Bias

How do we handle bias?

1) Identify the source (selection, information) 2) Estimate the magnitude (large bias vs small bias) 3) Assess the direction

How to minimize random error?

1) Increase precision of instrument 2) Take more measurements on more people 3) ALWAYS calculate power necessary to find the statistical and clinical difference you are looking for. This ensures that you have sufficient sample size to effectively answer your research question.

How do we prevent interviewer bias?

1) Masking the interviewers to the subject's disease or exposure status but not easy 2) Closed ended questions and better questionnaire design 3) Training interviewers not to ask leading or probing questions

How does random error influence studies?

1) Statistical fluctuations in measurement (noise) 2) Can go in either direction (overestimate or underestimate the true measurement) 3) Increases standard deviations around point estimates 4) Decreases precision (e.g., our ability to accurately estimate measures of association)

What are ways to control information bias?

1) Using two different controls (normal and diseased to see if they are different) 2) Avoiding open ended questions vs specific drug names 3) Self report for sensitive data 4) Biomarkers 5) Medical records 6) Knowledge of the hypothesis

What are examples of selection bias in case-control studies?

1) control selection bias 2) self-selection bias 3) differential surveillance 4) diagnosis/referral bias

When can bias be introduced to a study?

1) design (selection of controls) 2) data collection (interviewer) 3) analyses (statistical methods) 4) publication (reviewer)

What are three types of information bias?

1) interviewer bias 2) recall bias 3) differential misclassification

What is interviewer bias?

A systematic difference in collecting data by interview. Occurs in every study when cases or exposed are questioned differently

What is information bias?

A systematic difference in measuring exposure, outcome or covariates

What is bias?

A systematic error (not random) in the design or conduct of the study that leads to erroneous associations between the exposure and disease.

What affects the validity of a study?

Bias, confounding, and random error

What types of studies does selection bias affect?

Case control, retrospective cohort, as well as cohort (differential loss to follow up)

Occurs in which types of studies?

Case-control and cohort

What is recall bias?

Differential level of accuracy in information by compared groups. Case control = recall of exposure; in cohort =reporting of disease different among exposed vs unexposed

What is misclassification (measurement error)?

Exposed is classified as unexposed and vice versa or the diseased is classified as non-diseased or vice versa.

What is differential misclassification?

Exposure misclassification is different for the diseased compared to the non-diseased (away or towards the null)

What is external validity?

Generalizability

What is confounding?

Mixing of effects between the exposure, disease and a third factor (called a confounder) . This distorts the relationship between and exposure and disease

What is non-differential misclassification?

Nondifferential: misclassification occurs equally for diseased and non-diseased (always towards the null)

What is selection bias?

Produce results among participants that are different from the results that would occur among eligible individuals in the source population .

How does selection bias occur in case-controls?

Refusal, non-response, or agreement based on the exposure and disease

How is selection bias introduced?

Results from how subjects are selected and from factors that can influence participation

What study types are more vulnerable to bias?

Retrospective studies (case-control or retrospective cohort)

What are selection biases in cohort studies?

Retrospective: choice of exposed individuals related to outcome Prospective/ experimental: loss to follow up related to both the exposure and outcome Healthy worker effect: only healthy workers stay to be selected

What are the types of bias?

Selection bias and information bias

What is internal validity?

That the study conducted produced valid results