Vous êtes sur la page 1sur 2

Bias & Validity

Validity is the degree to which a measurement reaches a correct


conclusion
Internal validity: The extent to which the results of an investigation
accurately reflect the true relationship in a study population
External validity: The extent to which results of a study can apply to
other more general populations

Particularly important to physicians who need to appraise research


and determine whether its applicable to their patient

Bias is a systematic error in a study that leads to distortion of results. It is


a threat to validity. Bias can have direction, but if we know the direction of
the bias, we can make a clinical decision.
Selection bias: A selection process that may change the relationship
between exposure and and the disease of interest

Particularly important in case-control studies

Information bias: Random or systematic inaccuracies in


measurement. Also known as exposure misclassification

Non-differential = when all classes, groups, or categories of a


variable (whether exposure or outcome) have the same error rate or
probability of being misclassified.

Differential = when only particular groups are more likely to be


misclassified. Errors in information about one variable are affected
by the status of the other variable

Recall = Different abilities to recall previous activities &


exposures (i.e. someone who is sick might remember smoking
more than someone who is healthy)

Interviewer = Using leading or loaded questions

You can use biological markers to reduce misclassifications

Confounding bias: extraneous variables impact the relationship


between the exposure and the disease of interest. It can be reduced by
randomization, and also by stratifying the confounder into people who
have the confounder (ie. Obesity) and people who dont have the
confounder (ie. not obese). There is also often loss to follow-up.
CRITICAL APPRAISAL

Can I believe the results?

What are the results?

Can I apply the results to my patient?

You can have confounding, misclassification of exposures, and


misclassification of outcomes across the course of a study
Outcome Misclassification

Non-differential = giving false positives or negatives regardless of


which group the patient is in (i.e. 20% of false positives for both
untreated and treated subjects)

Differential = giving false positives or negatives of outcome status


more in one group than in the other (ie. 20% of false positives only
in the treated subjects)

Loss to Follow-Up Bias

If there are systematic differences in the amount of retention


between groups. For example, one group has lost 50 participants,
and the other group has only lost one.

Vous aimerez peut-être aussi