Lecture 13 - Bias

Question 1

How asses for internal validity of study design?

Answer 1

Look for bias

Question 2

3 Components of bias

Answer 2

1. Source/type
2. Magnitude/strength
3. Direction

Question 3

Can bias be fixed after study end?

Answer 3

No

Question 4

Measurement related bias

Answer 4

Errors in data collection that creates systematic difference in groups

Question 5

Selection related bias

Answer 5

Error in way researcher selects study subjects that creates systematic difference in groups

Question 6

Selection biases:

1. Healthy worker bias
2. Self-selection/participant bias
3. Control selection bias

Answer 6

1. Pick those working but people working like healthy
2. Volunteers might be different than those that choose not to
3. If choose control as first 500 to answer phone that don’t have diabetes (some may not have phone or are at work)

Question 7

Recall (Reporting) bias

Answer 7

Exposed or diseased have better recollection or exaggerate responses

Question 8

Hawthorne effect (Part of recall bias)

Answer 8

Subjects report effects differently becasue part of study. Overly enthusiastic

Question 9

Contamination bias

Answer 9

Control group accidentaly receive treatment that exposed did or received it outside of study

Question 10

Compliance/adherence bias

Answer 10

Groups being interventionally studied have diff compliances

Question 11

Lost to follow up bias

Answer 11

Groups have diff withdrawal or lost to follow up rates or other differences btw those that stay in study and those withdraw or lost to follow up (unreachable)

Question 12

Subject related biases

Answer 12

Things related to subject groups that creates systematic difference in groups

Question 13

Observer related variation

Answer 13

Systematic difference in groups caused by study coordinators/observers

Question 14

Interview (proficiency) bias

Answer 14

Systematic difference in soliciting, recording, or interpreting on part of researcher, unconscious or conscious. Interviewers knowledge may affect how ask questions which influence response from subjects

Question 15

Diagnostic/surveillance bias

Answer 15

Different evaluation, classification, diagnosis, or observation btw study groups.

Question 16

Hawthorne like effect in surveillance bias

Answer 16

Observers may have preconcieved expectations of what they should find in study.

Question 17

Misclassification bias

Answer 17

Measurement bias where there’s an error in classifying either the disease or exposure status or both. Put people in wrong group.

Question 18

Non-differential misclassification bias

Answer 18

Error both groups equally

Question 19

Effect on non-differential for dichotomous (2 category) variables

Answer 19

Moves measure of association (RR/OR) towards 1.Reduces magnitude of (attenuates) effect

Question 20

Differential misclassification bias

Answer 20

Error in one group different than in other

Question 21

Effect of differential

Answer 21

Move the OR/RR away from or towards 1.0. Can increase magnitude or decrease magnitude of association.

Question 22

Non-differential also called…..

Answer 22

Conservative

Question 23

Ways to control for bias

Answer 23

1. Most accurate data collection tools and methods
2. Blinding/masking
3. multiple sources to gathers info
4. Randomly allocate observers/interviewers
5. decrease lost to follow up
6. Use technology
7. Train observer/interviewers

Question 24

Ways to control for misclassification

Answer 24

Same as for biases. Misclassif is a measurement bias

Question 25

If going to be errors want to be non-differential or differential?

Answer 25

Non-differential becasue will be balanced and will only decrease magnitude of effect