Bias And Misclassifications Flashcards
What do researchers evaluate for internal validity?
Confounding and effect mod.
Bias
Statical significance
What is selection bias
Selection process used to pick patients.
Exchangeability or comparability; not similar variables b/w groups
Only testable variable should be different
Ex: healthy-worker bias (include alll workers, past or present dead or alive)
Ex: self selection/participant: volunteers have different beliefs or traits than other part of the population
t.
What is differential misclassification bias?
Misclassification of exposure or disease is related to the other depending on the study design
Bias can move measure of assoc. in either direction it can inflate or attenuate your estimates of associations
Affects one group unequally
What is observer related bias ?
A type of measurement bias
Bias due to observer
What is contamination bias?
A type of subject related bias
Patients of control group or treated group receive outside treatment/change
Ex: prescribed drugs by unrelated DR.
What is non differential misclassification?
Both groups experience error EQUALLY
Misclassification of exposure or diseases is UNRELATED to the other.
It’s still an error but making claims because its pushing it towards neutrality.
It causes your value to show 1.0
What is screen-related bias?
“Lead time bias”
A type of measurement bias.
An apparent benefit from a healthcare screening due
to the early detection of disease despite an
unchanged clinical outcome
‘Extra time’ (knowing patient’s have the disease) affordedby early detection (screening) is designated as the LeadTime
What is compliance (Adherence) bias?
Groups may not stay on the studies protocol (taking pills every day is one example )
What is Hawthorne effect?
A type of subject related bias.
Ppl modifies their behaviors when they know they are be watched
What is a bias?
A systematic (non random) error in study design or conduct to lead to erroneous results
DISTORTS RELATIONSHIP B/W ASSOC.
What is interviewer bias?
A type of observer bias.
Observer will do things different because they know who’s in what group Interventions not implied equally
Different interviewers, expertise are different
What are tips for controlling boas?
Select most precise, accurate and medically appreciate measures of assessment and evaluation and observation
Get data as often as possible
High sensitivity and validated screening tools
To reduce the 2 main categories of bias what’s the most important thing you should do???
Remember,
EXCHANGEABILITY = COMPARABILITY
DONT DO ANYTHING DIFFERENT OR CREATES DIFFERENCE B/W GROUPS
SAME, SAME, SAME
What are examples of controlling bias?
Binding/ masking
Randomly allocate observers/interviewers for data collection (train them; use tech)
Build in as many methods necessary to minimize lost to follow up
When assessing biases and their potential impact what elements relate to bias impact?
Source/type
Magnitude/strength (weak or strong asoc.)
Direction (over or under estimate of true measure of assoc.)
