GATE: practicing EBP with pictures (3/10/05)
GATE Frame
GATE: ‘EBP with pictures’
Graphic Approach to Teaching EBP
Graphic Appraisal Tool for Epidemiology
Graphic Architectural Tool for Epidemiology
Rod Jackson
EPIQ group
University of Auckland, NZ
www.epiq.co.nz
2/11/05
©
Sicily
GATE: Graphic Appraisal Tool for Epidemiology
©
GATE Frame: PECOT (design)
GATE Frame: RAAMbo (study appraisal)
P
• Population
• Exposure
• Exposure
E
(Intervention)
C
T
• Outcome
O
• Time
E
• Comparison
• Comparison
• Outcome
P
• Population
T
• Time
©
R
• Representative?
A
• Allocated or
Adjusted?
A
• Accounted for?
C
O
Mbo • Measurement
•blind or
•objective?
©
‘hang the study on the GATE frame’
Exercise for Q1. Ask a focussed 5-part question
The 4 + 1 steps of EBP (AAAAA)
1. Population……
P
1. Ask a focussed question
2. Access (ie. search) appropriate evidence
2. Exposure……..
3. Appraise evidence for validity, impact & precision
E
4. Apply evidence accounting for patient values, clinical
C
& policy issues (i.e. answer question)
5. Audit personal skills in doing steps 1-4; Audit your
practice (do you apply step 4 in usual practice).
T
Modified from DS et al
©
1
O
3. Comparison…….
4. Outcome…….
5. Time……….
GATE: practicing EBP with pictures (3/10/05)
Q1. Ask a focussed 5-part question: HRT example
Q2. Access appropriate epidemiological evidence
1. P: In post menopausal women with CHD,
P
1. Population……
P
2. E: does HRT,
E
C
E
3. C: compared with no HRT,
4. O: reduce the risk of future CHD,
O
T
2. Exposure……..
T
5. T: over 3-5 years
©
O
3. Comparison…….
4. Outcome…….
5. Time……….
©
Q2. Access appropriate epidemiological evidence:
HRT example
P
E
T
C
C
O
Q3. Appraise evidence for validity, impact & precision
1. P: postmenopausal
P
2. E: hormone replacement therapy
E
3. C: (placebo)
C
4. O: coronary disease
O
5. Time……….
T
©
R
A
A
Mbo
©
Q4. Apply the evidence
Q4. Apply evidence accounting for patient values,
clinical & policy issues
Epidemiologic
evidence
Patient preferences
Policy issues
clinical
considerations
‘clinical expertise:
other ways of knowing’
Apply
©
X-factor: ‘integrating & applying’ = CLINICAL expertise
2
GATE: practicing EBP with pictures (3/10/05)
CATs:
CAT forms:
(in Excel)
Intervention
Diagnosis
Prognosis/Risk
Critically
Appraised
Topics:
Systematic Reviews
download from:
‘a tool for
modeling the 5
steps of EBP’
www.epiq.co.nz
www.epiq.co.nz
GATE Frame: PECOT exercise:
HERS study on the GATE frame’
P
‘hang the
• Population ………………………………….
*Hulley et al. HERS JAMA. 1998;280:605-613
• Exposure ………………………………….
E
T
C
Read abstract and fill in the GATE Frame: PECOT
• Comparison ………………………………….
• Outcome ………………………………….
O
• Time ………………………………….
*Hulley et al. HERS JAMA. 1998;280:605-613
Source Population:
GATE: epi study design (P)
Population
68,561 women screened from
20 outpatients/community
screening centres
P
Eligible Population:
Post-menopausal,
established CHD, <
80 yrs, no MI in
last 6 mths, no
HRT last 3 months
©
3
P
Participant Population:
all eligibles
invited (2763)
GATE: practicing EBP with pictures (3/10/05)
GATE: epi study design (E&C)
Population
Exposure (E)
[intervention]
E
Identical Placebo
(n=1383)
HRT (n=1380)
P
C
Exposure (E)
[intervention]
Comparison (C)
[control]
E
C
Comparison (C)
[control]
©
GATE: epi study design (O)
Population
1º outcome: non fatal
MI or CHD death
P
E
yes
Exposure
E
yes
a
no
c
C
O
Comparison
no
C
O
Outcomes (O)
b
Outcomes (O)
d
©
1º outcome: non fatal
MI or CHD death
E
yes
no
172
mean HDL cholesterol (mmol/L)
HRT
Placebo
1.4
1.27
C
176
O
Outcomes (O)
4
Outcomes (O)
GATE: practicing EBP with pictures (3/10/05)
GATE: epi study design (T)
Population
Exposure
P
E
E
Comparison
C
O
Time (T)
Time (T)
yes
O
no
C
CHD outcomes measured
over 4.1 years
(longitudinal)
Outcomes (O)
©
GATE: HERS
E
Time (T)
Population =
women with CHD
C
Exposure = HRT
HDL cholesterol
measured at 1 year
(cross-sectional)
O
Time =
4.1 yrs
2763
1380
172
1383
O
176
Comparison = placebo
Outcomes =
non-fatal MI
& CHD death
©
GATE: epi study design
Population
Multiple Exposure
categories
GATE: epi study design
P
Population
Continuous measure
of Exposure: e.g.
body mass index
Comparison
E1 E2 E3 C
Continuous measure
of Outcomes
e.g. lipids
Multiple Outcome
categories
©
5
P
E
high…med..low
Correlation coefficient
low
O
medium
high
©
GATE: practicing EBP with pictures (3/10/05)
Cohort (Follow-up) study: archetypal epidemiological study
The GATE Approach: every epidemiological study
hangs on the GATE frame
there is only one study design:
Population
P
cohort
• RCT - interventions
• Cohort studies - prognosis / interv./
aetiology
E
Exposure
Real life time
• Cross-sectional studies - diagnosis
• Case-control studies - interv./aetiol.
Time
C
O
T
nonrandomised
allocation
Comparison
ill-health
Outcomes
©
GATE: Graphic Appraisal Tool for Epidemiology
“Life” is a cohort study: a “natural experiment”
Cohort study (aetiology): British Doctors Study*
Population
Smokers
Cohort of British
Doctors established
in 1951
P
EG
Non-smokers
CG
T
yes
Lung Cancer .
no
Cohort study (prognosis): *Danish Breast Cancer Cooperative
Population
Non-randomised
allocation (selfreported smoking)
childbirth <2yrs
before diagnosis
Cohort of women
diagnosed with
breast cancer
P
EG
CG
No childbirth <2 yrs
before diagnosis
50 years
O
T
10 years
yes
O
death
no
©
©
*Doll et al. BMJ 2004;328:1519
Cohort study (non-randomised intervention):
Population
HRT
No HRT
CG
Randomised Controlled Trial (RCT): HERS
Nurses Health Study*
Population
Exposure (intervention):
HRT
Non-randomised
allocation (selfreported use)
P
Randomised
EG
CG
10 years
T
yes
CHD
.
no
*Kroman et al. BMJ 1997;315:851-5
Cohort of US
Nurses
P
EG
Non-randomised
allocation to
prognostic groups
Comparison (control):
allocation
Placebo
T
Outcomes: CHD
O
Time
O
©
©
*Stampfer et al. NEJM 1991;325:756-62
Hulley et al. JAMA1998;280:605-13
6
GATE: practicing EBP with pictures (3/10/05)
Case series
Population
Population
Before-After Study
(& cross-over study)
P
P
CG
Exposure Grp
EG
No one allocated to
Comparison Group
C
Time
T
O
Outcomes
Allocation:
randomised or
non-randomised
EG
Exposure Grp
T
Comparison Grp
O
Outcomes
©
©
Cross-sectional (prevalence) study/survey: PECO
Real life time
Population
↑ LTPA
EG
Diagnostic Test Accuracy Study (cross-sectional)
P
CG
Population
O
Time ?
P
EG:
Gold Standard +ve
↑ angina
EG
CG
GS+
O
↓ angina
EG
CG
+
↓ LTPA
Test Outcomes .
-
©
a
CG:
Gold Standard -ve
GSO
Time
d
c
©
Case-control study (aetiology):
Diagnostic Test Study Application (cross-sectional)
Population
P
Auckland Heart Study
Aucklanders aged
35-74 yrs on
electoral roll
P
CONTROLS
EG:
Test +ve
EG
CG
T+
+ve predictive value.
a
c
CG:
Test -ve
TO
Allocation
by measurement
drinkers
CASES
Time
yes
MI
.
no
b
d
Allocation
by measurement
b
E/C assignment & Outcomes assessed in cohort at same Time
Population
Time
-ve predictive value.
non-drinkers
eg
cg
EG
CG
T
2 years
O
©
Controls sampled from
electoral roll: selfreported drinking
Cases identified
from hospital
admission lists
(& on elec. roll)
all case-control studies are nested in ‘virtual’ cohort studies
Jackson et al. BMJ 1991;303:211-6
©
7
GATE: practicing EBP with pictures (3/10/05)
GATE: the NUMBERS
Population
P=
EG= CG=
Exposure Group
Population
Comparison Group
T=
Outcomes
GATE: the numbers = HERS
a=
b=
c=
d=
Exposure Group
P= 2763
EG= CG=
1380 1383
Comparison Group
Time
Time
Outcomes
a=
172
b=
176
T= 4.1 yrs
c ~
d ~
1208 1207
* 99% follow-up 1º outcome
©
Occurrence = outcome / population
GATE: occurrence = numerator / denominator
P=
Population
P=
Denominators =
EG (x T) and CG (x T)
sub-populations
T
EG= CG=
+
-
a=
b=
c=
d=
EG= CG=
outcomes
©
Exp. Group Occurrence
EGO = a / EG x T
Numerators =
a or c and b or d
Epidemiology = numerator / denominator (E=N/D)
a=
b=
c=
d=
Comp. Group Occurrence
CGO = b / CG x T
T=
©
Occurrence
EGO = Exposure Group Occurrence (A/[EGxT])
= 30.40 / 1000 persons / year
CGO = Comparison Group Occurrence (B/[CGxT]
= 30.40 / 1000 persons / year
8
GATE: practicing EBP with pictures (3/10/05)
GATE: occurrence HERS
Effects: comparing occurrences
P= 2763
Relative Effect/Risk (RR)
=
EGO
CGO
e.g. relative risk, risk ratio, prevalence ratio, incidence ratio
EG= CG=
1380 1383
T= 4.1 yrs
Absolute Effect/Risk Difference (RD) = EGO - CGO
Comp. Group Occurrence
Exp. Group Occurrence
EGO = A / EGxT
A=
172
= 172/1380 x 4.1
B=
176
= 30.40/1000/yr
CGO = B / CGxT
e.g. risk difference, absolute risk
= 176/1383 x 4.1
= 31.04/1000/yr
Number Needed to Treat (NNT) to prevent/cause 1 event = 1/RD
GATE : RAAMbo (study appraisal exercise)
R
P
A
A
E
R
• Representative……………...
A
• Allocated or Adjusted……….
• Accounted for………………
C
A
Mbo
T
O
Mbo
• Measurement
•blind ………….
•Objective………………..
©
*Hulley et al. HERS JAMA. 1998;280:605-613
Q3. Appraise evidence for validity, impact & precision
R -
multiple recruitment strategies, well
Step 4
described include. Criteria, probably
representative of women with CHD
A -
random allocation
A -
100% mortality & 99% 1° outcome f/u,
good
Mbo -
1° outcome based on standardised
reasonably objective algorithm & assessed blind
to exposure
©
*Hulley et al. HERS JAMA. 1998;280:605-613
9
GATE: practicing EBP with pictures (3/10/05)
The 4 + 1 steps of EBP (AAAAA)
1. Ask a focussed question
the CAT:
2. Access (ie. search) appropriate evidence
3. Appraise evidence for validity, impact & precision
a tool for life
long learning
4. Apply evidence accounting for patient values, clinical
& policy issues (i.e. answer question)
5. Audit personal skills in doing steps 1-4; Audit your
practice (do you apply step 4 in usual practice).
Modified from DS et al
10