Exam 2: Bias & Misclassification Flashcards
What is the definition of Bias?
Systematic (non-random) error in study design or conduct leading to erroneous results.
How can researchers be assured they found a TRUE relationship between “exposure” and “outcome”?
They need to evaluate 3 aspects of their study (internal Validity)
1) Check for Bias
2) check for confounding or effect modification (interaction)
3) Check for statistical significance
What does bias do?
distorts the relationship (association) between exposure and outcome.
*design flaw.
After calculating various measures of association (RR/OR/HR), how can researchers assess internal validity (study design/methodology) for areas of possible error?
By check for bias, bias is one those possible sources of error
What can be done to “fix” a bias one it has already occurred?
Nothing can be done to fix it.
How can bias and it’s impact be minimized?
Prospective (pre-study) consideration and adjustment.
What 3 components do investigators evaluate when assessing for bias and its impact?
1) Source/Type
2) Magnitude/Strength
3) Direction
T/F bias can account entirely for a weak association (small RR/OR/HR) but is not likely to account entirely for a very strong association (large RR/OR/HR)
TRUE
How can bias affect the direction of a measure of the true association?
They can over or under-estimate the true measure.
In relation to the null hypothesis does a bias have any effect?
yes, a bias can have an enhancing or minimizing effect on the true measure of association (towards or away from the null hypothesis)
Define information/observation/measurement -related biases:
any aspect in the way the researcher collects information, or measures/observes subjects (and their variable) which creates a systematic difference between groups in the quality/accuracy of their information.
*errors in measurement or classification (misclassification)
What are selection-related biases?
Any aspect in the way the researcher selects subjects (cases or controls, exposed or unexposed, study sample) which creates a systematic difference in the composition between groups.
What is a commonly seen selection-related bias?
When comparing groups not coming from same population/group or not being representative of the full population.
What is a selection bias?
Selecting study subjects that are not representative of your primary population of interest or that generates differences in groups being compared (very commonly encountered)
What are 3 key examples of selection bias?
1) Healthy-worker bias
2) self-selection/participant (responder) bias
3) control selection bias
Define self-selection/participation (responder) bias.
those that wish to participate (volunteer) may be different in some way to those that don’t volunteer or self-select (refusal/non-response) to participate
When can a control selection bias be easily seen?
In a Case-Control study.
What issue do you run into with information/observation/measurement bias?
Subject-related variations
Observer-related Variations.
What is a recall (reporting) Bias?
A subject-related bias.
A differential level of accuracy/detail in provided information between study groups. exposed or diseased may have greater sensitivity for recalling their history, or the severity may amplify their response.
What is the Hawthorne effect?
A subject-related variation in which individuals can report their “effects” of exposure, disease symptoms or Tx differently b/c they are part of a study.
What is a contamination Bias?
a Subject-related variation in which members of the control group accidentally, or outside of the study protocol, receive the Tx (or similarly) or are exposed to the intervention being studied.
What is compliance Bias?
a subject-related variation in which the groups being interventionally studied have different compliances. ( just don’t take rx b/c of side effect)
What is a Lost to Follow-up bias?
A subject-related variation in which the groups being studied have different withdrawal or lost to follow-up rates OR there are other differences between those that stay in the study and those that withdraw or are lost to follow-up
** differential vs. non-differential.
Can a lost to follow-up occur in a cohort study?
Yes, in a prospective cohort. While waiting for the disease to occur they can be lost to follow-up
What is interviewer (proficiency) bias?
an observer-related variation.
-A systematic difference in soliciting, recording, or interpreting on the part of the researcher (or their assistants)
What type of bias would have an issue with the interviewers knowledge influence the response from the study subjects, either conscious or unconscious?
Interviewer (proficiency) bias.
How can you assure that you will get the same info with multiple people involved when interviewing pts?
Take the interviewer out of the equation. I.e Have a recording, or by not providing additional info if pts have questions.
What is a diagnosis/surveillance (expectation) bias?
Different eval/classification/Dx/observation between study groups.
Observers may have preconceived expectation of what they should find in exam/eval/follow-up
What is it called when the researcher is trying to make the study have the results they believe should be there?
Hawthorne-like effect
What is kappa statistic?
Agreement between evaluators
T/F blinding didn’t work if they pt doesn’t know what drug they were on.
False. Blinding did work!
How can you control for biases?
Select most precise, accurate, & medically-appropriate measures of assessment and eval/observation
What are ways you can have the highest sensitivity/specificity and valid screening tools?
Use published/past-utilized techniques, if possible.
Calibrate & test equipment, techniques and processes.
List 4 options for controlling biases:
1) blinding/masking
2) use multiple sources to gather all info
3) randomly allocate observer/interviewers for data collection (train them)
4) build in as many methods necessary to minimize loss to follow-up
Define misclassification:
error in classifying either disease or exposure status, or both
Define non-differential misclassification
misclassification of exposure or disease which is unrelated to the other.
Effect= for dichotomous (2) variable, bias can move the measure of association (RR/OR/HR) towards the null hypothesis.
define differential misclassification:
error in one group differently than other.
misclassification of exposure or disease is related to the other
effect= bias can move the measure of association in either direct in relation to the null hypothesis. can inflate or attenuate your effect estimates of association.
