Bias and confounding Flashcards
When may epidemiological studies go wrong?
Poor study design Random error Bias Confounding Interaction
How can random error in a study be reduced?
Increase sample size
Does increasing sample size reduce bias?
No
What are 3 the types of bias?
Selection bias
Information bias
Interpretive bias
When does selection bias occur?
When selected individuals are systemically different to those not selected for the study
Doesn’t accurately represent population of interest
Selection bias can be healthcare access bias. What is this?
Bias due to valuable, unusual or serious cases being more likely to be sampled
Selection bias can be spectrum bias. What is this?
Cases with certain signs are more likely to be detected
Selection bias can be volunteer bias. What is this?
Volunteers in study are different from those refusing
Selection bias can be non-response bias. What is this?
People that choose to respond (e.g. survey) are different from those that don’t respond
Selection bias can be loss-to-follow-up bias. What is this?
People lost to follow are different to those that remain in the study
Selection bias can be missing information in analysis. What does this mean?
Only individuals with relevant information are included in analysis
What is information bias?
Bias due to misclassification of diseased/non-diseased
What 2 factors influence information bias?
Specifity
Sensitivity
Information bias can be measurement bias. What does this mean?
Cases can be mild or severe - what is classed as diseased?
Recall/repsonder bias is a type of information bias. What is this?
People that have had a particular problem may be more inclined to deny previous exposures
What is interpretive bias?
Bias due to emphasis given on certain evidence during result evaluation
What is confirmation bias? (Type of interpretive bias)
Selectively reporting information that agrees with previous information
What is rescue bias?
Discounting uncomfortable data by finding faults with the study
What is confounding?
Association between a study factor and an outcome is distorted by a third variable
(3rd variable affects both study factor and outcome)
What is required for a variable to be considered as confounding?
Must affect both study factor and outcome
Must not be an intermediate step in a casual pathway
What is type I error? When might this occur?
Study detects statistically significant findings by chance rather than being real
If investigating multiple study factors
What is type two error?
A study is not sufficiently powered to detect real statistical differences
What is external validity?
How well the study results apply to the wider population
What is internal validity?
How well the true situation in the population is represented by the sample used