Upload
lisa-tompson
View
51
Download
0
Tags:
Embed Size (px)
Citation preview
Agenda for workshop
12.30 Introduction to EMMIE and rationale of approach (Shane Johnson)
12.45 Evidence appraised to date
13.00 The coding instrument
13.30 Exercise
14.30 The EMMIE narratives (and meta-synthesis)
15.00 Finish
Focus of reviews
326 reviews
~125 on single
intervention topics
~60 interventions
in total
+12 reviews in production
Reviews appraised so far
1. CCTV
2. Lighting
3. Multi-systemic therapy
4. Alcohol ignition interlocks
5. Sobriety checkpoints
6. CPTED (retail robbery)
7. Neighbourhood Watch
8. Music making interventions
9. Electronic monitoring
10. Increased police patrols
Experience of applying EMMIE
• Most reviews don’t use the language of EMMIE
– Different fields have very different reporting conventions
• The evidence is generally weak on effect, and often on other
dimensions
– But need to remember that reviews rely on primary study evidence
• BUT, weak evidence on effect doesn’t undermine other dimensions
– I.e. reviews can be strong on moderators or implementation
• Appraising quality is subjective, so we automated the scoring
Experience of applying EMMIE
Codebook has constantly been challenged and refined
• Effect: meta-analyses conducted in various ways
• Mechanism: presented (or not) in many different ways
• Moderators: working out a priori / post hoc can be challenging
• Implementation: teasing this out from MM tricky sometimes
• Economics: evidence rare on this
The coding instrument
• EMMIE-E - relates to the ‘evidence’ that emerges from reviews
• EMMIE-Q – relates to the ‘quality’ of it
• Both are needed for prospective users to gauge what is (not) known
and a level of confidence associated with findings
• Coding instrument appraises both of these for each dimension of
EMMIE
Exercise
• 4 groups
• Identify information on EFFECT and one other dimension
• 45 mins reading, discussing and annotating
• 15 minutes for group discussion
The EMMIE narratives (and meta-synthesis)
• Turning the coding into an accessible format
• Quality assurance
• The narratives
Summarising the effect
Rating Interpretation
X X Overall, evidence suggests an increase in crime
X X ! Overall, evidence suggests an increase in crime (but some
studies suggest a decrease)
X X Overall , no evidence to suggest an impact on crime (but some
studies suggest an increase)
X X No evidence to suggest an impact on crime
X X Overall, evidence suggests no impact on crime (but some
studies suggest either an increase or a decrease)
X X Overall, evidence suggests no impact on crime (but some
studies suggest a decrease)
X ! Overall, evidence suggests a decrease in crime (but some
studies suggest an increase)
X X Overall, evidence suggests a decrease in crime
Summarising EFFECT-Q
Star rating for Effect Q Associated text
★★★★ The review was sufficiently systematic that
most forms of bias that could influence the
study conclusions can be ruled out.
★★★★ The review was sufficiently systematic that
many forms of bias that could influence the
study conclusions can be ruled out.
★★★★ Although the review was systematic, some
forms of bias that could influence the study
conclusions remain.
★★★★ Although the review was systematic, many
forms of bias that could influence the study
conclusions remain.
★★★★ Text to reflect specific review
Star rating for
Moderator Q
Associated text
★★★★ Collection and analysis of relevant data relating
to theoretically grounded moderators and
contexts
★★★★ Theoretically grounded description of relevant
contextual conditions
★★★★ Tests of the effects of contextual conditions
defined post hoc using variables that are at
hand
★★★★ Ad hoc description of possible relevant
contextual conditions
★★★★ No reference to relevant contextual conditions
that may be necessary
Summarising MODERATOR-Q
Meta-synthesis
• Synthesis methods needed for integrating reviews
• Two overarching decision rules for narratives:
1. For each EMMIE element use the HIGHEST QUALITY (highest Q score) scoring
review to populate the EMMIE-E and EMMIE-Q scores
2. For each piece of information make it clear which source is being referred to
• Use sub-group analysis FROM ALL REVIEWS to work out the ‘inside’
cross and tick
– i.e. evidence of statistical reduction or backfire under certain conditions)
– Use the HIGHEST QUALITY meta-analysis to populate the overall effect