Bias and Confounding

Question 1

What kind of error is due to bias?

Answer 1

Systematic

Question 2

Is there a formal way to deal with systematic error?

Answer 2

NO

Question 3

What is validity in epidemiological context?

Answer 3

Relates to the absence of systematic error in a study result

*a valid measure of association in a study will have the same value as the true measure in the source population, except for error due to random variation

Question 4

What is bias in the context of epidemiological studies?

Answer 4

The extent to which a measure of association from a study differs from the true measure of association in the source population
**only for differences due to systematic error

Question 5

T/F: Systematic error can make a study result biased

Answer 5

True

systematic error
causes bias

Question 6

Bias causes a study to be _______

Answer 6

Invalid

Question 7

T/F: Random error can make a result biased

Answer 7

false

just makes the study wrong

Question 8

What does it mean when a study value is accurate?

Answer 8

The study was free from bias and is valid

Question 9

Can you determine whether a study has a systematic error when results are inaccurate the first time the study is run?

Answer 9

NO

systemic errors are repeatable - so on the first go at the study - you can’t determine whether your error is random or systematic

Question 10

Will random errors result in inaccurate values that are far off from expected or close to expected?

Answer 10

Far off or close

Random errors can be large or small: the key is - each value will differ greatly (making it non repeatable and unbiased)

Question 11

When you perform a study multiple times, and get values that are far from expected, but similar values each time, what kind of error(s) do you have?

Answer 11

Systematic - since the study repeatedly failed in the same way

*random error (in this case small) because the values were not exactly identical

Question 12

What will ensure validity in epidemiological studies?

Answer 12

Proper study design

Question 13

What is internal validity?

Answer 13

The study result is valid with respect to the population under study

Question 14

What is external validity?

Answer 14

The study result is valid to a wider population.. beyond the study population and or source population

aka generalizability

Question 15

Example study: assessing associations between risk factors and lymphoma in a sample of GR without lymphoma in Maine..

The study may be ____ valid if the study association was true

But may note be ____ invalid for other breeds, or for dogs in different populations/locations

Answer 15

internally valid (within the study population)

externally invalid (within all populations)

Question 16

What is the study population, source population, and other populations, in the context of epi studies?

Answer 16

Study pop = the subjects in the study

Source pop = population from which the subjects were drawn

Other pop = target populations —> populations to which we may want to generalize our results

Question 17

____ causes the measure of association estimated from the study population to be different from the true measure of association in the source population

Answer 17

Bias

**bias causes the study results to be different from those that truly exist in the source population

Question 18

What is a non-differential bias?

Answer 18

A bias that equally affects the study groups

• diseased and non-diseased are equally biased
• exposed and non-exposed are equally biased

Question 19

What is a differential bias?

Answer 19

A bias that affects one group more than another

Question 20

Data from a study:

In the source population: OR = 2

• the true proportion of the dz cases with exposure was 20%
• the true proportion of controls with exposure was 10%

In the sample collected from the study: )R = 2

• the dz cases had 10% with the exposure
• the controls has 5% with exposure

Is this differential bias or non-differential bias?

Answer 20

Non differential

The bias was a 50% reduction in the proportion exposed in both the dz cases and the control populations

Question 21

Two general sources of bias are ____ bias and _____ bis

Answer 21

Selection bias: error in selection of study subjects (the sample is different from the population)

Information bias: error in measurement (measure things with error or incorrectly) aka misclassification bias

Question 22

Besides bias, what can contribute to systematic errors?

Answer 22

Confounding - an unknown factor distorts the relationship between the exposure and outcome

Question 23

What does selection bias usually result from?

Answer 23

From groups being selected for the study not coming from the complete source population

Question 24

T/F: Most veterinary observational studies have selection bias

Answer 24

TRUE

This is because samples are collected from convenient locations that provide easy access to animals

Ex: veterinary clinics, abattoirs, shelters (these don’t represent entire animal populations)

Question 25

How do you reduce selection bias in a cross sectional study?

Answer 25

Random samples: identify all the individuals in the target population and randomly sample them

Question 26

How do you reduce selection bias in case control studies?

Answer 26

Matching - ensure that controls have similar demographics, age, breed, gender, etc as cases

Question 27

How do you reduce selection biased in cohort studies?

Answer 27

Matching: ensure that exposed and non exposed have similar demographics

Question 28

What are some examples of selection bias?

Answer 28

1. self selection bias - studies based upon volunteers 2. Healthy worker effect: studies or working people or working animals -who are healthier than the population as a whole 3. Diagnostic bias: diagnosis of a dz may be influenced by their knowledge of the exposure and dz (this can be reduced by having a clear, well defined case definition)

Question 29

What will reduce information bias in cross sectional descriptive studies?

Answer 29

evaluate the accuracy of measuring tools and adjust your estimates to reflect the error **error in measurement of dz in the sample may result in the prevalence estimate in the sample being different than the true prevalence of in the target population

Question 30

If there is informational bias in a case control study, what will be affected?

Answer 30

The association will be bias Ex: in a case control study, owners of sick pets (cases) are more likely to remember exposures than owners of healthy pets (controls)

Question 31

What are some examples of information bias?

Answer 31

1. observer variation (this is minimized by using standardized protocols) 2. Deficiency of tools and technical errors 3. Recall bias (those affected by dz have a greater sensitivity for recalling exposure) 4. Reporting bias

Question 32

The confounding factor influences the exposure, the outcome, or both? This results in distorting the measure of association

Answer 32

BOTH the exposure and the outcome the distortion can be large and lead to over estimation or under estimation of an association it can even change the apparent direction of an association

Question 33

What three criteria must a third factor meet to be a confounder?

Answer 33

1. associated with the exposure 2. associated with the outcome 3. Not in the casual pathway between exposure and outcome

Question 34

What three ways can you reduce confounding variables in a study?

Answer 34

Before the study starts: if you predict that there is a confounder: 1. Match the study - select cases and controls so that the confounding factor is equally represented in both groups 2. Restriction - do not enroll any animals that have the confounding factor 3. Randomization - will reduce confounding

Question 35

What do you do if you realize there is a confounding variable after the study is completed?

Answer 35

Deal with the confounding factor in the analysis by stratifying Stratify: partition the results based on the confounding factor Ex: if the confounding factor is sex - split the data into males and females, then analyze both groups separately