unit 7 pt 2 - Observation Bias Bias Evaluating Valid...

Info iconThis preview shows pages 1–3. Sign up to view the full content.

View Full Document Right Arrow Icon
Evaluating Valid Statistical Associations Statistical Associations: Bias – Part 2 Part 2 Dr Borenstein Dr. Borenstein NO AUDIO ON THIS SLIDE. CLICK NEXT TO CONTINUE >> Observation Bias ± Information ² Reca Recall ² Interviewer Data from cases and controls are not comparable Recall Bias ± Subject recalls exposures erroneously ² In case-control study ± Cases may over- or under-report (over-report socially-desirable behaviors (e.g., reading to baby in utero) or under-report undesirable behaviors, (e.g., smoking, drug exposures) ± Cases may have done research (web, reading) about their disease and be more knowledgeable than controls (unless controls have a different disease of similar severity/quality of life impairment) ± Controls may over-or ±under-report (same as above, but controls may also under-report because they have not thought in as much detail about causes for their own/child’s/other’s disease) their own/child s/other s disease) Example 1: Recall Bias Alzheimer’s disease Community controls (No Alzheimer’s disease) Example 1: Recall Bias Family history of Alzheimer’s disease (No Alzheimer s disease) OR = ad No family history of Alzheimer’s disease OR bc ± AD case (informants/proxies) recall subject’s family history better than controls (control proxies have no reason to recall such family history) ± Note: this affects cells a and d selectively, causing OR to be extremely over-estimated (biased away from the null) – Type I error ± Solution: none or select controls with another disease Example 2: Recall Bias - Occupational Exposures to Anesthetic Gases and Risk of Spontaneous Abortion ± Women with miscarriages may recall Women with miscarriages may recall exposures better than women giving birth to healthy babies to healthy babies. Example 3: Recall Bias - Lung Cancer and Cigarette Consumption ± Cases may exaggerate their cigarette consumption (they know that tobacco consumption is related to their disease) ± Controls have little reason to recall or exaggerate their tobacco consumption – (can’t choose diseased controls with another tobacco-related disease because p will be incorrect disease because p o will be incorrect) ± Observed OR will be over-estimated (biased away from the null: Type I error away from the null: Type I error)
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Recall Bias unlikely to be a problem in a Cohort Study… ± Why? ² Exposures are ascertained before knowledge of who is going to get the disease of interest Interviewer Bias ± Interviewer elicits or interprets information Interviewer elicits or interprets information differently between cases/controls or exposed/unexposed ² Interviewers know who is a case and who a control ± May probe in a biased way, eliciting more information from cases ± Trytomaskcase-control/exposed-unexposed Try to mask case control/exposed unexposed status from interviewer; very difficult ± Try to mask principal hypotheses from interviewer/subjec interviewer/subject Example of Interviewer Bias Alzheimer’s Community controls Example of Interviewer Bias
Background image of page 2
Image of page 3
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 07/15/2011 for the course PHC 6000 taught by Professor Staff during the Summer '08 term at University of South Florida.

Page1 / 8

unit 7 pt 2 - Observation Bias Bias Evaluating Valid...

This preview shows document pages 1 - 3. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online