Module 12

Question 1

What are two major sources of error in epidemiologic research?

Answer 1

Random (or chance) errors (reliability)
Systemic errors- all occurring for a specific reason (validity, bias)

Question 2

What can cause a lack of precision?

Answer 2

Sampling error, imprecision in measurement, and variability of the data

Question 3

What is the result of lack of precision?

Answer 3

Wider confidence interval
Makes it more difficult to reject the null hypothesis

Question 4

How can you reduce random error d/t lack of precision?

Answer 4

Increase the sample size and/or increase the number of measurements and taking the average of those measurements

Question 5

What can systematic errors be caused by?

Answer 5

Selection bias, information bias, or confounding

Question 6

What occurs as a result of systemic error?

Answer 6

Changes the value of results (odds ratio or relative risk) by moving it up or down

Question 7

How to reduce systemic bias

Answer 7

Prevent it when designing or carrying out the study or adjusting for confounding in the analysis

Question 8

What are the two components of validity?

Answer 8

Internal validity
External validity

Question 9

Characteristics of lack of internal validity

Answer 9

It’s the same as bias in a study.
It is systematic error.
It is affected by how the study was designed and carried out

Question 10

A study is said to have high internal validity when…

Answer 10

There have been proper selection of study groups-little selection bias
A lack of systematic error in measurement- little information bias,
And the results are not due to confounding

Question 11

High internal validity = ?

Answer 11

Low bias

Question 12

Definition of systematic errors?

Answer 12

Errors in how samples were selected or the quality of the data used

Question 13

What does external validity imply?

Answer 13

The ability to generalize beyond a set of observations to some universal statement

Question 14

What is external validity affected by?

Answer 14

The source population from which the sample is drawn and by the level of internal validity in the study

Question 15

Question asked for internal validity

Answer 15

Could the association be d/t bias (internal validity issues?) Selection bias and/or information bias

Question 16

Question regarding external validity

Answer 16

To whom does this association apply? To what extent may the findings from the study be generalized to other pops?

Question 17

Which is more important: internal or external validity?

Answer 17

A study can have good internal validity (not much bias) but still not be externally very valid (limited generalizability to a source pop)
However, a study cannot be externally valid without being internally valid
In other words, if a study has a great deal of bias, you cannot accurately generalize to a universal statement, or to any population

Question 18

Definition of bias

Answer 18

Deviation of results or interferences from the truth, or processes leading to such deviation. Any trend in the collection, analysis, interpretation, publication, or review of data that can lead to conclusions that are systematically different from the truth.

Question 19

When does overestimation of association occur?

Answer 19

If the two comparison groups are more different in the study than they are in reality

Question 20

Overestimation of association in case-control studies

Answer 20

The odds of exposure to a risk factor is overestimated in the cases
The odds of exposure to a risk factor is underestimated in the controls

Question 21

Overestimation of association in cohort studies

Answer 21

The incidence of the outcome in the exposed is overestimated
The incidence of the outcome in the unexposed is underestimated

Question 22

When does underestimation of the association occur?

Answer 22

If the two comparison groups are less different in the study than they are in reality

Question 23

Underestimation of the association in case-control studies

Answer 23

The odds of exposure to a risk factor is underestimated in the cases and overestimated in the controls

Question 24

Underestimation of the association in cohort studies

Answer 24

The incidence of the outcome in the exposed is underestimated and overestimated in the unexposed

Question 25

How bias occurs in descriptive studies

Answer 25

The prevalence (not just the prevalence ratio) in descriptive cross-sectional studies can be biased up or down This can be caused by selection bias of some sort (improperly selected sample) or information bias (data obtained from sample not correct)

Question 26

Definition of selection bias

Answer 26

A bias d/t differences in the manner in which study groups are formed: bias in how people are selected into the sample

Question 27

When does a selection bias occur?

Answer 27

Arises when the relationship between exposure and dz is different for those who participate and those who theoretically would be eligible for the study but do not participate

Question 28

Selection bias is not the same as...

Answer 28

Random sampling error

Question 29

Why is selection bias not the same as random sampling error?

Answer 29

They are both sampling problems, but for different reasons Sampling error is due to low sample size But selection bias is d/t recruitment of the wrong ppl. It cannot be corrected by increasing the sample size

Question 30

Types of selection bias

Answer 30

Non-response bias LTFU bias Healthy worker effect Exclusion bias Incidence-prevalence bias Hospital admission bias

Question 31

When does non-response bias occur?

Answer 31

When the response rate is low and non-respondents are systematically different than respondents

Question 32

Definition of response rate

Answer 32

The proportion of potential participants who are asked to participate in the study and actually did participate divided by the total who were asked to take part in the study (both those who accepted and those who refused)

Question 33

When can volunteer bias occur?

Answer 33

When people volunteer for a study, as volunteers are often systematically different than the general pop. This occurs when sample participants have been self-selected

Question 34

Which type of sample involves some degree of volunteer bias?

Answer 34

Convenience (nonprobability) survey samples

Question 35

When does LTFU bias occur?

Answer 35

When ppl drop out of a study, and those who drop out are systematically different than those who remained

Question 36

In what two situations does healthy worker effect occur?

Answer 36

1. When an occupational group is compared to the general pop, which includes ppl too unhealthy to work. 2. When "sicker ppl choose to work in environments where exposures are low, are excluded from being hired, or once hired, are transferred to less exposed jobs or leave work"

Question 37

What does healthy worker effect lead to?

Answer 37

Underestimate of the association

Question 38

When can exclusion bias occur?

Answer 38

If the exclusion criteria for participants are different for the comparison groups in the study

Question 39

When does hospital admission bias occur?

Answer 39

When hospitalized participants are used as a "healthy" comparison group, but actually are not representative of people without the dz being studied. This can happen if hospitalization rates differ for the exposed and not exposed. The association between exposure and dz in the hospitalized sample may not reflect what it is in non-hospitalized ppl

Question 40

Why can incidence-prevalence bias occur?

Answer 40

Because prevalence is influenced by the duration of the dz and whatever is influencing that- not just what causes one to get the dz in the first place. Including participants who are prevalent cases can introduce bias, since what led to their survival (and hence inclusion in the study) may be something different than what caused them to get the dz

Question 41

Common forms of selection bias in cross-sectional studies

Answer 41

Non-response bias Volunteer bias Incidence-prevalence bias

Question 42

Common types of selection bias in case-control studies

Answer 42

Non-response bias Exclusion bias Incidence-prevalence bias Hospital admission bias

Question 43

More details about selection bias in case-control studies

Answer 43

More of a problem for case-control studies than some other study designs Has often to do with how the controls are selected

Question 44

How can bias occur in selection of controls in a case-control study?

Answer 44

Controls may be selected into the study based in some way on the exposure, so that controls do not represent the source pop in terms of exposure status. As a result, the relationship between exposure and dz among participants in the study differs from what the relationship would have been along individuals in the pop of interest.

Question 45

How to reduce selection bias among controls

Answer 45

Recruit controls from the same source pop as the cases If unsure of the exact sampling frame, attempt to draw controls from a variety of sources. Compare the prevalence of the exposure among controls with other sources to evaluate credibility

Question 46

How to reduce selection bias due to unrepresentative cases

Answer 46

Develop an explicit (objective) case definition Enroll all cases in a defined time and region Ensure that all medical facilities are thoroughly canvassed Consider whether all cases require medical attention; consider possible strategies to ID where else the cases might be ascertained Try to recruit incident (new) rather than prevalent (all) cases, if possible

Question 47

Common forms of selection bias in cohort studies

Answer 47

LTFU bias Healthy worker effect

Question 48

How to minimize LTFU

Answer 48

Restricting the study participants to those likely to remain in the study Collecting personal identifying info Making periodic contact, and providing incentives Recruiting a new cohort as time goes on to replace those LTFU (and who die)

Question 49

How to minimize healthy worker effect

Answer 49

Select a comparison group made up of a similar type of worker, who were unexposed

Question 50

Definition of information bias

Answer 50

Results from systematic differences in the way data on exposure or outcome are obtained from the various study groups It is a bias in terms of the data itself and its ascertainment Can be introduced as a result of systematic measurement error in assessment of either exposure and/or dz

Question 51

Types of information bias

Answer 51

Recall bias Reporting bias ---Social desirability bias ---Wish bias Interviewer/abstractor bias Surrogate interview bias Exposure suspicion bias Detection bias

Question 52

Definition of recall bias

Answer 52

Occurs when one group remembers the past differently (more or less accurately) than the other

Question 53

Definition of reporting bias

Answer 53

Can occur whenever any self-reported data are collected. This is any incorrect information given by the respondents for whatever reason, intentional or unintentional

Question 54

Definition of social desirability bias

Answer 54

Aka the Clever Hans Effect or Obsequiousness Bias. A type of reporting bias in which participants lie about the data in order to please the interviewers

Question 55

Wish bias definition

Answer 55

A type of reporting bias in which cases seek to show that the dz was not their fault. Such participants deny certain negative lifestyle exposures, or overestimate workplace exposures. This form of bias may be unconscious.

Question 56

Interviewer/abstractor bias or observer/interviewer bias

Answer 56

Occurs when interviewers probe more thoroughly for an exposure or outcome in one comparison group than in the other (or when they abstract data more thoroughly in one group than another)

Question 57

When can interviewer/abstractor bias occur?

Answer 57

In any type of unblinded or single blinded study where there are two or more comparison groups -Case-control studies -Cohort studies -Clinical trials that are not double blinded

Question 58

How to prevent interviewer/abstractor bias and the placebo effect

Answer 58

In a double-blind study neither the participants nor the investigators know who is receiving the active tx

Question 59

How to minimize interviewer/abstractor bias when there is no double blinding possible in observational studies

Answer 59

If interviewers do not know the exposure status of the participants, there may be less likelihood of this sort of bias (a kind of blinding) Even if the interviewers are blinded as to who is exposed and who is not, the participants themselves know

Question 60

When does surrogate interview bias occur?

Answer 60

When participants are dead or unable to be interviewed, and surrogates are interviewed in their place

Question 61

Exposure suspicion bias definition

Answer 61

This type of information bias is closely related to, but not quite the same as interviewer/abstractor bias. It occurs when knowledge of cases' and controls' outcome status determines how the exposure is assessed.

Question 62

Detection bias definition

Answer 62

Aka surveillance bias. When one pop or group is monitored more thoroughly over time than the other

Question 63

Common forms of information bias in cross-sectional studies

Answer 63

Reporting bias Recall bias Surrogate interview bias

Question 64

Common forms of information bias in case-control studies

Answer 64

Reporting bias Recall bias Interviewer/abstractor bias Surrogate interview bias Exposure suspicion bias

Question 65

Common forms of information bias in cohort studies

Answer 65

Reporting bias Interviewer/abstractor bias Detection or surveillance bias

Question 66

What is the result of information bias?

Answer 66

People are misclassified into the wrong groups Should be exposed but are listed as unexposed or vice versa Or should be diseased but are listed as not diseased or vice versa The right people are in your study, but under the wrong labels Can occur in any study design

Question 67

What are two types of misclassification?

Answer 67

Non-differential (less serious) Differential (more serious)

Question 68

Definition of non-differential misclassification

Answer 68

When the direction and extent of misclassification is equal between comparison groups. It occurs in both groups

Question 69

What does non-differential misclassification always bias?

Answer 69

Odds ratio, relative risk, prevalence ratio or correlation coefficient toward the null

Question 70

Differential misclassification definition

Answer 70

When the direction and/or extent of misclassification is not equal between comparison groups, or misclassification only occurs in one of the comparison groups

Question 71

What calculations can differential misclassification bias?

Answer 71

The odds ratio, relative risk, etc. in any direction

Question 72

Techniques to reduce information bias

Answer 72

Use memory aids to avoid recall bias Validate exposures via medical records or blood tests as opposed to questioning the participant Provide standardized training sessions and protocols for interviewer to avoid interviewer/abstractor bias Use standardized data collection forms to avoid interviewer/abstractor bias Blind or mask when possible in clinical trials

Question 73

What are key questions when evaluating an epidemiologic study in terms of information bias?

Answer 73

What bias might have been introduced into the findings by the approach to selection of participants? How were the comparison groups chosen? Were selection criteria clearly defined? What were the response rates? Were all of the participants in the sample used in the analysis of the article? Why were some left out?

Question 74

What are key questions when evaluating an epi study in terms of information bias?

Answer 74

Is there any evidence for bias by misclassification of the outcome or the exposure (information bias)? Is there likely to be recall or reporting bias? Could there be any interviewer/abstractor bias? Were the measurement and classification methods consistent for all participants?

Question 75

When to control selection and information bias

Answer 75

Selection and information bias can only be prevented and controlled during the design (planning) and conduct of a study- not later -Choice of a study population (selection) -Methods of data collection (information)