5) Bias And Confounding

Question 1

Sources of selection bias

Answer 1

• inapporpriate selection of subjects from the study population
• non random allocation of exposure
• failure to include subjects in analysis

Question 2

RCT

Answer 2

• randomly allocate exposure and conceal allocation
• minimise loss to follow up
• analyse by intention to treat in order to pressure the two groups exactly how they were at the outset
• losses before randomisation do not lead to selection bias

Question 3

Performance bias

Answer 3

• systematic difference in the care provided to the participant in the comparison groups other than the intervention under investigation

Question 4

Measurement bias

Answer 4

• Systematic error in the estimate of frequency or effect due to measurement error or misclassification of the study factors, outcome factors or confounding factors

Question 5

Validity

Answer 5

• The degree to which a measurement or test measures what it is suppoes to measure

Question 6

Repeatability/reliability

Answer 6

Ability of a measure or test to produce the same result when used repeatedly in the same person

Question 7

Sensitivity

Answer 7

• the proportion of people who in truth have the attribute who also have a positive test
• the proportion of people with a disease that test positive

Question 8

Specificity

Answer 8

• the proportion of people who have a negative result and do not have the disease

Question 9

Confounding

Answer 9

• occurs when a measure of effect is biased because of the association of the exposure with other factors that influence the outcome

Question 10

Criteria for confounding

Answer 10

• independent predictor of outcome
• not an intervening variable
• associated with study facrtor - unequally distributed in the exposure group

Question 11

Practical test to know whether or not there is confounding

Answer 11

• if there is a difference between crude and stratified measures then there is confounding, if no - no confounding

Question 12

Control of confounding at the design stage

Answer 12

• randomisation
• restriction
• matching

Question 13

Control of confounding at the analysis stage

Answer 13

• standardisation
• stratification
• multivariate modellinG

Question 14

Effect modification

Answer 14

• an innate aspect of the relationship between the exposure, the effect modifier, and the outcome
• cannot be removed through design or analysis, unlike the confounding effect