Bias

Question 1

Bias

Answer 1

Factor which causes SYSTEMATIC over- or under- estimate of a particular result

Question 2

Types of biases that can occur in the designing of the study?

Answer 2

Selection bias, volunteer bias, channelling bias

Question 3

Selection bias

Answer 3

Participants in research may differ systematically from the population of interest.

Question 4

Biases that can occur when conducting the research?

Answer 4

Interview bias, recall bias, Hawthorne bias

Question 5

Biases that can occur in writing and publishing the study?

Answer 5

Publication bias, confounding

Question 6

What is confounding bias…?

Answer 6

Confounding is when you have an independent factor which is associated with both the exposure factor and the outcome of interest with the disease

Question 7

How can you minimise confounding factors?

Answer 7

Minimise through stratification, regression analysis, randomisation

Question 8

What is internal validity and external validity?

Answer 8

Internal validity - accuracy of conclusions
External validity - generalisability of results

Question 9

Membership bias:

Answer 9

Explanation: Included patients are already participating in a study so might be more likely to look after their health, see the benefits of research, and educate themselves about treatment.

Question 10

Neyman/survival bias

Answer 10

Explanation of an example: Included patients who had to be alive months after the treatment, out-selecting those with more rapidly evolving conditions who may have realistic idea of the treatment effects

Question 11

Examples of confounders:

Answer 11

Age, ethnicity, sex, comorbidities (DM, alcohol intake), past surgery, previous family history, drug history, etc.

Question 12

Diagnostic purity bias

Answer 12

Comorbidity is excluded in the sample population, such that it does not reflect the true complexity of cases in the population

Question 13

Neyman bias (survival bias)

Answer 13

There is a time gap between the onset of a condition and the selection of the study population, such that some individuals with the condition are not available for selection.

Question 14

Response bias

Answer 14

Individuals volunteer for studies but they differ some way from the target population e.g. they are more motivated to improve health

Question 15

Lead-time bias

Answer 15

Screening/testing increases the perceived survival time without affecting the course of the disease

Question 16

Publication bias

Answer 16

A study that shows a significant difference between two interventions is more likely to be published than a negative study

Question 17

Observation bias

Answer 17

There is a problem with the way data is collected in the study, such that data has been unduly influenced by the expectations of the researchers and subjects

Question 18

Response bias

Answer 18

The subject answers questions in the way in which they think the researcher wants them to answer e.g. subjects in the experimental arm are more likely to give favourable responses

Question 19

Recall bias

Answer 19

Subjects selectively remember details from the past

Question 20

Exclusion bias

Answer 20

Results from differences in dropout rates between groups

Question 21

Hawthorne effect

Answer 21

Subjects alter their behaviour because they are aware they are being observed in a study

Question 22

Performance bias

Answer 22

Differences in care provided aside from intervention

Question 23

Detection bias

Answer 23

Differences in how outcomes are assessed between groups

Question 24

Aggregation bias

Answer 24

Occurs when it is wrongly assumed that the trends seen in aggregated data also apply to individual data points

Question 25

What is a method to assess the potential role of publication bias?

Answer 25

Funnel plot

Question 26

Ecological fallacy

Answer 26

The failure in reasoning that arises when an inference is made about an individual based on aggregate data for a group. This is an association for aggregate data in which the unit of observation is the country.

Question 27

Strategies to reduce confounding

Answer 27

Randomisation, restriction, matching, stratification, adjustment and multivariate analysis

Question 28

How does randomisation reduce confounding?

Answer 28

Aims to randomly distribute confounders between study groups

Question 29

How does restriction reduce confounding?

Answer 29

Restricts entry of individuals with confounding factors (risks bias in itself)

Question 30

How does matching reduce confounding?

Answer 30

Matching of individuals/groups aim for equal distribution of confounders

Question 31

How does stratification reduce confounding?

Answer 31

Confounders are distributed evenly within each stratum. Stratification allows the association between exposure and outcome to be examined within different strata of the confounding variable. This technique allows observation effects of an intervention in different subgroups (however, this becomes more difficult if there are several confounders). Creating several smaller groups reduces the power of the study.

Question 32

How does adjustment reduce confounding?

Answer 32

Usually distorted by choice of standard

Question 33

How does multivariate analysis reduce confounding?

Answer 33

Only works if you can identify and measure the confounders. This analysis is a statistical procedure for analysis of data involving more than one type of observation. Multivariate regression is an extension of multiple regression with one dependent variable and multiple independent variables.

Question 34

What are the advantages of multivariate analysis

Answer 34

Limited loss of power and ease of combining several confounders in one study