Lecture 13: Bias

Question 1

Give the definition of bias

Answer 1

• Systematic (non-random) error in study design or conduct leading to Erroneous Results.
• Bias distorts the relationship (association) between exposure and outcome

Question 2

What can be done to fix a bias?

Answer 2

Nothing can be done to “fix” a bias once it has already occurred (after study end). He laid that one on thick.

• Prospective (pre-study) consideration and adjustment can minimize bias and its impact o
• Still must assess for it to confirm internal validity and that conclusions are correct

Question 3

What are the 3 elements to be associated with a bias?

Answer 3

• Source/Type
• Magnitude/Strength - Bias can account entirely for a weak association (a small RR/OR) but is not likely to account entirely for a very strong association (a large RR/OR)
• Direction - Bias can over- or under-estimate the true measure of association oBias can have a enhancing or minimizing effect on the true measure of association (towards or away from 1.0)

Question 4

What are the 2 Main Categories of Bias?

Answer 4

1. Measurement-Related

2. Selection-Related

Question 5

Describe Measurement-Related Bias

Answer 5

Measurement’-related (Information/Observation):

• Any aspect in the way the researcher collects information, or measures/observes subjects which creates a systematic difference between groups
• Errors in measurement can also cause a resultant error in patient classification (misclassification, more later in lecture)

Question 6

Describe Selection-Related Bias

Answer 6

• Any aspect in the way the researcher selects or acquires study subjects which creates a systematic difference between groups
• Commonly seen when comparative groups not coming from same population/group or not being representative of the full population or even differentially selected (processes)
• This is the #1 bias in scientific literature. Just because it’s really hard to get a truly random selection. And that might be testable.

Note: Intentionally restricting a population with intention and reason to do so up front is not a bias.

Question 7

What types of selection biases are common?

Answer 7

They way study subjects are selected generates differences in groups (very commonly encountered)

• Key Examples:
• Healthy-Worker bias: Can easily be seen in prospective Cohort studies
• Self-Selection/Participant (Responder) bias: Those that wish to participate (volunteer) may be different in some way to those that don’t volunteer or self-select (refusal/non-response) to participate
• Control Selection bias: Can easily be seen in Case-Control studies

Question 8

What can occur in cross-sectional studies?

Answer 8

Cross-Sectional studies are subject to Neyman bias (a.k.a., selective survival)

• More easily descriptive for longer-lasting & more indolent diseases
• Not effective for acute or rapidly fatal diseases

Question 9

Name 4 types of Subject-Related Biases.

Answer 9

1. Recall/Reporting Bias
2. Contamination
3. Compliance/Adherence Bias
4. Lost to Follow-Up Bias

Question 10

Describe a Recall/Reporting Bias

Answer 10

A differential level of accuracy/detail in provided information between study groups

• Exposed or diseased subjects may have greater sensitivity for recalling their history (better memory; easier to remember if more severe) or amplify (exaggerate) their responses
• Individuals can report their “effects” of exposure, disease symptoms or treatment differently because they are part of a study
• “Hawthorne Effect” (I dunno)

Question 11

Describe the other 3 subjective biases

Answer 11

• Contamination bias: Members of the control group accidently, or outside of the study protocol, receive the treatment (or similar) or are exposed to the intervention being studied
• Compliance/Adherence bias: Groups being interventionally studied have different compliances
• Lost to Follow-up bias: Groups being studied have different withdrawal or lost to follow-up rates OR there are other differences between those that stay in the study and those that withdraw or are lost to follow-up.
• -Differential vs. Non-Differential

Question 12

What are 2 types of Observer-Related Biases

Answer 12

1. Interviewer/Proficiency Bias

2. Diagnosis/Surveillance Bias

Question 13

Describe Interviewer/Proficiency Biases

Answer 13

• A systematic difference in soliciting, recording, or interpreting on the part of the researcher (or their assistants)
• Interviewers knowledge may influence the structure, or tone, of questions or follow-up questions which may influence response from the study subject OR,
• Interventions/treatments are not applied equally between groups due to skill or training differences of study personnel or differences in study procedure compliance by staff at different sites
• Can be conscious or unconscious actions of the interviewer

Question 14

Describe the Diagnosis/Surveillance (Expectation) Bias

Answer 14

• Different evaluation, classification, diagnosis, or observation between study groups
• Observers may have preconceived expectations of what they should find in examination, evaluation, or follow-up
• “Hawthorne-Like Effect” from the researchers’ perspective

Question 15

What is a Misclassification Bias

Answer 15

This is a type of measurement bias. Basically stating individual must be accurately placed into the right group. Such as person with TB getting placed into group without TB because that person was asymptomatic or the measurement was wrong.

Question 16

Describe a Non-Differential Misclassification Error

Answer 16

Non-differential (error in both groups equally):
• Misclassification of exposure or disease which is unrelated to the other (disease or exposure), depending on study design
• Effect = For dichotomous (2 category) variables, bias can move the measure of association (RR/OR) towards 1.0; it attenuates your effect estimates of association
- Example: RR of 0.3 moves to 0.7 -> attenuation towards 1.0
- Example: OR of 1.9 moves to 1.2 -> attenuation towards 1.0
- In both cases, they both move closer to 1.0

This is the situation where if a screw-up is non-differential, it doesn’t severely affect the study. But it’s still an error.

Question 17

Describe a differential misclassification error

Answer 17

Differential (error in one group differently than other):
• Misclassification of exposure or disease is related to the other (disease or exposure), depending on study design
• Effect = Bias can move the measure of association (RR/OR) in either direction in relation 1.0; it can inflate or attenuate your effect estimates of association
- Example: RR of 0.8 moves to 0.2 or 1.4 moves to 2.1 -> inflation away from 1.0
- Example: OR of 2.3 moves to 1.1 or 0.6 moves to 0.9 -> attenuation towards 1.0
- These Can move away from or towards 1.0

Question 18

What are the most important aspects in controlling biases?

Answer 18

Select the most precise, accurate, & medically-appropriate measures of assessment and evaluation/observation
• Highest sensitivity/specificity and validated screening tools
- Use published/past-utilized techniques, if possible
- Calibrate & Test equipment, techniques and processes
•Plan specifics of data collection as much as possible, test-run forms, interviews and survey-questions
- Train physicians, researchers and assistants on processes
- Use technology as much as possible
- Blinding/Masking
- Use multiple sources to gather all information
- Randomly allocate observers/interviewers for data collection (and train them!; use technology!)
- Build in as many methods necessary to minimize loss to follow-up
–Lost-to-follow-up bias (Differential Attrition bias)