2. Lecture Learning Outcomes
By the end of this lecture, you should be able to:
Definition of bias
Types of bias
Understanding & Handling bias
Practice & examples of bias
4. Bias
(Definition)
Systematic deviation of results or inferences from truth.
Systematic error (as opposed to a random error) that skews
the observation to one side of the truth.
Processes leading to such deviation.
An error in the conception and design of a study—or in the
collection, analysis, interpretation, reporting, publication, or
review of data—leading to results or conclusions that are
systematically (as opposed to randomly) different from truth.
5. Concept of Bias
Thus, if we use a scale that is not calibrated to zero, the
weights we obtain using this scale will be biased.
Similarly, if a sample is biased (for example, more males in
the sample than the proportion of males in the population, or
selecting cases from a hospital and controls from the general
community in a case-control study), the results tend to be
biased.
Bias are often inevitable
Since it is often difficult to correct for biases once the data
have been collected, it is always advisable to avoid bias when
designing a study.
6. Bias As Deviation From the Truth
Ways in which deviation from the truth can occur include:
Systematic variation of measurements or estimates from the true values.
Variation of statistical estimates (means, rates, measures of
association, etc.) from their true values as a result of statistical artifacts or
flaws in study design, conduct, or analysis.
Deviation of inferences from truth as a result of conceptual or
methodological flaws in study conception or design, data collection, or the
analysis or interpretation of results.
A tendency of procedures to yield results or conclusions that depart from
the truth (in study design, data collection, analysis, interpretation, review,
or publication)
Prejudice leading to the conscious or unconscious selection of research
hypotheses or procedures that depart from the truth in a particular direction
or to one-sidedness in the interpretation of results.
8. Allocation Bias
Allocation bias:
An error in the estimate of an effect caused by failure to
implement valid procedures for random allocation of subjects
to intervention and control groups in a clinical trial or in
another type of randomized study (randomized field trials,
randomized community trials).
9. Ascertainment Bias
Ascertainment bias:
Systematic failure to represent equally all classes of cases or
persons supposed to be represented in sample
This bias may arise because of the nature of the sources from
which persons come (e.g., a specialized clinic);
From a diagnostic process influenced by culture, or
idiosyncrasy; or,
In genetic studies, from the statistical chance of selecting
from large or small families.
10. Attrition Bias
Attrition bias:
A type of selection bias due to systematic differences between
the study groups in the quantitative and qualitative
characteristics of the processes of loss of their members
during study conduct; i.e., due to attrition among subjects in
the study.
Different rates of losses to follow-up in the exposure groups
may change the characteristics of these groups irrespective of
the studied exposure or intervention, or losses may be
influenced by the positive or adverse effects of the exposures.
11. Auxiliary Hypothesis Bias
Auxiliary hypothesis bias:
A form of rescue bias and thus of interpretive bias, which
occurs in introducing ad hoc modifications to imply that an
unanticipated finding would have occurred otherwise had the
experimental conditions been different.
Because experimental conditions can easily be altered in
many ways, adjusting a hypothesis is a versatile tool for
saving a cherished theory.
12. Berksonian & Berkson’s Bias
Berksonian bias:
A general term to indicate all types of bias that have the
structure of selection bias, based on the assumption that
Berkson originally described a bias with that structure.
Berkson’s bias: (Syn: Berkson’s fallacy):
A form of selection bias that arises when the variables whose
association is under study affect selection of subjects into the
study.
It is a particular concern in hospital-based studies, especially
when prevalent or previously diagnosed cases are not
excluded.
13. Berkson’s Bias
Berkson’s bias: (Syn: Berkson’s fallacy):
Joseph Berkson (1899-1982) described an imaginary hospital-
based case-control study wherein the controls are patients
with other diseases, and the “exposure” is also a disease; he
noted that in such a study the association between the
disease prevalences is expected to differ from the
corresponding association in the general population.
This difference in the association has historically been
referred to as Berkson’s bias.
14. Bias Due to Withdrawals
Bias due to withdrawals:
A difference between the true effect and the association
observed in a study due to characteristics of subjects who
choose to withdraw.
15. Bias Due to Instrument Error
Bias due to instrument error:
Systematic error due to faulty calibration, inaccurate
measuring instruments, contaminated reagents, incorrect
dilution or mixing of reagents, etc. See also contamination,
data; information bias; measurement bias.
17. Bias in Epidemiologic Studies
(Types)
Several types of bias exist in research.
Sackett et al. have listed 19 types of bias commonly
encountered in epidemiological studies.
Choi has expanded this list further to 65.
18. Bias in Epidemiologic Studies
(Types)
Selection bias
Information bias
Confounding
19. Bias in Epidemiologic Studies
(Selection Bias)
It is a systematic error resulting from participant
selection procedures or factors influencing
participation.
Occurs in all types of study (observational &
experimental).
Cannot be corrected analytically (must be prevented)
20. Bias in Epidemiologic Studies
(Selection Bias)
Prevalence-incidence bias:
The high case-fatality rate in the early stages of clinically
manifested coronary artery disease may invalidate the study
of possible etiological factors, since the persons available for
study as cases are the survivors (severe cases are absent).
Likewise, myocardial infarction may be silent. Clinical features
may be absent, and the biochemical and electrocardiographic
changes in myocardial infarction may return to normal after an
infarct (these mild cases will not appear among cases for
study). The type of bias introduced into the study may be clear
by contrasting a cohort study (where the disease is identified
in all its forms).
22. Bias in Epidemiologic Studies
(Selection Bias)
Minimising selection bias:
Clear definition of study population
Explicit case and control definitions
Cases and controls from same population
23. Bias in Epidemiologic Studies
(Information Bias)
Systematic error in the measurement of information on
exposure or outcome.
Differences in accuracy:
of exposure data between cases and controls.
of outcome data between different exposure groups.
Study subjects are classified in the wrong category.
24. Types of Information Bias
Interviewer bias
Recall bias
Reporting bias
Publication bias
Follow up bias
25. Types of Information Bias
Interviewer bias:
Investigator asks cases and controls differently about
exposure
Cases of
listeriosis
Controls
b
Eats soft cheese a
Does not eat
soft cheese
c d
Investigator may probe listeriosis cases about consumption of
soft cheese
Overestimation of “a” → Overestimation of OR
26. Types of Information Bias
Recall bias:
Cases remember exposure differently than controls
Mothers of
Children with
malformation
Controls
b
Took tobacco,
alcohol, drugs
a
Did not take c d
Mothers of children with malformations will remember past exposures better
than mothers with healthy children.
Overestimation of “a” → Overestimation of OR
27. Types of Information Bias
Follow up bias:
Unexposed are less likely diagnosed for disease than
exposed
Example:
• Cohort study to investigate risk factors for
mesothelioma.
• Difficult histological diagnosis.
• Histologist more likely to diagnose specimen as
mesothelioma if asbestos exposure known.
28. Berkisonian Bias
A special example of bias after Dr. Joseph Berkson
who recognized this problem.
The bias arises because of the different rates of
admission to hospitals for people with different
diseases (i.e., hospital cases and controls)
29. Berkisonian Bias
Example:
Household interviews were performed on random samples of
the general population asking about musculoskeletal and
respiratory diseases and recent hospitalizations. In the
general population, there appeared to be no association
between these two disorders (OR = 1.06), but in the subset of
the population who had been in hospital during the previous
six months, individuals with musculoskeletal disorders were
more likely to have respiratory disease than not (OR = 4.06).
This occurred because individuals with both disorders were
more likely to be hospitalized than those with only one of the
disorders.
30. Available at: http://www.sph.emory.edu/activepi/Instructors/Kevin_MSword/Lesson_9boh.htm. Accessed
on Oct 18, 2011.
Why misclassification of disease status?
•Incorrect Diagnosis
•Limited knowledge
•Diagnostic process complex
•Inadequate access to technology
•Laboratory error
•Disease subclinical
•Detection bias (e.g. more thorough exam in exposed)
•Subject Self report
•Incorrect recall
•Reluctant to be truthful
•Records incorrectly coded in data-base
32. Differential misclassification:
◦ When the misclassification results in exposure is
incorrectly more in cases than in controls. Or vice
versa; like one group has a lot more incorrect
information than the other group
Non-differential misclassification:
◦ When the misclassification is not related to
exposure status or disease status. And is occurring
at the same proportion in both groups; e.g. if 20%
of cases are classified incorrectly on exposure in
cases and about 20% in controls too
33. An obstetrician wanted to study the
association between congenital
malformations and history of infections
during pregnancy.
He interviewed women (a group who
delivered children with malformations, and a
group of women with normal children). He
asked about history of all types of infections
during pregnancy.
34. After finishing the interviews, he also wanted to
go through the women's’ medical records, in
order to minimize recall bias.
He discovered that women who had a baby with
malformation tended to remember all infections
during pregnancy more than the mothers with
normal babies.
What kind of misclassification is this?
35.
36. Bias in Prospective Cohort Studies
Loss to follow up:
The major source of bias in cohort studies
Assume that all do / do not develop outcome?
Ascertainment and interviewer bias:
Some concern: Knowing exposure may influence how
outcome determined.
Non-response, refusals:
Little concern: Bias arises only if related to both exposure
and outcome.
Recall bias:
It is not a problem: Exposure determined at time of
enrolment.
37. Bias in Retrospective Cohort & Case-control
Studies
Ascertainment bias, participation bias, interviewer bias:
Exposure and disease have already occurred →
differential selection / interviewing of compared groups
possible.
Recall bias:
Cases (or ill) may remember exposures differently than
controls (or healthy)
38. Controlling for Information Bias
- Blinding
prevents investigators and interviewers from
knowing case/control or exposed/non-exposed
status of a given participant
- Form of survey
mail may impose less “white coat tension” than a
phone or face-to-face interview
- Questionnaire
use multiple questions that ask same information
acts as a built in double-check
- Accuracy
multiple checks in medical records
gathering diagnosis data from multiple sources
39. Bias Due to Confounding
Bias due to confounding factors.
Confounding factors: associated with both exposure and out
come
Differs from selection and information bias because it can be
evaluated and controlled to some extent in the analysis phase
of the study.
It can be removed by matching cases and controls.
40. References
Main Textbook:
1.K. Park's (2015): Textbook of Preventive and Social Medicine. Banarsidas Bhanot-Jabalpur. 23rd edition.
Other references:
1.Text Book of Public Health and Community Medicine. RajVir Bhalwar, Department of Community
Medicine, Armed Forces Medical College, Pune, in collaboration with WHO, India Office, New Delhi
(2009).
2.Lucas, A. and Gilles, H. (2003): Short Textbook of Public Health Medicine for the tropic, 4th edition,
Oxford University Press Inc., New York, USA.
3.Portney, L. G. and Watkins, M. P
. (2008): Foundation of Clinical Research. Applications to Practice. 3rd
edition.
4.Kumar, R. (1996): Research methodology. A step by step guide for beginners. 3rd edition.
5.Miller, D. C. (1991): Handbook of Research Design and Social Measurement. 5th edition.
6.Altman, D. G. (1991):Practical statistics for medical research. Boca Ratón, Chapman & Hall/ CRC;
Websites:
1. World Health Organization (WHO): http://www.who.ch
2. Centers for Disease Control and Prevention (CDC), USA: http://www.cdc.gov
3. The Johns Hopkins Bloomberg School of Public Health, OPENCOURSEWARE (OCW) project:
http://ocw.jhsph.edu
4. The WWW Virtual Library (Medicine and Health): Epidemiology
(http://www.epibiostat.ucsf.edu/epidem/epidem.html).