Lecture #20 - Bias Pt2 (attempt 2)

Question 1

1. Define information bias

Answer 1

Systematic error arising from procedures used for classification or measurement of outcome/exposure

This (systematic) measurement error is the same between the two groups = non-differential

This (systematic) measurement error is not the same between the two groups = differential

Question 2

1. How is data collected in a study and how can measurement error occur in these two ways?
2. What kind of information is it that is gathered in these two ways?
3. How do we want the data to be collected?

Answer 2

3. Same for everyone

Question 3

Measurement error - it can be both random and systematic. True or false?

Why is measurement error a problem?

What effect can measurement error have in a descriptive study? What about in an analytic study?

Answer 3

-

Question 4

Measurement error in a descriptive study (cross-sectional) - what might you want to ask yourself?

Answer 4

-

Question 5

Measurement error in an analytic study - can lead to what? What are the two types? How does each affect the estimate?

Answer 5

Estimate closer to null in non diff and can be either away or towards null in differential

Question 6

Measurement error in analytic studies; differential misclassification (or differential error):

1. Cross Sectional studies - what could lead to differential misclassification here? Like, what could lead to the measurement error being different in these two groups of exposure vs non exposure?
   - What did Page 19 say for example

Answer 6

1. People with the outcome might report the exposure differently to those without the outcome. Like, those people with diabetes are more likely to say they ate more sugar than those without diabetes. I think. Or with objective info - if the machine weighed heavy people 2kg heavier and the normal people normally?
   - “People with outcome might report the exposure differently to those without the outcome”

Question 7

2 Case-control - what can you have problems with?

• Define this problem
• Explain this
• How does it affect the MoA?

-Page 19

Answer 7

2. “Systematic error due to differences in accuracy or completeness of recall to memory of past events or experiences.”
   - It’s basically when cases are trying to actively recall why they got the disease (their exposure status) and the controls aren’t too bothered (differential recall).
   - So the measure of association: It affects the controls because they aren’t remembering their exposure right and it decreases the odds of ratio in controls therefore increases the true odds ratio
   - Can try minimise this by using objective measures that doesn’t involve subjective measures and not self-reported because that could lead to difference in recall. Also, can validate self-reported measures with other information (e.g. biochemical markers to make sure they’re saying the right thing). Also, memory aids - show them pictures to minimise recall bias.

• “Cases might more accurately recall past exposure compared to controls”
• “An interviewer who’s aware they are interviewing a case might ask more probing questions about exposure of interest” (So can get interviewer bias here too?)

Question 8

3. Cohort studies
   - Potential for what?
   - What two things do we consider carefully?
   - Differential misclassification of two things - how? What kind of bias is the second one?
   - How do you minimise interviewer bias? (4)

Answer 8

• Potential for misclassification of exposure/outcomes
• Consider carefully: have participants been correctly classified? Has outcome status been correct classified?
• Differential misclassification: if classification of exposure depends on outcome (outcome has occurred in historical cohort - not a problem in prospective
• Differential misclassification: If classification of outcome depends on exposure e.g. if interview knows exposure + examined the outcome differently for those in exposed group compared with those in comparison group = interviewer bias

-Can minimise by
=have a clearly defined study protocol and measures
=structured questionnaire and standard prompts (same questions etc for all participants)
=training of interviewers (so they don’t ask more questions or more probing questions to the exposed group - so you’re systematically measuring the exposed group differently and this can lead to information bias)
=Blinding
(can also like time the interviewer)

-“An interviewer aware of the exposure status may ask more probing questions about outcome among exposed compared with those in comparison group (similar can happen in RCT)”

Question 9

4. RCTs

- What are two ways in which bias can occur?

Answer 9

-

Question 10

How do you minimise bias? (3)

Answer 10

-

Question 11

Publication Bias

1. Define it
2. Explain it

Answer 11

-