GATE: a Graphic Approach To Evidence based practice

Critically Appraised Topic (CAT): Applying the 5 steps of Evidence Based Practice
Using evidence from iagnostic test acc!racy st!dies
Assessed "y: ate:
Pro"lem
Describethe problem thatled you to seek an answer from the literature about diagnostic accuracy.
#tep $: As% a foc!sed 5&part '!estion !sing PEC(T frame)or%
Population /
patient / client
Describe relevant patient/client/population group (be specifc about: symptoms, signs,
medical condition, age group, se, etc.! that you are considering testing
"posure
(#arget
disorder!
Describe the #arget disorder (disease/condition! to be diagnosed. $s it relevant to
consider levels/categories of severity/stage%
&omparison(n
o #arget
disorder!
Describe the typical health status of those without the target disorder who would also
receive the test. 're they likely to be disease free or have other co(morbidities%
)utcome
(#est!
Describe the test, including levels/categories if relevant, that you are considering doing
(note the *outcome+ in a diagnostic test accuracy study is the test result.
#ime #ime is not usually considered eplicitly in a diagnostic test accuracy ,uestion
#tep *: Access (#earch) for the "est evidence !sing the PEC(T frame)or%
P"&)# item Primary -earch
#erm
-ynonym . -ynonym /
Pop!lation +
Participants /
patients /
clients
"nter your key search
terms for P, " 0 ). & 0
# seldom useful for
searching. 'dd mesh
terms (e.g. sensitivity
0 specifcity! 1/or
diagnostic flter to
refne. 2se 3"-4
terms (from Pub3ed! if
available, then tet
words.
)5 $nclude relevant
synonym
)5 $nclude relevant
synonym
'6D
Eposure
(#arget
disorder!
's above )5 's above )5 's above '6D
Comparison
(no #arget
disorder!
's above )5 's above )5 's above '6D
(utcomes
(#est!
's above )5 's above )5 's above '6D
(Time! 's above '6D 's above '6D 's above
,imits
-.ilters
Pub3ed has ,imits (eg age, "nglish language, years!0 Pub3ed &linical 7ueries has
.ilters (e.g. study type! to help focus your search. 8ist those used.
ata"ases searched:
Database &ochrane -5s )ther -econdary
-ources
Pub3ed /
)vid3edline
)ther
6umber of
"nter number of hits "nter number of hits "nter number of hits "nter number of hits
9:: /;.:: Please contribute your comments and suggestions on this form to:
rt.<ackson=auckland.ac.n>
.
GATE CAT / iagnostic Test Acc!racy
#t!dies
publications
(4its!
from
&ochranedatabase
search for -ystematic
5eviews (-5!.
from other
secondary sources
(specify source!
from Pub3ed /)vid/etc
(specify database!
from other sources (e.g.
?oogle scholar, ?oogle!
"vidence -elected
"nter the full citation of the publication you have selected to evaluate.
@ustifcation for selection
-tate the main ob<ectives of the study.
"plain why you chose this publication for evaluation.
9:: /;.:: Please contribute your comments and suggestions on this form to:
rt.<ackson=auckland.ac.n>
/
iagnostic test acc!racy st!dies
#tep 0: Appraise #t!dy
0a1 escri"e st!dy "y hanging it on the GATE frame (also enter st!dy n!m"ers into the
separate e2cel GATE calc!lator)
P
o
p
!
l
a
t
i
o
n
-tudy
-etting
Describe when 0 from where participants
recruited (e.g. what year(s!, which country,
urban/rural/ hospital/community!
"ligible
population
5ecruitme
nt process
Defne eligible population / main eligibility
(inclusion and eclusion! criteria (e.g. was
eligibility based on presenting symptoms / signs,
results of previous tests, or participants who had
received the test or reference standard%
Describe recruitment process(e.g. were eligibles
recruited from hospital
admissions/electoral/birth register, etc!. 4ow
they were recruited (e.g. consecutive eligibles!%
Participant
s
Ahat percentage of the invited eligibles
participated%Ahat reasons were given for non(
participation among those otherwise eligible%
E
2
p
o
s
!
r
e

-

C
o
m
p
a
r
i
s
o
n
E2pos!re Gro!p
Comparison Gro!p
(EG)
(CG)
'llocation
method
'llocated by measurement of #arget
disorder into those with disorder (5ef
standard 1ve! 0 those without disorder (5ef
standard (ve!
"posure
(5ef -td.
1ve!
Describe reference standard positive disorder:
what, howdefned, how measured, when, by
whom (level of epertise%!. $nclude description
of categories if more than yes/no
&ompariso
n
(5ef -td. B
ve!
Describe reference standard negative
disorder(as above!
(
!
t
c
o
m
e
s
)utcome(s
!
(#est!
Describe the diagnostic test: what, how defned,
how measured, when, by whom (level of
epertise%!. $nclude description of categories if
more than yes/no
T
i
m
e
#ime -tatewhen test was donein relation to when the
reference standard was done.
3
e
p
o
r
t
e
d
Enterthe main reported
res!lts
)utcome
5isk
estimat
e
&onfdence $nterval
-ensitivity
9:: /;.:: Please contribute your comments and suggestions on this form to:
rt.<ackson=auckland.ac.n>
:
-pecifcity
1ve 85
(ve 85
PP9
6P9
Complete the 4!m"ers on the separate GATE Calc!lator for iagnostic#t!dies
iagnostic test acc!racy st!dies
#tep 0: Appraise #t!dy
0"1 Assess ris% of errors !sing 3A5"o5A4
Appraisal'!estions (3A5"o5A4)
3is%
oferrors
67 27 87
na
6otes
5ecruitment/'pplicability*errors9: ,uestions on risks to application of results in practiceare in
blue boes
$nternal study design errors: ,uestions on risk of errors within study (design 0 conduct! are in
pink boes
'nalyses errors: ,uestions on errors in analyses are in orange boes
5andom error: ,uestions on risk of errors due to chance are in the green bo
:ey for scoring ris% of errors: 6 ; lo)< 2 ; of concern< 8 ; !nclear< na ; not
applica"le
P
a
r
t
i
c
i
p
a
n
t

P
o
p
!
l
a
t
i
o
n
3ecr!itment appropriate to study ob<ectives%$s it possible to defne groups/populations
that the fndings are applicable to%
-tudy -etting appropriate% -core risk
of error
as: 1, , %
or na (see
key
above!
Doesthe study setting (e.g. what year(s!, which
country, urban / rural, hospital / community!
support the applicability of the study results%
-tudy planned before reference
standard and tests done%
Aas the study done prospectively or was it a
retrospective use of available data%
$f retrospective was the participant population
chosen primarily because of available test data
or target disorder data%
"ligible population appropriate% Aas the eligible population from which
participants were identifed relevant to the
study ob<ective%
Aere inclusion 0 eclusion criteria well defned
0 applied similarly to all potential eligibles%
Participants similar to all
eligibles%
Did the recruitment process identify participants
likely to be similar to all eligibles% Aas suCcient
information given about eligibles who did not
participate%
Dey personal (risk/prognostic!
characteristics of participants
reported% 'ppropriate spectrum
of participants%
Aas there suCcient information about baseline
characteristics of participants to determine the
applicability of the study results% Aas any
important information missing% Aas there an
appropriate spectrum of people similar to those
in whom the test would be used in practice%
9:: /;.:: Please contribute your comments and suggestions on this form to:
rt.<ackson=auckland.ac.n>
E
E
2
p
o
s
!
r
e
s

-
C
o
m
p
a
r
i
s
o
n
s
Allocationto "? 0 &? done well%
5eference standard suCciently
well defned and well measured
so participants allocated to
correct #arget disorder groups%
Aere reference standarddefnitions described in
suCcient detail for the measurements to be
replicated% Aere the measurements done
accurately% Aere criteria / cut(oF levels of
categories well <ustifed!
5eference standard measured
prior to #est% $f not, was it
measured blind to #est result%
Aas reference standard administered whatever
the test result and interpreted without
knowledge of the test result% $f not, was it likely
to cause bias%
Prevalence (pre(test probability!
of #arget disorder typical of usual
practice%
6ote: $f prevalence (pre(test probability! of
target disorder similar to usual practice, these
data can be used to help determine post(test
probabilities in practice (also need 85s!
5aintenance in allocated groupsand throughout study suCcient%
Proportion of intended
participants receiving both #est
and 5eference
-tandardsuCciently high%
Aas there a particular subgroup of the eligible
participants not given either the #est or the
5eference -tandard%Aas this suCcient to cause
important errors%
&hange in #arget disorder/#est
status in periodbetween #est and
5eference -tandard being
administered
$f there was a considerable delay between #est
and 5eference -tandard then study could some
new events have occurred or treatment may
have been started that could inGuence the
results of the #est/5ef -tandard% $f so, was this
suCcient to cause important bias%
(
!
t
c
o
m
e
s
"lind or o"=ective 5eas!rement of )utcomes: were they done accurately%
#est measured blind to 5eference
-tandard -tatus status%
Aere #esters aware of whether participants
were 5eference -tandard positive or negative%
$f yes, was this likely to lead to biased
measurement%
#est measured ob<ectively% 4ow ob<ective was the#est measurements (e.g.
automatic test, strict criteria!%
Ahere signifcant <udgment was re,uired, were
independent ad<udicators used%
Aas reliability of measures relevant (inter(rater
0 intra(rater!, 0 if so, reported%
#est safe, available, aFordable 0
acceptable in usual practice%
Aould it be practical to implement this #est in
usual practice% 4ow safe, available, aFordable
0 acceptable might it be%
3
e
s
!
l
t
s
A4alyses: were they done appropriately%
$f 5ef -tandard 1ve 0 (ve groups
not similar at baseline was this
ad<usted for in the analyses%
-ome factors that diFer between those with 0
without the target disorder could eFect test
accuracy (e.g. age, obesity, co(morbidities!,
although these are typically not reported.
"stimates of #est
sensitivity/specifcity etc given or
calculable% Aere they calculated
correctly%
Aere raw data reported in enough detail to
allow // tables to be constructed (i.e. #P, HP,
H6 0 #6! in ?'#" frame 0 to calculate
estimates of test specifcity and sensitivity if
entered into ?'#" calculator% Aere ?'#" results
similar to reported results%
3easures of the amount of Aere confdence intervals 0/or p(values for
9:: /;.:: Please contribute your comments and suggestions on this form to:
rt.<ackson=auckland.ac.n>
I
random errorinestimates given or
calculable% Aere they calculated
correctly%
study resultsgiven or possible to calculate% $f
they could be entered into ?'#" calculator,
were ?'#" results similar to reported results%
#!mmary of #t!dy Appraisal
Aas therisk of errordue to
internal study design 0 conduct
low enough for the results to be
reasonably unbiased%
2se responses to ,uestions in pink boes above
Aas the risk of errors due to
study analyses suCciently low for
the results to be reasonably
unbiased%
2se responses from the orange boes above
Aas the amount of random error
in study results eFects low
enough for the results to be
meaningful%
2se responses to ,uestions in green bo above.
Aould you make a diFerent decision if the true
eFect was close to the upper confdence limit
rather than close to the lower confdence limit%
Ahat is your level of concern
about the applicability of these
fndings to your problem%
2se responses to ,uestions in blue boes above
9:: /;.:: Please contribute your comments and suggestions on this form to:
rt.<ackson=auckland.ac.n>
J
iagnostic test acc!racy st!dies
#tep >: Apply1 Consider+)eigh !p all factors - ma%e (shared) decision to act
"pidemiological evidence: summarise the
,uality of the study appraised, the magnitude
and precision of the measure(s! estimated and
the applicability of the evidence.
'lsosummariseits consistency with other
studies (ideally systematic reviews! relevant to
the decision.
&ase circumstances: what circumstances(e.g.
disease process/ co(morbidities Kmechanistic
evidenceL, social situation! specifcally related
to the problem you are investigating may
impact on the decision%
-ystem features: were there any system
constraint or enablers that may impact on the
decision%
Ahat values 0 preferences may need to be
considered in makingthedecision%
ecision: #aking into account all the factors above what is the best decision in this case%
#tep 5: A!dit !s!al practice (.or ?!ality @mprovement)
$s there likely to be a gap between your usual practice and best practice for the problem%
9:: /;.:: Please contribute your comments and suggestions on this form to:
rt.<ackson=auckland.ac.n>
M
The A&factor