Bias, Confounding and

the Role of Chance

Principles of Epidemiology
Lecture 5

Dona Schneider, PhD, MPH, FACE

To Show Cause We Use

Kochs Postulates for Infectious Disease

Hills Postulates for Chronic Disease and Complex Questions

Strength of Association Tonights entire lecture

Biologic Credibility

Specificity

Consistency with Other Associations

Time Sequence

Dose-Response Relationship

Analogy

Experiment

Coherence

Epidemiology (Schneider)

To Show a Valid Statistical Association

We need to assess:

Bias: whether systematic error has been built

into the study design

Confounding: whether an extraneous factor is

related to both the disease and the exposure

Role of chance: how likely is it that what we

found is a true finding

Epidemiology (Schneider)

BIAS
Systematic error built into the
study design
Selection Bias
Information Bias

Types of Selection Bias

Berksonian bias

There may be a spurious

association between diseases or between a

characteristic and a disease because of the
different probabilities of admission to a
hospital for those with the disease, without the
disease and with the characteristic of interest
Berkson J. Limitations of the application of fourfold table
analysis to hospital data. Biometrics 1946;2:47-53

Epidemiology (Schneider)

Types of Selection Bias (cont.)

Response Bias those who agree to be in

a study may be in some way different from
those who refuse to participate

Volunteers may be different from those who

are enlisted

Epidemiology (Schneider)

Types of Information Bias

Interviewer Bias an interviewers knowledge

may influence the structure of questions and the
manner of presentation, which may influence
responses

Recall Bias those with a particular outcome

or exposure may remember events more clearly
or amplify their recollections

Epidemiology (Schneider)

Types of Information Bias (cont.)

Observer Bias observers may have

preconceived expectations of what they
should find in an examination

Loss to follow-up those that are lost to

follow-up or who withdraw from the study may
be different from those who are followed for
the entire study

Epidemiology (Schneider)

Information Bias (cont.)

Hawthorne effect an effect first documented

at a Hawthorne manufacturing plant; people act
differently if they know they are being watched

Surveillance bias the group with the known

exposure or outcome may be followed more
closely or longer than the comparison group

Epidemiology (Schneider)

Information Bias (cont.)

Misclassification bias errors are made

in classifying either disease or exposure
status

Epidemiology (Schneider)

Types of Misclassification Bias

Differential misclassification Errors in

measurement are one way only

Example: Measurement bias

instrumentation may be inaccurate, such as
using only one size blood pressure cuff to
take measurements on both adults and
children

Epidemiology (Schneider)

Misclassification Bias (cont.)

True Classification
Cases
Exposed
100
Nonexposed
50
150

Controls

Total

50
50
100

150
100
250

OR = ad/bc = 2.0; RR = a/(a+b)/c/(c+d) = 1.3

Differential misclassification - Overestimate exposure
for 10 cases, inflate rates
Exposed
Nonexposed

Cases

Controls

Total

110
40
150

50
50
100

160
90
250

OR = ad/bc = 2.8; RR = a/(a+b)/c/(c+d) = 1.6

Misclassification Bias (cont.)

True Classification
Cases

Controls

Total

Exposed

100

50

150

Nonexposed

50

50

100

150

100

250

OR = ad/bc = 2.0; RR = a/(a+b)/c/(c+d) = 1.3

Differential misclassification - Underestimate exposure
for 10 cases, deflate rates
Cases

Controls

Total

Exposed

90

50

140

Nonexposed

60

50

110

150

100

250

OR = ad/bc = 1.5; RR = a/(a+b)/c/(c+d) = 1.2

Misclassification Bias (cont.)

True Classification
Cases

Controls

Total

Exposed

100

50

150

Nonexposed

50

50

100

150

100

250

OR = ad/bc = 2.0; RR = a/(a+b)/c/(c+d) = 1.3

Differential misclassification - Underestimate exposure
for 10 controls, inflate rates
Cases

Controls

Total

Exposed

100

40

140

Nonexposed

50

60

110

150

100

250

OR = ad/bc = 3.0; RR = a/(a+b)/c/(c+d) = 1.6

Misclassification Bias (cont.)

True Classification
Exposed
Nonexposed

Cases
100
50
150

Controls
50
50
100

Total
150
100
250

OR = ad/bc = 2.0; RR = a/(a+b)/c/(c+d) = 1.3

Differential misclassification - Overestimate exposure
for 10 controls, deflate rates
Cases

Controls

Total

Exposed

100

60

160

Nonexposed

50

40

90

150

100

250

OR = ad/bc = 1.3; RR = a/(a+b)/c/(c+d) = 1.1

Misclassification Bias (cont.)

Nondifferential (random)
misclassification errors in assignment
of group happens in more than one
direction

This will dilute the study findings BIAS TOWARD THE NULL

Epidemiology (Schneider)

Misclassification Bias (cont.)

True Classification
Cases

Controls

Total

Exposed

100

50

150

Nonexposed

50

50

100

150

100

250

OR = ad/bc = 2.0; RR = a/(a+b)/c/(c+d) = 1.3

Nondifferential misclassification - Overestimate
exposure in 10 cases, 10 controls bias towards null
Cases

Controls

Total

Exposed

110

60

170

Nonexposed

40

40

80

150

100

250

OR = ad/bc = 1.8; RR = a/(a+b)/c/(c+d) = 1.3

Controls for Bias

Be purposeful in the study design to minimize the chance for bias

Define, a priori, who is a case or what constitutes exposure so

that there is no overlap

Example: use more than one control group

Define categories within groups clearly (age groups, aggregates of

person years)

Set up strict guidelines for data collection

Train observers or interviewers to obtain data in the same fashion

It is preferable to use more than one observer or interviewer, but not so

many that they cannot be trained in an identical manner

Controls for Bias (cont)

Randomly allocate observers/interviewer data

collection assignments

Institute a masking process if appropriate

Single masked study subjects are unaware of whether they are in

the experimental or control group

Double masked study the subject and the observer are unaware of
the subjects group allocation

Triple masked study the subject, observer and data analyst are
unaware of the subjects group allocation

Build in methods to minimize loss to follow-up