Epidemiology: Bias, Chance Flashcards
Define selection bias
Where the relation between exposure and disease is different for those who participate Vs those who could but dont.
Meaning the populations may not be representative
Define information bias. What are the two types?
Occurs when there are systematic errors in measurement of information for the subjects. And the consequences are different for the different study groups
Reporting bias
Observer bias
Define observer bias
Where the person doing the observations may have a tendency to classify one thing more greatly than another due to some preconceptions
Define subject bias
Where a participant may be more likely to classify things than their identical counterpart e.g recall bias
Define misclassification bias in the context of cases control studies
Where true cases are wrongly labelled as controls and vice versa
If random then will weaken observed effect
Define survivor bias in the context of case control studies
If exposure is rapidly fatal then patient will die before they are defined as a case HENCE will identify factors that increase survival among the diseased as risk factors
Define a cofounder
A variable that is independently associated with exposure and disease e.g a 3rd factor
E.g looking at use of bifocals vs bed wetting. There will be an association but this is false as the use of bifocals does not cause bedwetting. Rather it is as people who need them are usually elderly
How do you mitigate for confounding factors?
Stratify the population however this leads to reduced power and makes presentation and interpretation difficult
Can also use logistic regression
Name the ways that selection bias occurs in case control studies
Cases selected for study are not representative of all eligible cases
Controls not representative of population which produced the cases
Name the way selection bias can occur in cohort studies
Where the comparison groups (exposed and unexposed) are not truly comparable
- poor choice of unexposed group
- differences in follow up between groups
- missing data
Name the ways selection bias can occur in cross sectional studies
Participation rarely 100%
Participants may have different exposures
Available data on non responders should be examined
What is reporting bias
When subjects with a specific health outcome report previous exposures with a different degree of accuracy than those without the outcome
OR
Subjects give an answer they think the investigator wants to hear
How do you avoid observer bias in
A. Case control studies
B. Cohort studies
A. People responsible for classifying exposure status dont know the disease status of participant
B. People responsible for classifying outcome don’t know subjects exposure category
What are the two types of misclassification bias?
Non-differential (random)- where all exposure categories have equal chance to be in wrong category
Differential- where a certain exposure category is more likely to be categorised than another
What P value would you make you think that the results conclude a real effect?
P<0.05
