1 Department of Veterans Affairs Epidemiologic Research and Information Center at Durham, NC Michel Ibrahim, MD, PhD Director of Education Program Lorraine Alexander, DrPH Carl Shy, MD, DrPH Sherry...

1 Department of Veterans Affairs
Epidemiologic Research and Information Center at Durham, NC
Michel Ibrahim, MD, PhD
Director of Education
Program
Lorraine Alexander, DrPH
Carl Shy, MD, DrPH
Sherry Farr, Graduate
Research Assistant
Sandra Deming, MPH
Graduate Research Asst.
www.sph.unc.edu/courses/eric
January / February 2001 Issue 16
Ron Horner, PhD
ERIC Director
http://hsrd.durham.med.va
.gov/ERIC/
The ERIC Notebook is
funded by the Department
of Veterans Affairs (DVA),
Veterans Health
Administration (VHA),
Cooperative Studies
Program (CSP), to
promote the strategic
growth of the
epidemiologic capacity of
the DVA.
Information Bias
Bias occurs when an estimated association (RR,
OR, difference in means, etc) deviates from the
true measure of association. The source of error
can be random or systematic, but the
consequence of the bias is systematic error in the
RR or OR estimate. Bias may be introduced at
the design or analysis phase of a study.
Major types of bias
Selection bias
Information bias
Confounding bias
Information bias is a distortion in the measure of
association caused by a lack of accurate
measurements of exposure or disease status.
Information bias, also called Measurement Bias,
arises when study variables (exposure, disease or
confounders) are inaccurately measured or
classified. Bias in the RR or OR can be
produced even if measured errors are equal
between exposed and non-exposed, or between
diseased and non-diseased study participants.
Non-differential Misclassification
Non-differential misclassification occurs if there
is equal misclassification of exposure between
diseased and non-diseased subjects, or if there is
equal misclassification of disease between
exposed and non-exposed subjects. If exposure
or disease is dichotomous, then non-differential
misclassification causes a bias of the RR or OR
towards the null.
Non-differential misclassification of exposure
status
Non-differential misclassification of exposure
status in a case control study occurs when
exposure status is equally misclassified among
cases and controls. Non-differential
misclassification in a cohort study occurs when
exposure status is equally misclassified among
persons who develop and persons who do not
develop disease.
Non-differential misclassification of disease
status occurs in a case-control study when
disease status is equally misclassified among
exposed and non-exposed subjects. Nondifferential
misclassification of disease status
occurs in a cohort study, when a study subject
who becomes diseased is equally misclassified
among exposed and non-exposed cohorts.
Effect of non-differential misclassification of
exposure
Non-differential misclassification always biases
the RR or OR towards the null if exposure
classification is dichotomous, i.e., either exposed
or non-exposed. If exposure is classified into 3
or more categories, intermediate exposure groups
may be biased away from the null, but the
overall exposure-response trend will usually be
biased towards the null.
Effect of non-differential misclassification of
disease
In most cases, non-differential misclassification
of disease will produce bias toward the null, i.e.
the RR or OR will be biased towards 1.0. If
errors in detecting presence of disease are equal
between exposed and non-exposed subjects
(sensitivity is less than 100%), but no errors are
made in the classification of disease status
(specificity is 100%), the RR in a cohort study
will not be biased, but the risk difference will be
biased towards the null.
Effect of non-differential misclassification of
disease status
If no errors are made in detecting presence of
disease (100%sensitivity), but equal errors are
made among exposed and non-exposed in the
classification of disease status (specificity less
than 100%), both the RR and risk difference will
be biased towards the null.
Combined errors in both sensitivity and
specificity further increase the bias towards the
null, but specificity errors produce larger biases
overall.
2 Department of Veterans Affairs
Epidemiologic Research and Information Center at Durham, NC
Differential misclassification
Differential misclassification occurs when
misclassification of exposure is not equal between diseased
and non-diseased subjects, or when misclassification of
disease is not equal between exposed and non-exposed
subjects.
Differential misclassification causes a bias in the RR or
OR either towards or away from the null, depending on the
proportions of subjects misclassified.
Effect of differential misclassification of exposure or
disease
Differential misclassification of exposure or disease can
lead to bias the RR or OR in either towards or away from
the null. The direction of bias is towards the null if fewer
cases are considered to be exposed or if fewer exposed are
considered to be diseased. The direction of bias is away
from null if more cases are considered to be exposed or if
more exposed are considered to be diseased.
Effect of differential misclassification of exposure or
disease can lead to bias the RR or OR in either direction.
The direction of bias is towards null if fewer cases are
considered to be exposed or if fewer exposed subjects are
considered to be diseased. The direction of bias is away
from the null if more cases are considered to be exposed or
if more exposed subjects are considered to be diseased.
Interviewer bias
Interviewer bias is a form of information bias due to:
1) lack of equal probing for exposure history between
cases and controls (exposure suspicion bias); or
2) lack of equal measurement of disease status between
exposed and non-exposed (diagnostic suspicion bias)
Solutions:
1) blind data collectors regarding exposure or disease
status
2) develop well standardized data collection protocols
3) train interviewers to obtain data in a standardized
manner
4) seek same information about exposure from two
different sources, e.g. index subject and spouse in
case-control study
Recall or reporting bias
Recall or reporting bias is another form of information
bias, due to differences in accuracy of recall between cases
and non-cases or of reporting of disease between exposed
and non-exposed.
Cases may have greater incentive, due to their concern, to
recall past exposures. Exposed persons, in a cohort study,
may be concerned about their exposure and may overreport
or more accurately report the occurrence of
symptoms or disease.
Solutions:
1) add a case group unlikely to be related to exposure
2) add measures of symptoms or disease unlikely to be
related to exposure
Complications in predicting direction of
misclassification bias
Misclassification of confounders results in unpredictable
direction of bias. Non-differential misclassification of a
polychotomous exposure variable (3 or more categories)
may result in bias away from null, though this is less likely
than bias towards the null.
Non-differential misclassification of disease limited to a
loss of sensitivity of detecting disease, without any loss in
specificity, does not bias toward null, whereas a loss of
specificity always biases toward the null.
Conclusions
Some inaccuracies of measurement of exposure and
disease occur in all studies.
If a positive exposure-disease association is found and nondifferential
measurement errors are more likely than
differential ones, measurement error itself cannot account
for the positive finding since non-differential error nearly
always biases towards the null.
Strive to reduce errors in measurement:
· develop well standardized protocols
· train interviewers and technicians well
· perform pilot studies to identify problems with
questionnaires and measuring instruments
· Attempt to assess the direction of bias by
considering likelihood of non-differential or
differential misclassification.
3 Department of Veterans Affairs
Epidemiologic Research and Information Center at Durham, NC
Self-Evaluation
Q1: Indicate which of the following statements are
true or false:
a) Non-differential misclassification of either the
exposure or outcome will, in general, bias the
association between the two toward the null.
b) Differential misclassification of either the
exposure or outcome will, in general, bias the
association between the two away from the null.
Q2: You are doing a case-control study of the effect of
folic acid intake on neural tube birth defects, and are
concerned about recall bias affecting your estimate of
association. Which of the following potential control
groups would reduce the likelihood of recall bias?
a) Women with normal births.
b) Women who gave birth prematurely.
c) Women with newborns who had other types of
birth defects without a known association to folic
acid intake.
Answers:
Q1a) True, one can conclude that non-differential
misclassification will generally bias the estimate of the
association toward the null.
Q1b) False, one cannot determine how non-differential
misclassification will change the estimate. The estimate
of association may change in either direction. See the
example below:
Hypothetical true classification of exposure and disease
status: (Note: in many studies, you many not know the
unbiased distribution of exposure and disease)
diseased non-diseased
exposed 100 110
not-exposed 140 300
OR = (100 X 300) / (140 X 110) = 1.90
If 20 people who had the disease were misclassified as noexposed,
and 10 people who had the disease were
misclassified as being exposed, the resulting table would
be:
diseased non-diseased
exposed 110 110
not-exposed 130 300
OR = (110 X 300) / (130 X 110) = 2.31
Non-differential misclassification of exposure status in this
case caused a bias of the estimate away from the null.
If 20 people who had the exposure were misclassified as
diseased, and 10 people who had the exposure were
misclassified as being non-diseased, the resulting table
would be:
diseased non-diseased
exposed 90 120
not-exposed 140 300
OR = (90 X 300) / (140 X 120) = 1.60
Non-differential misclassification of disease status caused
a bias of the estimate toward the null in this case.
Q2: The correct answer is c. Women who have negative
pregnancy outcomes may be more likely to recall events or
exposures that occurred during their pregnancy due to
concern or guilt. Women who give birth to children with
birth defects unrelated to folic acid intake are likely to have
similar incentive or concern (due to a negative pregnancy
outcome) when recalling exposures during pregnancy as
mothers who had children with neural tube defects. Thus
recall bias is assumed to be similar between the two
groups, which should minimize the effect on the measure
of association.
Glossary:
Information bias: A distortion in the measure of
association caused by a lack of accurate measurements
of exposure or disease status which can results from
poor interviewing techniques or differing levels of
recall by participants.
Non-differential Misclassification: It is equal
misclassification of exposure between diseased and
non-diseased subjects, or equal misclassification of
disease between exposed and non-exposed subjects.
Differential Misclassification: is unequal
misclassification of exposure between diseased and
non-diseased subjects, or unequal misclassification of
disease between exposed and non-exposed subjects.
4 Department of Veterans Affairs
Epidemiologic Research and Information Center at Durham, NC
ERIC NOTEBOOK IS PRODUCED BY THE EDUCATIONAL ARM (MICHEL A. IBRAHIM, MD, PHD, DIRECTOR) OF THE
EPIDEMIOLOGIC RESEARCH AND INFORMATION CENTER AT DURHAM, NC (RON HORNER, PHD, DIRECTOR)
BETH ARMSTRONG, ERIC PROGRAM MANAGER
VA MEDICAL CENTER (152)
508 FULTON STREET
DURHAM, NC 27705
Upcoming Topics and
Special Notes:
ERIC Notebook can be viewed
online at
http://sph.unc.edu/courses/eric
or
http://hsrd.durham.med.va.gov/
ERIC/Education/Education.html.
Please let Beth Armstrong know
which topics are of special
interest to you so that we can
If you would like to receive ERIC Notebook please fill out the
form below:
Name: ________________________________________________
Degree(s):_________________________
Address:______________________________________________
City, State, Zip:________________________________________
Telephone Number:_____________________________________
Fax Number:___________________________________________
E-mail Address:________________________________________
Please fax to: 919-416-5836 – Attn: Beth Armstrong Or
Mail to : Beth Armstrong, ERIC Program Manager, VA Medical
Center (152), 508 Fulton Street, Durham, NC 27705