Information bias

Question 1

Synonym for information bias?

Answer 1

Measurement error

Question 2

What is information bias?

Answer 2

A systematic error that arises due to the collection of erroneous information from study participants

Question 3

Regarding what types of variables is information bias relevant?

Answer 3

• Exposure
• Outcome
• Covariate

Question 4

When is misclassification not differential?

Answer 4

When errors on the variable do not depend on the value of other variables

Question 5

What is the effect of non-differential misclassification of a dichotomous exposure/outcome?

Answer 5

Bias towards the null

Question 6

The degree of non-differential misclassification bias is a function of:

Answer 6

1. Sensitivity
2. Specificity
3. Prevalence of exposure

Question 7

About non-differential misclassification bias: when is specificity critical?

Answer 7

When exposure is rare

Question 8

About non-differential misclassification bias: when is sensitivity critical?

Answer 8

When exposure is common

Question 9

About non-differential misclassification bias: what is critical, at fixed values of specificity and sensitivity?

Answer 9

The prevalence of the exposure

Question 10

What is the effect of non-differential misclassification of a multi-categorical exposure/outcome?

Answer 10

Bias in either direction (toward or away from the null)

Question 11

When does non-differential misclassification NOT result in bias?

Answer 11

• When there is perfect specificity, for RR estimates (e.g., odds ratio)

Question 12

When do dependent errors in the measurement of the treatment and outcome occur?

Answer 12

When the same source is used to gather information on the exposure and outcome

Question 13

What is the result of dependent non-differential misclassification?

Answer 13

Bias in either direction

Question 14

3 exceptions to “non-diff misclassification biases towards the null rule”

Answer 14

1. Multi-categorical exposure (away or towards)
2. Perfect specificity (no bias for RR estimates)
3. Dependent non-differential misclassification (away or towards)

Question 15

The effect of differential misclassification is a function of:

Answer 15

The sensitivity, the specificity and the prevalence of exposure between cases and control

Question 16

3 types of information bias in the collection of the exposure information

Answer 16

1. Recall bias
2. Interviewer bias
3. Changes in exposure status over time

Question 17

2 types of information bias in the collection of outcome information

Answer 17

1. Observer/detection/surveillance bias

2. Reporting bias

Question 18

Difference between poor recall and recall bias?

Answer 18

Poor recall is non-differential

Question 19

In what type of study design is recall bias especially an issue?

Answer 19

Case-control studies

Question 20

What is recall bias?

Answer 20

Differential errors in recall of past exposures based on disease status

Question 21

What is interviewer bias and when does it occur?

Answer 21

Bias in the ascertainment of the exposure due to the interviewer not being blind to the disease status (e.g., prompt or emphasize certain questions)

Question 22

What is the observer bias?

Answer 22

The collection of different qualities of information about the outcome for exposed vs unexposed, affected by knowledge of the exposure

Question 23

What is reporting bias?

Answer 23

When the information on the outcome is obtained by participant response, especially for events where it is hard to have objective information (may overreport if they know the study hypothesis or underreport if socially undesirable)

Question 24

What is the link between information bias and confounding?

Answer 24

Information bias can cause residual confounding, such as when non-differential misclassification of a confounding results in imperfect adjustment

Question 25

How can we prevent recall bias?

Answer 25

• Frame questions to aid accurate recall (design level)
• Use different control group (e.g., different disease)
• Objective/administrative information
• Nested case-control studies (provides longitudinal exposure information)

Question 26

How can we prevent interviewer bias?

Answer 26

• Standardization
• Blinding interviewers to the outcome
• Validated instruments
• Quality control of interviews (must be systematic)
• Diagnose interviewer bias with “phantom” control

Question 27

What is a phantom control?

Answer 27

Manipulated controls, e.g. they are mistaken for a case, so you can compare if they’re assessed like the other controls or like the other cases

Question 28

How do we prevent observer bias?

Answer 28

• Blind to exposure status

- Multiple observers

Question 29

How do we prevent reporting bias?

Answer 29

• Confirm self-report with administrative data

- Use of validated instruments and computerized surveys for sensitive information