Causal inferences: bias, confounding and interaction

Question 1

Bias

Answer 1

occurs as a result of an error in the design, organization or analysis of a study and may result in an incorrect assessment of an exposure’s influence on the risk of developing a specific disease

Question 2

Two common types of bias

Answer 2

information bias

selection bias

Question 3

Selection bias

Answer 3

if cases and controls that were selected showed some noticeable link between exposure and disease but there was no association this may have occurred as a result of selection bias

Question 4

Selection bias id divide in two

Answer 4

Non-respondent bias: when participants that respond to a survey respond in a way that is diff from the people who do not respond
Exclusion bias: when investigator uses dissimilar eligibility criteria to select the cases and controls

Question 5

Information bias

Answer 5

can arise when the way info is gathered about participants in a study is not complete or accurate enough. This may result in a portion of the info related to exposures or disease outcome being wrong

Question 6

Misclassification bias

Answer 6

participants are misclassified for instance cases are misclassified as controls or vice versa. Exposure status may also be inaccurate if person is not aware that they have been exposed or feels that they were exposed when they were not

Question 7

Two types of misclassification bias

Answer 7

differential: rate of misclassification varies in separate groups
non-differential: problems with accuracy of data collection from both separate study groups, cases and controls

Question 8

Some forms and sources of info bias

Answer 8

bias in abstracting records
interviewing
from surrogate interviews
surveillance bias
recall bias
reporting bias

Question 9

Confounding

Answer 9

true association is observed but in actual fact does not exist, caused by a third factor
e.g. we find a link between factor A and a disease we may assume that the factor A causes the disease. But factor X which is associated with factor A ( but is not a result of factor A) may also be a risk factor for the disease. Factor X may then be the real factor that lead to the development of disease and not factor A but because factor A was studied, the results seem to indicate that it is associated with the development of the disease

Question 10

What is a confounder

Answer 10

a 3rd factor that must be considered when deciding whether an association is causal

Question 11

Problems with confounding

Answer 11

designing and carrying out a study where the cases are matched to the controls for the factor suspected to be a confounder
analysis of the data by stratification or adjustment

Question 12

Interaction

Answer 12

when two or more risk factors modify the incidence of disease and the joint effect of the two causal factors differ from what would be expected when adding their effects when they are acting independently

Question 13

Synergism

Answer 13

positive interaction

occurs when the combined effect of the two factors is greater than what we would expect

Question 14

Antagonism

Answer 14

negative interaction

combined effect of the two factors is less that what we anticipate

Question 15

Steps in looking for an interaction

Answer 15

establish if there is confounding
if not, then there is an interaction
If association is equally strong in diff strata that were determined according to a third variable then there is no interaction
If there is an interaction then calculate incidence after exposure using addition model and multiplicative model
Choice of which model to use depends on biology of disease