Making epidemiological evidence more accessible using pictures Rod Jackson Updated November 09
Jan 17, 2016
Making epidemiological evidence more accessible
using pictures
Rod Jackson
Updated November 09
What is Evidence Based Practice?
What is Evidence Based Practice?
The 6 steps of Evidence Based Practice
1. ASK questions relevant to your clinical problem using the PECOT (PICO) framework
2. ACCESS - search for epidemiological evidence to help answer your questions using the key PECOT terms
3. APPRAISE the evidence for its validity, effect size, precision)
4. AGGREGATE the aggregated (systematically reviewed) evidence with patient/community, clinical/health & policy issues & make an evidence-based decision
5. APPLY (implement) your decision6. [AUDIT your usual practice (i.e. compare your usual
practice for this clinical problem against ‘best’ evidence-based practice) – is there an evidence-practice gap?].
The GATE frame
Graphic Approach To Epidemiology
©
Participants
Exposure Group Comparison Group
OutcomesTime
P
E C
O
T
PECOT: the 5 parts of every epidemiological study
All epidemiological studies can be hung on the GATE frame
P
E C
O
T
EBP Step 1: ASK - turn your question into 5 parts (PECOT)
1. Participants (patient(s) you want to treat)
2. Exposure (an intervention if about therapy)
3. Comparison (there is always an alternative! - another therapy, nothing …
4. Outcome (usually a disease or condition you want to prevent or manage)
5. Time frame (over which you expect a result)
P
E C
O
T
EBP Step 2: ACCESS - search for the best evidence to answer your questions
Use the key PECOT components to choose search terms
P
E C
OT
• P
• E
• C
• O
• T
• Recruitment
• Allocation
• Maintenance
• Blind or
• Objective measurements of outcomes
EBP Step 3: Appraise the evidenceusing PECOT & RAMBO on the GATE frame
EBP Step 4: AGGREGATE the relevant information & make an evidence-based decision:’ the X-factor
©
Epidemiologic evidence (ideally
from a SR)
Clinical / population
health considerations
Policy issues
Patient / community preferences
X-factor: making evidence-based decisions
expertise: ‘putting it all together’ the art of practice
Step 5
APPLY
Implementation!
Step 6: AUDIT - evaluate & improve performance
1. Determine ‘best’ practice (EBP Steps 1-4)
2. Assess current practice: survey
3. Compare with best practice - is there a gap?
4. Consider reasons for gap, identify processes to
reduce gap & implement
5. Re-survey: is there any improvement?
= quality improvement / audit
GATE Graphic Approach To Epidemiology
Graphic Appraisal Tool for Epidemiology
Graphic Architectural Tool for Epidemiology
www.epiq.co.nz
The GATE frame
©
the shape of every epidemiological study
GATE study design (PECOT)
P
E C
O
T
GATE study analyses (EGO & CGO)
a b
c d
EG CG
GATE study appraisal (RAMBO)P
E C
OT
Recruitment
Allocation
Maintenance
Blind or Objective measurements of
outcomes
GATE study design (PECOT)
Participants
Exposure Group Comparison Group
OutcomesTime
P
E C
O
T
Participants
Study Setting
Eligible Participants
ParticipantsP
Exposure & Comparison Groups
Exposure or Intervention Group
(EG)
Comparison or Control Group
(CG)EG CG
Outcomes (O)
Outcomes (O)Oa b
c d
yes
no
‘Dis-ease’
Time (T)
T
incidence
prevalence
GATE study analyses
Denominator (Participants)D
N Numerator (Outcomes)
O = N÷DO = N÷D
All epidemiological studies involve measuring the OCCURRENCE of disease
Occurrence = Numerator ÷ Denominator
GATE study analyses
P
EG CG
O
Denominator 1:Exposure Group
EG
Numerator 1:a
Denominator 2:Comparison Group
CG
Overall Denominator
a b
c d
Numerator 2:b
Occurrence = N ÷ D
P
EG CG
O
Denominator 1:Exposure Group
EG
Numerator 1:a
Denominator 2:Comparison Group
CG
a b
c d
Numerator 2:b
Exposure Group Occurrence:EGO = a ÷ EG
Comparison Group Occurrence:CGO = b ÷ CG
Estimating effects & associations involves comparing occurrences
Relative Effect or Risk = EGO ÷ CGO
e.g. relative risk (RR), risk ratio, prevalence ratio, incidence ratio
Absolute Effect or Risk Difference = EGO - CGO
e.g. risk difference (RD), absolute risk
Number Needed To Treat (NNT) = 1 ÷ RD
Analyses
it’s all about EGO & CGO
Occurrence = N÷D per unit of timeP
EG CG
O
Denominator 1:Exposure Group
EG x T
Numerator 1= a
Denominator 2:Comparison Group
CG x T
a b
c d
Numerator 2 = b
Exposure Group Occurrence:EGO = a ÷ (EG x T)
Comparison Group Occurrence:CGO = b ÷ (CG x T)
‘person-time exposure’
GATE study appraisal (RAMBO)P
E C
OT
Recruitment
Allocation
Maintenance
Blind or Objective measurements of
outcomes
Study appraisal scores
How well was the study done?
Was it ok ( or +) or not ok ()?
or unclear (?) or not applicable (n/a)
‘no study is perfect!’
RAMBO
E C
OT
appropriate Recruitment processes?P
Study setting & eligibility criteria well described?
e.g. Recruit random/representative sample ORRecruit consecutive eligibles
‘appropriateness’ depends on study question
RAMBO
EG CG
OT
appropriate Allocation process?to EG & CG?
P
Allocation process well described?
If allocated by investigators was it done well?- was allocation random (e.g drugs)and was allocation concealed? ORIf allocated by measurement (e.g. smoking) - were E & C measured well
Allocate
EG CG
OT
P
RCT: Allocate randomly by investigators (e.g drugs)
EG CG
OT
P
Cohort: Allocate by measurement (e.g. smoking)
RAMBO
EG CG
OT
good Maintenance?did most participants remain in allocated groups (EG &
CG)
P
Participants &/or investigators blind to exposure (and comparison exposure)?
Compliance high & similar in EG & CGContamination low & similar in EG & CGCo-interventions low & similar in EG & CG
Completeness of follow-up high & similar in EG & CG
RAMBO
EG CG
OT
Blind or Objective?outcome measurements
P
If outcome measurements not Objective (eg. automated or definitive) were investigators blind to exposure (and comparison exposure)
A
The 4 (GATE) study biases
P
E C
OT
Recruitment bias
Allocation bias
Maintenance bias
Outcome measurement bias
The GATE frame (design & bias)
P
E C
OT
Recruitment
Allocation
Maintenance
Blind or Objective measurement
Participants
Exposure
Comparison
OutcomesTime
The different study designs
illustrated with GATE
The GATE approach: every epidemiological study hangs on the GATE frame
There is only one basic study design:
• Cohort (& case-control) studies - aetiology / prognosis / intervention
• RCT (a randomised cohort study)- interventions
• Cross-sectional studies - diagnosis
Cohort (follow-up) study: archetypal epidemiological approach
Participants
Exposure Group Comparison Group
OutcomesTime
P
E C
OT
Allocated by measurement (not by randomisation)
Best design for investigating aetiology (risk), prognosis
Randomised controlled trial - cohort study where exposure allocated by randomisation process
Participants
Exposure Group Comparison Group
OutcomesTime
P
E C
OT
Allocated by randomisation
Best design for investigating treatments
Case series is a Cohort study with no comparison group
Participants
Exposure Group
OutcomesTime
P
E C
OT
Allocated by measurement
Before-after study
Participants
Exposure Group
Comparison Group
OutcomesTime
P
C
OT
Allocated by timing of intervention
E
Cross-over trial
Participants
Exposure Group 2
Comparison Group 2
OutcomesTime
P
C2
OT
Allocated by randomisation
E2
E1
C1
Exposure Group 1
Comparison Group 1
Cross-sectional study
Participants
Exposure Group Comparison Group
OutcomesTime
P
E C
OT
Allocated by measurement
real-life time
best design for prevalence and diagnostic test accuracy
Diagnostic test accuracy studyP
EG CG
O
Disease +ve(reference standard +ve)
Test
Disease –ve(reference standard –ve)
a bc d
Likelihood +ve test if D+ve:EGO = a ÷ EG
Likelihood +ve test if D -ve:CGO = a ÷ CG
+
-
EGO/CGO = +ve LR
Diagnostic test accuracy studyP
EG CG
OTest a b
c d
Likelihood -ve test if D+ve:EGO = c ÷ EG
Likelihood -ve test if D -ve:CGO = d ÷ CG
+
-
EGO/CGO = -ve LR
Disease +ve(reference standard +ve)
Disease –ve(reference standard –ve)
Diagnostic test for disease prediction
P
EG CG
O
Test +ve
Disease(reference standard)
Test -ve
a b
c d
Likelihood of D if test +ve:EGO = a ÷ EG
Likelihood of no D if test -ve CGO = d ÷ CG
+
-
Positive predictive value Negative predictive value
Diagnosis: test accuracy
EG CG
OT
est
a b
c d
+
-P
P
CG
EG
Disease + -
+
- + -
Diagnosis: test accuracy
EG CG
OT
est
a b
c d
+
-P
P
CG
EG
Disease + -
+
- + -
Diagnosis: disease prediction
Case control study
for investigating aetiology, interventions when outcomes rare
Exposed Not Exposed
Cases a bControls c d
Case control study
Exp. Not Exp.
Cases a bControls
Participants
Exp Group Comparison Group
OutcomesTime
P
EG CG
OT
cases
‘nested in a virtual cohort study’
a b
controls
eg cg
P
ComparisonE1
CE2E3Multiple Exposure
categories
Multiple Outcome categories
GATE: multiple categories
Participants
P
Continuous measure of Outcomes e.g. lipids
Olow
medium
high
high..med..low
Continuous measure of Exposure: e.g. body mass index
E
Correlation coefficient
GATE: continuous measurements
Participants
Life is a non-randomised trial
The 6 steps of EBP
A CAT documents the steps for a specific question
1. ASK questions relevant to your clinical problem using the PECOT (PICO) framework
2. ACCESS - search for epidemiological evidence to help answer your questions using the key PECOT terms
3. APPRAISE the evidence for its validity, effect size, precision)
4. AGGREGATE the aggregated (systematically reviewed) evidence with patient/community, clinical/health & policy issues & make an evidence-based decision
5. APPLY (implement) your decision6. [AUDIT your usual practice (i.e. compare your usual
practice for this clinical problem against ‘best’ evidence-based practice) – is there an evidence-practice gap?].
GATE-lite(simple1 page CATs)