Lec 11. Bias and Misclassification Flashcards
Researches evalutate 3 aspects of their study (internal validity) before declaring a real true association, what are they?
Check for Counfounders or effect modifiers, check for bias and check for statistical significance.
What is bias?
Systematic (NOT RANDOM) error in study design or conduct leading to erroneous results.
What does bias do to the relationship between exposure and outcome?
Distorts it
Can bias be fixed once it already occurred (after study ends)?
No
How can you minimize bias and its impact?
Prospective (pre-study) consideration and adjustment
What are three elements of bias?
Source/type, Magnitude/Strength and Direction
What is Magnitude/Strength pertaining to one of the elements of bias?
Can account entirely for a weak association (RR/OR) but not likely for a very strong one
What is Direction pertaining to one of the elements of bias?
Can over or under estimate true association. Towards or away from the ratio of 1.0
What are the two main categories of Bias?
Selection-related and Measurement-related
What is selection-related?
How we go and pick people, this will screw up soley by pts you pick.
What is a way to avoid selection-related bias?
Dont do anything that is different between groups
What is measurement related bias?
Anything we are measuring into observation, all data collection. Errors in here
What is a way to avoid measurement-related bias?
Dont do anything that is different between groups
What is selection bias?
The way study subjects are selected in different ways causing differences between groups
What are examples of selection bias?
Healthy-worker bias, self-selection/participant(responder) bias
What is Healthy-worker bias?
Only ppl included in study are those who are healthy enough to be there. The sick and dead are left out. w/ environmental studies
What is self-selection/participant(responder) bias?
Those who wish to volunteer maybe different than those who didnt volunteer
Pertaining to measurment bias via subject-related what are some examples?
Recall bias, Hawthorne effect, Contamination bias, compliance/adherence bias, Lost to follow up bias (Differential attrition)
What is recall bias?
Pts recall things differently,memories etc
What is Hawthorne effect?
Pts act differently because they know they are being watched
What is contamination bias?
Control group receive treatment not part of study, altering results
What is compliance/adherence bias?
Pts dont comply with medication/treatment
What is lost follow up (differential attrition) bias?
Pts leave due to side effect etc
Pertaining to measurement bias via observer related bias what are some examples?
Interviewer bias, diagnosis/surveillance bias, lead-time bias, publication bias
What is interviewer bias?
Anything the interviewer does that is different from the 2 groups
What is Diagnosis/surveillance bias?
Different interviewers have different skills/view points. Screen related.
What is lead-time bias?
Pts screened and are becoming aware of disease longer than those unaware.
What is publication bias?
Not everything gets published. Only want positive or desired outcome for researchers or authors.
What is misclassification bias?
Error in classifying either disease or exposure status, or both. Source of measurement bias.
What are two primary forms of missclassifcation?
Non differential and differential
What is non differential and differential?
Non differential: mistakes happen equally in both groups (OR/RR) move to 1. Accept errors. Differential: Mistakes happen in one group unequally. RR/OR either direction
What are two ways to control for bias?
Blinding/masking, and Randomly allocate observers/interviewers for data collection
What is blinding/masking?
We use multiple sources to gather info
what is Randomly allocate observers/interviewers for data collection?
Building many methods necessary to minimize loss to follow up.
