an exposure and outcome – or the odds that an outcome will occur given a particular exposure, vs the odds that the outcome will occur in the absence of that exposure

Question 3

Attributable risk

Answer 3

AR = (incidence in exposed) - (incidence in unexposed)

Difference in risk between exposed and unexposed groups; or the proportion of disease occurrences attributable to that exposure.

Question 4

Population attributable risk

Answer 4

PAR = (attributable risk) * (prevalence of exposure to risk factor in the population)

PAR = (AR) * (P)

Question 5

Number needed to treat

Answer 5

NNT = 1 / ARR

ARR: absolute risk reduction

Question 6

Type I error

Answer 6

Type I error (alpha error):

• accepting a statistically significant effect which exists due to chance / no such effect exists
• rejecting the null hypothesis when it is true
• p-values influence likelihood of making this error

Question 7

Type II error

Answer 7

Type II error (beta error) =

• there is a difference or relationship but it has not been detected
• inaccurately determining there is no difference, when in fact there is
• accepting the null hypothesis when it is actually false
• sufficiently powered studies reduce this error likelihood

Question 8

Selection bias

Answer 8

Selection of individuals, groups or data for analysis in such a way that proper randomization is not achieved - sample obtained is not representative of the population intended to be analysed
E.g. recruiting for a general population hypertension study from a renal clinic.

Question 9

Observation or measurement bias

Answer 9

Use of unreliable or invalid measurement of study variables, or obtaining non-comparable information between study groups.
E.g. asking for a patient’s weight instead of measuring it.

Question 10

Recall bias

Answer 10

Differences in accuracy of recalling past events and exposures by cases and controls. Common in retrospective studies, may be reduced by blind assessment of exposure and outcome variables.
E.g. Patients with cancer may more thoroughly recall and report risk factor exposures that have been publicised compared with controls.

Question 11

Lead time bias

Answer 11

Early detection confused with increased survival.

E.g. Erroneously determining increased survival in patients whose diseases are detected through screening tests rather than through normal detection, where screening has simply identified the disease earlier in its natural history without altering it.

Question 12

Sensitivity, specificity

Answer 12

Sens: A / A+C or TP / TP + FN
“Rules out”

Spec D / B+D or TN / FP + TN
“Rules in”

Question 13

Positive and negative predictive values

Answer 13

PPV = A / A + B or TP / TP + FP
NPV = D / C + D or TN /  FN + TN

Question 14

How is prevalence calculated?

Answer 14

Prevalence = (individuals with disease or attribute) / (population at risk) * 100

= (A + C) / (A + B + C + D) * 100

= (TP + FN) / ( TP + FP + FN + TN) * 100

Question 15

Positive and negative likelihood ratios

Answer 15

Informs the probability that a patient has the condition with a positive test, or does not have the condition with a negative test.
LR+ = sens / (1 - spec)
Likelihood of a positive test in a patient with the condition vs those without it
LR- = (1 - sens) / spec
Likelihood of a negative test in a patient without the condition vs those with it

Question 16

Experimental and control event rates

Answer 16

EER = TP / (TP + FP) = A / (A + B)

CER = FN / (FN + TN) = C / (C + D)

Question 17

Absolute risk reduction

Answer 17

ARR = [EER - CER]

EER = experimental event rate
CER = control event rate

Difference in risk between exposed and unexposed