Lecture 7: Bias and confounding Flashcards
What is Bias in biostatistics?
Bias is any systematic error in design/implementation of a study
•Systematic error in selecting participants
•Systematic error in data collection
Wha is the effect of bias?
It creates the tendency to overestimate or underestimate a measure of association.
What are the types of bias?
Random error
Systematic error:
- information bias
- selection bias
What is Selection bias?
- Errors introduced when the study population does not represent the target population
- Systematic difference between those included in a study and those who are not
How do you reduce selection bias?
The most effective method to minimise selection bias in intervention studies is random allocation to treatment and control groups.
What is Volunteer or Non- response bias?
Typically occurs in surveys/observational studies, where individuals decide whether or not to participate.
What is the Healthy worker effect ?
Occurs in occupational cohort studies or in studies comparing working to non-working populations.
The effect:
- Workers have lower rate of disease than non-workers
- Lower mortality in the employed population compared to the general population.
- Comparison of workers to non-workers may underestimate the health effects of occupational exposures
What is Neyman’s bias .
Neyman Biasis a selection bias where the very sick or very well (or both) are erroneously excluded from a study.
What are the effects of Neyman bias?
- Excluding patients with severe disease will make treatment effective
- Excluding patients who have recovered will make conditions lookmoresevere.
It affects studies on long-term conditions e.g. HIV or tuberculosis. Less of a problem with acute conditions.
What is Attrition bias?
Loss to follow up
It can occur in Cohort studies and randomized controls trials (RCT’s) which involve following people over a certain period of time
Those that are lost to follow up may be different from those who complete the study
What is information (measurement) bias?
Information bias is a systematic error in exposure assessment or outcome assessment
Biased data (even with random sample)
What is Recall bias?
Error due to differences in accuracy or completeness of recall to memory of past events or experiences.
Respondents memory vary according to whether they have experienced the outcome
- Cases with disease are more likely to recall the possible cause of their illness than healthy people.
What is Interviewer bias?
Interviewer bias occurs when there are systematic differences in the way information is collected for the groups being studied
How can interviewers influence respondents’ answers?
Structure of questions – e.g. putting emphasis on questions, use of gestures.
The interview situation
What is Observer bias?
Occurs when observers are not blinded to exposure or disease status
Observers may have preconceived expectations of what they should find in an examination
Example: Radiologist aware of a patient’s smoking status may look more critically at abnormalities in X-ray.
What is the Hawthorne effect?
Alteration in behaviour when subjects know they are being studied.
How does Controlling for bias occur?
- Ensure sample selection is random
- Measure and control all variables related to health behaviour
- Set up strict guidelines for data collection:
Use standardised questionnaires, train interviewers and observers, preferably use more than one observer. - Institute Blinding/masking where appropriate:
Blinding prevents interviewers or observers (and some times subjects) from knowing case/control status, treatment allocation, or exposure status of study subjects. - Use an appropriate method for data collection
E.g. for socially sensitive questions, such as alcohol and drug use or sexual behaviours, use a self-administered questionnaire instead of an interviewer. - Build in methods to minimise loss to follow-up:
Recruit people who can easily be tracked
Use of incentives
Define confounding
Lack of comparability: when certain background factors differ between groups being compared.
A mixing of effects: A situation in which the apparent effect of an exposure on risk is explained by other factors, resulting in a distortion of the true relationship
What is lack of comparability in confounding?
when certain background factors differ between groups being compared.
What is a mixing of effects in confounding?
A situation in which the apparent effect of an exposure on risk is explained by other factors, resulting in a distortion of the true relationship
What are Confounders?
Factors or variables that cause confounding
What are the conditions to be a Confounder?
Associated with the exposure
Associated with the outcome
Not on the causal pathway between the outcome and exposure
How to deal with confounding in design and statistical analysis?
Randomisation
Matching
Restriction
Stratification
Multi-variate regression analysis
Bias refers to errors in the design or conduct of a study that can result in false conclusions
True or false?
True
Selection bias can occur when the study sample does not reflect the population of interest
True or false?
True
Information bias can occur when the method used in collecting study data could lead to inaccurate responses or measurements
True or false?
True
Confounding can be minimised at ?
Design stage- randomisation
Analysis stage- stratification, regression analysis
