Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
35 Cards in this Set
- Front
- Back
|
What is Bias ? |
•Any systematic error that results in an incorrect estimate of the association between the exposure and the disease. |
|
|
When can bias occur? |
–design, –conduct, –analysis -- Interpretation --Publication |
|
|
What is selection bias? (participation bias)? |
Case control study = selection bias - important problem oin these Cohort study = participation bias "who get into the study" |
|
|
What is a common element of selection bias |
–The relationship (OR,RR) between the exposure and the disease (must include both) is different in the study participants that it is in the eligible population. |
|
|
What are 2 types of selection bias? |
Hospital controls Hospital cases |
|
|
Explain the effects of selection bias using 2x2 tables |
Over sample - cell d - OR goes up Over sample cell B OR decreases artificially |
|
|
What is detection bias? Example? |
–The exposure results in closer medical surveillance, and the disease is likely to be detected. Example: –Exogenous estrogens and uterine Cancer; women typically had more doctors visits than those not on estrogen therapy. More doctor visit more likely todetect cancer |
|
|
WHat is diagnostic bias Example? |
–The clinician may be aware of exposure history and is more likely to diagnose a specific condition - –Fetal alcohol syndrome; written on birth certificate not justfrom signs/symptoms but bcdoctor knows mothers exposure |
|
|
What are some selection bias red flags? |
–Low participation rates (<70%) – bias could be happening •Mathematically happening in those that aren’t participating à distorts results –Different participation rates between cases/controls, exposed/unexposed •Differential measurement on one variable |
|
|
WHat are some preventions for BIAs? |
–Think carefully about selection pressures for cases/controls, exposed/unexposed during the design phase. –Case-controls use multiple sets of controls –Maximize participation– recruit ppl motivated to be a part of the study –Obtain information on refusals (age, sex, race, smoking, SES) |
|
|
WHat is information bias? |
–A systematic difference in the collection of data (exposure and disease status) among the study subjects |
|
|
WHat are common elements of Information Bias? Cohort study and case - control the _____ of ____ is affected by ___ or ___ ______ |
–Cohortstudy: The reporting of disease is affected byperceived or actual exposure, or
–Case-controlstudy: thereporting of exposure is affected by perceived or actual disease. |
|
|
WHat MUST information bias include? case-control and cohort ____ reporting of _____ in ______ vs. ______ |
***Mustinclude both variables ***Overreporting of disaese inexposed vs. unexposed Overreporting of exposure in cases vs. controls |
|
|
What are 2 case-control specific information bias? |
1) recall bias 2) Interview bias |
|
|
Explain interviewer bias Examples? |
–More probing, detailed, or persistent questioning of cases/controls or across interviewers •Example: Woburn, MA, childhood leukemia –Had toxic waste site contaminated well water –Harvard included activist individuals – interviewers knew women were personally effected to interviewee tried to say what theywanted to hear bc ofwho the interview was |
|
|
Explain recall bias When the ____ or ____ of _____ reporting ____ between _____ and _____
Why doesn't this occur in cohort studies? **Who knows what in case control? Vs cohort _____ exposure _____ of outcome. |
–Whenthe accuracy or completeness of exposure reporting differs between cases andcontrols
–Incase control study everything you want to measure has already happened. Casesknow they have disease, controls know they don’t– effects what they report–VS:Cohort – measuring exposure with knowledge independent of the outcome |
|
|
Example of recall bis Prenatal birth defects |
–prenatalexposures and birth defects, childhood cancers
• Youask the parents/ mother – asking mother about acutely ill child about exposures– she prob hasvery good recall of all her exposures bcshe has already asked herself what happen
•Controlmother don’thave same motivation to recall these things – instead chose mother or childwith other diseases that might create same level of exposure recall out of thecontrols |
|
|
What are 2 cohort specific information bias? |
•Observation/ReportingBias •Loss to follow-up |
|
|
Explain Observation/Reporting Bias Example? |
–Knowledge of exposure status influences the determination of the outcome –GulfWarSyndrome– cohort: troops exposed in desert storm – so much reporting of what the symptoms were and were not they were reporting symptoms etc. bc they were paying attention to what they should be seeing |
|
|
Explain Loss to follow-up bias |
–When persons lost to follow-up differ from remaining subjects with respect to both exposure and disease. •Theymight be the ones with outcome of interest – cant measure and show connection –underestimates RR |
|
|
What are some INfo bias alarms? Includes... |
•Anytime the data is collected differently in cases vs controls, or exposed vs unexposed.– - Measured exposure/outcome differently Includes: –Methods– Interviews vs mailed questionnaires–Completeness– Accuracy, Follow-up•10%dropout (DO) in exposed, 20% DO in unexposed = problem |
|
|
What are some ways to prevent information bias? |
•Establish a clear definition of exposure and disease before the study begins •Use the same data collection methods for all subjects, following a clear and rigid protocol •Mask interviewer and participants to the study hypothesis. If possible,mask interviewers to disease or exposure status of subject •Closed–ended questions are better than open-ended questions •Find another source of information to validate disease/exposure classification –Check self-reports with med. records, employment data, family source –Includes:multiple questions in different formats/wording |
|
|
What are some ways to prevent loss to follow up? |
•Recruit motivated participants, and stay in contact •Identify someone who will know where they are (advantage of registries) –ask for contact info for close friend and family member. Registries – external system and can still get employment status •Collect identifying information so you can find them: social security number,credit bureaus, drivers license • •Finally,compare baseline characteristics of lost vs retained to determine if/how they differ.Differences in common characteristics of those dropping out? Not different =reassuring |
|
|
WHat is misclassification Error? ** |
•A type of information bias Exposure ---> Disease –Some misclassification of study subjects with respect to disease and exposure status is virtually inevitable. –Will underestimate or over estimate exposure or incorrectly diagnose disease
|
|
|
WHat are two types of misclassification Error? |
–Non-differential misclassification - Random error –Differential misclassification |
|
|
Explain Non-differential misclassification (random error) |
•Theprobability of misclassification is the same for all of the groups beingcompared. |
|
|
Explain Non-differential misclassification (random error) in a case-control study |
–Thelikelihoodof misclassification of exposure status is the same for diseased andnon-diseased |
|
|
Explain Non-differential misclassification (random error) in a cohort study |
The likelihood of misclassification ofdisease status is the same for the exposed and non-exposed |
|
|
What are the results of non-differential misclassification |
•The measure of association is biased toward the null •(Whatever the “true” odds ratio, the observed odds ratio will be closer to 1.0) •Protective– misclass will go towards .78, .89 will never cross 1.0• OR will go down from 4.5 toward 1 but wont cross that threshold 1.0 **** The observed association is a conservative estimate of the “true”association. |
|
|
Explain differential misclassification |
•The probability of misclassification differs across groups. |
|
|
Explain differential misclassification in a case-control study |
–The likelihood of misclassification of exposure depends on disease status. |
|
|
Explain differential misclassification in a cohort study |
–The likelihood of misclassification of disease depends on exposure status. |
|
|
What are the results of differential misclassification |
• Unpredictable–Canmake positive association go higher lower or cross 1.0 |
|
|
2 types of selection bias Hospital controls - advantages/disadvantages |
•Advantages– easy to find, better compliance/enthusiasm, easy to track, cheap–disAdvantages Their exposure status may be linked to their hospitalization, and/or may be very different from the exposure status of all eligible controls. |
|
|
2 types of selection bias Hospital cases advantages/disadvantages |
–May represent the more severe cases of a disease(often referrals)–Not representative of all cases–Problem: If exposure affects severity but does not cause the disease. |
