click below
click below
Normal Size Small Size show me how
PHC 6013: Bias types
Types of bias
Term | Definition | Type of bias | Effect on relative odds | Notes |
---|---|---|---|---|
Detection signal bias | An innocent exposure may become suspect if, rather than causing a disease, it causes a sign or symptom which precipitates a search for the disease, e.g. the current controversy over post-menopausal estrogens and cancer of the endometrium. | Selection | Ca/co: Up Preventable but if matching is used, overmatching may be a concern. Is measurable. Cohort: Up | AKA Unmasking bias |
Non-respondent bias | Non-respondents (or late-comers) may have exposures or outcomes which differ from those of respondents (or early comers). | Selection | Ca/co: Up or down Both preventable and measurable. Cohort: Up or down Both preventable and measurable. | Direct opposite of volunteer bias |
Membership bias | The degree of health within a group differs systematically from that of the general population. | Selection | Ca/co: Up or down Not preventable and only partially measurable. Cohort: Up or down Not preventable and only partially measurable. | A type of this is Health Worker Effect |
Diagnostic suspicion bias | Due to knowledge of subject's prior exposure, becomes suspicious of a particular diagnosis/outcome. Can lead to more intense surveillance/screening for the outcome among exposed than among unexposed. | Selection | Ca/co: Unlikely to occur but up Cohort: Up Both preventable and measurable. | AKA Case ascertainment or surveillance bias |
Exposure suspicion bias | Similar to diagnostic suspicion bias. Due to knowledge of subject's disease status, becomes suspicious of a particular exposure. Leads to more intense surveillance/screening for exposure among diseased. Asks questions differently about exposure. | Information | Ca/co: Up Both preventable and measurable. Cohort: Very unlikely to occur but up | AKA Diagnostic or interviewer bias |
Family information bias | Similar to exposure suspicion bias, but errors originate with the subject or his/her family. The flow of family information about exposures and illnesses is differential depending on disease status. Those diseased more likely to search/acquire the info. | Information | Ca/co: Up Both preventable and measurable. Cohort: Does not occur | Similar to recall bias |
Recall bias | Cases and controls recall exposures differently. May be most marked when the exposure of interest is rare or when controls are drawn from the community rather than from hospitalized patients. | Information | Ca/co: Up Both preventable and measurable. Cohort: Does not occur | AKA Memory bias |
Attention bias | Study subjects systematically alter their behavior when they know they are being observed. | Information | AKA Hawthorne Effect | |
Observer bias | A distortion, conscious or unconscious, in the perception or reporting of the measurement by the observer. It may represent systematic errors in the way an instrument is operated, or in the way an interview is carried out. | Information | ||
Instrument bias | Results from faulty function of a mechanical instrument (ex. machine that has not been calibrated). | |||
Subject bias | Distortion of the measurement by the study subject, for example, in reporting an event. | Contributes to respondent and recall bias. | ||
Selection bias | Distortion in measure of association caused by using improper procedures for obtaining persons from the target population to become members of the study population or as a result of factors that influence continued participation of subjects in a study. | Ca/co: Up or down Cohort: Up or down Can also produce an effect when none actually exists | ||
Information bias | Distortion in measure of association caused by a lack of accurate measurements of exposure or disease status. | AKA Measurement bias? | ||
Neyman bias | Disease is measured too late because of a time gap between exposure and selection of study subjects. Because of this, cases that were fatal or were only short episodes are missed. | Selection | Ca/co: Up or down Not preventable and only partially measurable. Cohort: Unlikely to occur but up or down Both preventable and measurable. | AKA Prevalence-incidence bias |
Berkson bias | In hospital-based studies, admission rates of exposed and unexposed cases and controls differ, distorting the odds ratio. Occurs due to differential burden of symptoms, access to care, popularity of disorders and institutions, etc. | Selection | Ca/co: Up or down Preventable but not without sacrificing its value. Not measurable. Cohort: Does not occur | AKA Admission rate bias |