BiasSyst
-
Upload
drnareshchauhan -
Category
Documents
-
view
216 -
download
0
Transcript of BiasSyst
-
8/9/2019 BiasSyst
1/30
-
8/9/2019 BiasSyst
2/30
Competency to be gained
from this lecturePrevent biases that may be avoided and
understand the effect of others
-
8/9/2019 BiasSyst
3/30
Key elements
Definition of biases Selection biases Information biases Controlling the effect of biases
-
8/9/2019 BiasSyst
4/30
-
8/9/2019 BiasSyst
5/30
Effect of misclassification
!on"differential misclassification&sually reduces the strength of associationbet'een outcome and e#posure$(( or )( closer to *%Is not a sufficient reason to dismiss the results ofa study reporting an association
Differential misclassification $bias%ay commonly increase the strength of the
associationay also reduce the strength of the association
Biases
-
8/9/2019 BiasSyst
6/30
-
8/9/2019 BiasSyst
7/30
Types of biases
Selection biasesBiases in the 'ay sub-ects enter a study
Information biasesBiases in the 'ay information is collected afterinclusion in a study
Biases
-
8/9/2019 BiasSyst
8/30
.evel of operation of biases
S y s t e m a t ic r e c r u i t m e n to f c a s e s / c o n t r o l s
' i t h s p e c i f ic e # p o s u r e s t a t u s
S y s t e m a t i c c o l l e c t i o n o fi n f o r m a t io n l e a n i n g t o ' a r d s
s p e c i f ic e # p o s u r e s t a t u s
C o l l e c t i o n o f i n f o r m a t i o na b o u t e # p o s u r e
I n c l u s i o no n t h e b a s i s o f o u t c o m e
C a s e c o n t r o l
S y s t e m a t ic r e c r u i t m e n to f e # p o s e d / u n e # p o s e d
' i t h s p e c i f ic r i s 0 o f o u t c o m e
S y s t e m a t i c c o l l e c t i o n o fi n f o r m a t i o n l e a n i n g t o ' a r d s
s p e c i f i c o u t c o m e s t a t u s
C o l l e c t i o n o f i n f o r m a t i oa b o u t o u t c o m e
I n c l u s i o no n t h e b a s i s o f e # p o s u r e
C o h o r t
T y p e o f s t u d y
Selection biases
Information biases
Biases
-
8/9/2019 BiasSyst
9/30
Selection biases
Type of study Inclusion into thestudySources of selection
biases
Case control )n the basis of
outcome
Selection of casesor controls 'ithspecific e#posurestatus
Cohort )n the basis of
e#posure
Selection ofe#posed orune#posed 'ithspecific ris0
Selection biases
-
8/9/2019 BiasSyst
10/30
E#amples of selection biases
in a cohort study Sub-ects selected on the basis of e#posure )utcome not independently distributed
among e#posed and une#posedSystematic selection of e#posed sub-ects 'ith1
+igher ris0 of outcome .o'er ris0 of outcome
Systematic selection of une#posed sub-ects 'ith1 +igher ris0 of outcome .o'er ris0 of outcome
2 b
c d
a B
c d
a b
c D
a b
C dSelection biases
-
8/9/2019 BiasSyst
11/30
E#amples of selection biases
in a case control study Sub-ects selected on the basis of outcome E#posure not independently distributed
among cases and controlsSystematic selection of case"patients 'ith1
+igher fre3uency of e#posure .o'er fre3uency of e#posure
Systematic selection of control"sub-ects 'ith1 +igher fre3uency of e#posure .o'er fre3uency of e#posure
2 b
c d
a b
C d
a b
c D
a B
c dSelection biases
-
8/9/2019 BiasSyst
12/30
Sources of selection biases
Surveillance Screening / diagnosis 2dmission to health care facilities Selective survival !on"response / loss to follo' up
Selection biases
-
8/9/2019 BiasSyst
13/30
Sources of selection biases
SurveillanceSystematic notification of cases e#posed
Screening / diagnosis 2dmission to health care facilities Selective survival
!on"response / loss to follo' up
Selection biases
-
8/9/2019 BiasSyst
14/30
Sources of selection biases
Surveillance Screening / diagnosis
Systematic case search among e#posed 2dmission to health care facilities Selective survival
!on"response / loss to follo' up
Selection biases
-
8/9/2019 BiasSyst
15/30
Sources of selection biases
Surveillance Screening / diagnosis 2dmission to health care facilities
Systematic admission of1 Case"patients e#posed / une#posed Control"sub-ects e#posed / une#posed
Selective survival !on"response / loss to follo' up
Selection biases
-
8/9/2019 BiasSyst
16/30
Sources of selection biases
Surveillance Screening / diagnosis 2dmission to health care facilities Selective survival
Systematic inclusion of cases 'ho survived and'ho may be more or less e#posed
!on"response / loss to follo' up
Selection biases
-
8/9/2019 BiasSyst
17/30
Sources of selection biases
Surveillance Screening / diagnosis 2dmission to health care facilities Selective survival !on"response / loss to follo' up
Systematic inclusion of sub-ects more li0ely toparticipate 'ho may be1 ore or less e#posed ore or less at ris0
Selection biases
-
8/9/2019 BiasSyst
18/30
Information biases
Type of study Collection ofinformationSources of
information biases
Case control 2bout e#posure
Collection ofinformation leaningto'ards specifice#posure status
Cohort 2bout outcome
Collection ofinformation leaningto'ards specificoutcome status
Information biases
-
8/9/2019 BiasSyst
19/30
E#amples of information biases
in a cohort study Sub-ects selected on the basis of e#posure )utcome not independently measured among
e#posed and une#posedeasurement of a higher incidence of outcome1 2mong e#posed 2mong une#posed
easurement of a lo'er incidence of outcome1 2mong e#posed 2mong une#posed
2 b
c d
a b
C d
a b
c D
a B
c dInformation biases
-
8/9/2019 BiasSyst
20/30
E#amples of information biases
in a case control study Sub-ects selected on the basis of outcome E#posure not independently measured among
cases and controls 1easurement of a higher fre3uency of e#posure1 2mong cases 2mong controls
easurement of a lo'er fre3uency of e#posure1 2mong cases 2mong controls
2 b
c d
a B
c d
a b
c D
a b
C dInformation biases
-
8/9/2019 BiasSyst
21/30
Sources of information biases
(ecall Investigator Data 3uality Prevarication
Information biases
-
8/9/2019 BiasSyst
22/30
Sources of information biases
(ecallCases may recall e#posure more than controls
Investigator Data 3uality Prevarication
Information biases
-
8/9/2019 BiasSyst
23/30
Sources of information biases
(ecall Investigator
Systematic collection of information supportinge#pected conclusions
&nconsciously Consciously
ay be e#amined in the analysis Data 3uality Prevarication
Information biases
-
8/9/2019 BiasSyst
24/30
Sources of information biases
(ecall Investigator Data 3uality
Better e#posure data available on casesBetter outcome data available on e#posed
Prevarication
Information biases
-
8/9/2019 BiasSyst
25/30
Sources of information biases
(ecall Investigator Data 3uality Prevarication
Systematic distortion of the truth by sub-ects
Information biases
-
8/9/2019 BiasSyst
26/30
)ther limitations also
referred to as biases Intervention biases
!on"specific effect of interventions in a trial
2nalysis biases!on"systematic statistical analyses
Interpretation biasesInvestigators 'ith pre"conceived ideas
Publication biases!egative studies do not get published
-
8/9/2019 BiasSyst
27/30
Chec0list to prevent biases 'hen
preparing a protocol $*/4%*5 Is the sample representative,
2re controls representative of the population
from 'hich cases are dra'n,45 2re the e#posure and outcome criteria1
Standardi6ed,Specific,
Clear,7ell measured,2pplied consistently by trained field 'or0ers,
Controlling biases
-
8/9/2019 BiasSyst
28/30
Chec0list to prevent biases 'hen
preparing a protocol $4/4%85 Could e#posure status affect the probability
to detect the outcome,
95 Could outcome status affect the probabilityto recall the e#posure,
:5 7hat 'ill be the proportion of response,;5 7hat 'ill be the proportion of
-
8/9/2019 BiasSyst
29/30
Discussing biases critically
(evie' critically selection and informationbiases that are li0ely to have operated
(evie' the potential effect of each of thesebiases $e5g5> under" or overestimation% (evie' 'hat can be done to address these
biases a posteriori $e5g5> analysis plan%
2naly6e possible summary effect of all thesebiases on the final results Interpret data in light of this summary effect
Controlling biases
-
8/9/2019 BiasSyst
30/30
Ta0e"home messages
+ave a strong design that prevents biases E#amine ho' sub-ects enter a study Collect information in a standardi6ed 'ay Prevent the biases you can avoid> understand
those you cannot avoid