7. ERROR AND BIAS-2010-2012
Transcript of 7. ERROR AND BIAS-2010-2012
-
7/28/2019 7. ERROR AND BIAS-2010-2012
1/18
Company
Logo
ERROR AND BIAS IN OBSERVATIONLilik Zuhriyah
20 April 2012
-
7/28/2019 7. ERROR AND BIAS-2010-2012
2/18
1
2
3
4
Sources of error and bias
Practical issues in reducingbias and error
Assessment the consistency of
information
Effects of misclassification onstudy results
Sub Topics
-
7/28/2019 7. ERROR AND BIAS-2010-2012
3/18
Introduction
Assessing error and bias is the first step in
the critical appraisal of a study
Error (Non-differential error) = inaccuracy
which is similar in the different groups ofsubjects being compared
Ex : bleeding, volume of breastmilk
Bias = inaccuracy which is different in sizeor direction in one of the groups under
study than in the others
-
7/28/2019 7. ERROR AND BIAS-2010-2012
4/18
Types of Error
Random Error
Systematic Error = Bias
any systematic error in the design,conduct or analysis of study that results in
a mistaken estimate of an exposureseffect on the risk of disease (Gordis,1996)
If the result of study close to the correctvalue, it has little error.
-
7/28/2019 7. ERROR AND BIAS-2010-2012
5/18
True value of factor
1. Random variation
2. Biological variation
True value of factor at the point of measurement
3. Performance of the instrument used to measure the factor : bias due to the
influence of the subjects being assessed or bias due to the influence of the
observers
4. Recording of the results
5. Computation of the results
Value of the variable used in the study
Sources of Error and of Bias
-
7/28/2019 7. ERROR AND BIAS-2010-2012
6/18
VALIDITY RELIABILITY RELATIONSHIP
Valid Tidak Valid
Reliabel A : Robin Hood
scenario: Your measure
is both reliable and valid
B : Consistent
but wrong
Tidak
Reliabel
C : Seldom hit the
center of the target but,on average, you are
getting the right answer
for the group (but not
very well for individuals.
In this case, you get a
valid group estimate,
but you are inconsistent
D : Your hits
spread acrossthe target and
you are
consistently
missing the
center. In this
case, your
measure neither
reliable nor valid
Matriks sasaran tembak (Bulls eye) (Murti, 2003 p 181 & Greenberg, 1996 p. 134)
-
7/28/2019 7. ERROR AND BIAS-2010-2012
7/18
VALIDITY RELIABILITY RELATIONSHIP
Validity and reliability are separate ideas, in fact, theyre
related to each other
If the spots are less scattered, it show less variability and
more precision (Confidence Intervals)
Confidence intervals are a numericmeasure of the range
within which such variation is likely tooccur. The 95%confidence intervals represent the range within which we
are likelyto find the underlying true treatment effect.
The greater the precision the narrower the confidence
intervals. The greater precision results from larger sample sizes
and consequent larger numberof events.
C versus A show increased random error (error)
B & D versus A & C show systematic error (bias)
-
7/28/2019 7. ERROR AND BIAS-2010-2012
8/18
Relation of Systematic Error & Random Error to
Study Size
Systematic Error
Random Error
Study Size
E
r
r
o
r
-
7/28/2019 7. ERROR AND BIAS-2010-2012
9/18
Sources of Bias
Selection Bias
The way the subjects have been selected
Information Bias
The way the study variables are measured
Confounding
Sourcesof bias
Some confounding factor that is not completely controlled
-
7/28/2019 7. ERROR AND BIAS-2010-2012
10/18
Selection Bias
When the association between exposure anddisease differs for those who participate andthose who do not participate in the study
Example :
New Screening Test to detect colon Cancer
Volunteer Vs Non Volunteer
Surveillance bias Disease maybe better in the monitored
population than in the general population
-
7/28/2019 7. ERROR AND BIAS-2010-2012
11/18
Information Bias
Bias may be introduced in the way information
is abstracted from medical, employment or
other records or from the manner in which
interviewers ask questions.
Ex : case control study of pancreaticcancer. The respondent is a family member
Non response are generally not a
representative group.
-
7/28/2019 7. ERROR AND BIAS-2010-2012
12/18
Evaluation of Bias
The evaluation of bias is subjective :Presence of bias
Its direction and potential magnitude of
effects on the results : Positive (OR target pop > OR sourcepop) or negative (OR target pop < ORsource pop)
Towards (underestimate) or away from(overestimate) null hypothesis
-
7/28/2019 7. ERROR AND BIAS-2010-2012
13/18
Bias in Case Control Studies
Ensuring that cases and controls are assessed by
the same observer or group of observers, and bythe same methods used under the same
circumstances
Need additional observation (recording) :
the length of time taken for examination/
interviews
the interviewers assessment of the cooperation
of the subject and the degree of difficultyexperienced with some of the key questions
Asking the subjects / interviewer whether they
aware of any relationship
-
7/28/2019 7. ERROR AND BIAS-2010-2012
14/18
Practical Issues in Reducing Bias and Error
The main approach to minimize error and
bias :
The definition of the item to be recorded
The choice of methods of measurementThe standardization of procedures
Quality control of all aspects of data
gathering and processing
-
7/28/2019 7. ERROR AND BIAS-2010-2012
15/18
-
7/28/2019 7. ERROR AND BIAS-2010-2012
16/18
Assessment of bias and error
ASSESSMENT OF OBSERVATION BIAS AND ERROR
What is the definition of the factor being
assessed ?
Is it the same for each group ?
Is it appropriate to the hypothesis ?
What is the same method of
assessment ?
Instrument used
Observer making the assessment
Circumstances of use
Subjects circumstances
Subjects knowledge and cooperation
Are the methods of assessment similar
for each group ?
Are the subjects, or the observers,aware of the groping of the subjects
when the assessment is made?
How accurate and reliable is the
method of assessment?
When is the observation made ?In calendar time
In relation to the hypothesis
Is it the same for each group
How are the data handled ?
Recording and coding
computation
Are the methods the same for each
group ?
-
7/28/2019 7. ERROR AND BIAS-2010-2012
17/18
Bias, and Methods to control bias
The most important source of bias arevariation in the subjects response to themethod of assessment, and variation in theobservers response.
The general principle for minimizing bias isto ensure that the methods used are appliedin the same mannerand with the same careto all subjects
Double blind (possible in clinical trial, but nopossible in prospective design) is to reduceerror
-
7/28/2019 7. ERROR AND BIAS-2010-2012
18/18
Company
Logo