Module 5

Question 1

Ecological fallacy

Answer 1

exposure-disease associations at the group level do not hold at the individual level

Question 2

Regression to the mean

Answer 2

extreme measure on a particular variable is more likely to be closer to its mean on a later measurement

Question 3

Bias

Answer 3

systematic error in the design or implementation of a study that results in an incorrect value for a measure of association
result = the study gets the wrong answer
epidemiologists are concerned with reducing bias to enhance internal validity
generalizability is less of a concern
biases vary in magnitude
impossible to know to what extent biases have actually affected the results of a particular study
even studies with a methodological flaw might be relatively unbiased

Question 4

Bias and study design

Answer 4

type of study design makes some designs more prone to bias than others
absence of randomization and retrospective nature of case-control and cohort studies makes these designs more susceptible to bias than RCTs
some study designs have specific types of bias
case-control = recall bias because subjects are often required to answer questions about exposures that happened long ago
cohort = selection bias because there could be different rates of loss to follow-up in the exposed and unexposed groups
RCT = contamination bias due to inadequate wash-out periods

Question 5

Selection bias

Answer 5

selection bias = are participants in the study similar in all important aspects except for exposure/disease
selection bias arises through different events = volunteer bias, non-response bias, exclusion bias, healthy worker effect

Question 6

Volunteer bias

Answer 6

volunteers who participate in a research project are different in some ways from the general population
sampled only a subset of the population
data gathered are not representative of all people
volunteer bias is a challenge to the external validity of any research project

Question 7

Non-response bias

Answer 7

individuals chosen are unwilling or unable to participate in the survey
people who we invite to participate and ignore our invitation are different from those who respond yes or no

Question 8

Exclusion bias

Answer 8

exclusion criteria for a study creates a group of people who are not representative of the population that give rise to the cases
arises in case-control studies if patients with a particularly high (or low) rate of prior exposure to the alleged etiologic agent are excluded from the selection of either cases or controls, but not from both

Question 9

Healthy worker effect

Answer 9

in occupational epidemiology, those who attend work are generally healthier than those who cannot work

Question 10

Information bias

Answer 10

information bias/observation bias/measurement bias = is information about the outcome or exposure obtained in the same manner
refers to measurements performed on participants
differential or nondifferential
information bias arises through different events = recall bias, interviewer bias, loss-to-follow up, misclassification

Question 11

Nondifferential

Answer 11

measurements of disease is not different for exposed and unexposed subjects
measurements of exposure is not different for diseased and non-diseased subjects

Question 12

Differential

Answer 12

measurements of disease is different for exposed and unexposed subjects
measurements of exposure is different for diseased and non-diseased subjects

Question 13

Recall bias

Answer 13

memory of exposure history is distorted by present health state
in a case-control study, cases recall and/or report previous exposure differently from controls
if better recall of cases is because they are searching for a cause of the disease, use controls with a disease with a similar potential for the search

Question 14

Interview bias

Answer 14

affects all study designs
systematic difference in soliciting, reporting, and interpreting information from participants
case-control = exposure histories
cohort = outcomes reported
solution
blind exposure status of participants in cohort studies when measuring outcome
blind case/control status when assessing exposure
mask hypotheses
use a standardized, closed-form questionnaire

Question 15

Loss to follow up

Answer 15

problem in prospective studies (RCTs, cohorts)
systematic difference in following and obtaining outcome information
problem if different mechanisms for follow-up exist for comparison groups
problem if reason for LTFU related to outcome
solution
use standardized follow-up procedures for all participants
minimize potential LTFU
backup contact information
different modalities to fill out questionnaires

Question 16

Reporting bias

Answer 16

reluctance to report an exposure for social or psychological reasons
social desirability bias = under report street drug use, smoking
wish bias = deny exposures to minimize self-blame

Question 17

Biases in RCTs

Answer 17

discussion of bias often focuses on observational studies, however bias also occurs in RCTs
selection bias = randomization
performance bias and detection bias = blinding
attrition bias = incomplete outcome data due to losses in follow-up

Question 18

Types of errors

Answer 18

random error reduces precision of the estimates (OR/RR)
systematic error reduces validity or accuracy of the estimates
accurate and precise = no error
precise, but not accurate = systematic error
accurate, but not precise = random error
neither accurate nor precise = systematic and random error

Question 19

Selection bias in case-control studies

Answer 19

exposure can influence the detection of cases
an inappropriate control group = not representative of the population that gave rise to cases
an appropriate control group, but participation is related to exposure status

Question 20

Selection bias in cohort studies

Answer 20

in a retrospective cohort study, all cases of disease have already occurred so participants usually know their exposure and outcome status
in a prospective cohort study, we enroll participants before we know their outcome
bias can occur if choice of enrolled exposed or unexposed participants relates to both exposure and outcome status