Week 6 Notes – Evidence Synthesis & Evidence Tables
Introduction to Evidence Synthesis
Purpose & Value
Compiles findings from multiple primary studies to build a coherent picture of a phenomenon.
Detects overall trends, reveals consistencies / discrepancies, and directs future research agendas.
Strengthens external validity by moving beyond single‐study idiosyncrasies.
Typical Workflow (recap of previous weeks)
Formulate focused question → Systematic search & screening → Critical appraisal → Synthesis (this week’s focus) → Interpretation & reporting.
Two Core Approaches
Narrative (qualitative) synthesis – text-based integration of heterogeneous evidence.
Systematic (quantitative) synthesis / meta-analysis – statistical pooling of comparable effect sizes.
Choice depends on data homogeneity, research objective, and available expertise.
Narrative Synthesis
Definition: Qualitative strategy that describes and interprets findings across diverse designs, measures, and contexts.
When Preferred: High methodological diversity, scarce quantitative data, exploratory questions.
Core Techniques
Thematic Analysis
Code recurring concepts, group them into overarching themes.
Example: Interventions for university anxiety → themes of coping strategies, social support, therapeutic modality effectiveness.
Tabulation
Build matrices to juxtapose study characteristics & outcomes.
Enables side-by-side comparison (e.g., population, intervention type, outcome score).
Textual Description
Rich, critical description of each study plus cross-study linkages.
Example: Contrast varying definitions of resilience and discuss how operationalisation affects reported outcomes.
Structuring a Narrative Review
Introduce scope → Present thematic clusters → Contrast findings → Identify gaps → Conclude with implications.
Integrate critical appraisal rather than listing studies uncritically.
Real-World Relevance: Often underpins practice guidelines when statistical pooling is impossible (e.g., early evidence on emerging therapies).
Quantitative Synthesis: Meta-Analysis
Goal: Produce a pooled effect size offering greater precision than any individual study.
Key Steps
Effect-Size Calculation
Convert disparate results to a common metric (e.g., Standardised Mean Difference, d; Odds Ratio, OR).
Formula example: \text{SMD}=\frac{\bar X{\text{treat}}-\bar X{\text{control}}}{SD_{\text{pooled}}}.
CBT example: 5-point drop on 20-pt scale vs 0.8-pt drop on 5-pt scale → compute comparable d values.
Statistical Integration
Choose fixed-effect (assumes common true effect) or random-effects (allows effect variance).
Weight each study (inverse of variance); compute pooled estimate & 95 % CI.
Heterogeneity Assessment
Cochran’s Q and I^{2} statistics quantify dispersion.
I^{2}=\frac{Q-df}{Q}\times100\% (values > 50 % ≈ substantial heterogeneity).Address via sub-group, meta-regression, or sensitivity analyses (e.g., age groups, baseline severity).
Quality Safeguards
Ensure comprehensive search, consistent coding, duplicate extraction, publication-bias tests (funnel plot, Egger).
Ethical Implication: Mis-pooled biased studies can misguide treatment guidelines; transparency is obligatory.
Mixed-Methods Synthesis
Concept: Integrates statistical generalisability of quantitative data with the contextual richness of qualitative findings.
Advantages
Validates quantitative trends with participant narratives; illuminates mechanisms.
Typical Challenges & Mitigations
Data Integration Complexity
Design a priori plan (convergent, sequential, or embedded synthesis design).
Example: Merge RCT anxiety score reductions with interview-based perceptions of CBT usefulness.
Consistency in Interpretation
Use joint-display matrices to align qualitative themes with quantitative outcomes.
Ensure narratives support rather than contradict statistical trends (or explain divergence).
Developing Synthesis Tables
Purpose: Condense complex multi-study information into a clear, accessible snapshot.
Construction Principles
Organise Information
Essential columns: Author/Year | Design | Sample (N, characteristics) | Intervention/Exposure | Outcome Measures | Effect Size / Key Findings | Quality Rating.
Example entry for CBT study: Jones et al., 2022 | RCT | N=120 undergraduates | 8-week group CBT | Beck Anxiety Inventory | d=-0.65 | Low risk of bias.
Visual Presentation
Bold headers, uniform fonts, shading/colour for subgroup distinctions.
Keep row heights aligned; wrap text judiciously.
Data Summarisation
Highlight repeating findings (e.g., majority show 10 %–20 % reduction).
Include footnotes for anomalies or methodological caveats.
Toolbox
Microsoft Excel / Google Sheets → Sorting, filtering, conditional formatting.
Reference-management plug-ins (e.g., Zotero → CSV export) accelerate population.
Critical Evaluation of Synthesised Evidence
Quality Assessment
CASP, Joanna Briggs, Cochrane Risk of Bias 2; record domains such as randomisation, blinding, attrition.
Applicability / External Validity
Compare study populations with target context; note cultural, age, or clinical severity boundaries.
Biases & Limitations
Publication bias (file-drawer effect) may inflate pooled effect; explore with funnel plots, trim-and-fill.
Language bias, selective outcome reporting, researcher allegiance.
Transparent Reporting: Follow PRISMA or ENTREQ checklists; supply supplementary tables & code.
Lecture 6.2 – Writing an Evidence Table
Introduction & Benefits
Converts narrative or quantitative synthesis into an immediately scan-able format.
Promotes clarity, organisation, and analytic rigor.
Enables rapid identification of patterns, themes, and outliers.
Designing the Table
Step 1 – Layout
Decide minimal yet sufficient columns (avoid clutter).
Step 2 – Software Choice
Choose tools offering sorting & conditional colour coding.
Automate bibliographic fields via citation-manager exports when possible.
Populating the Table
Efficient Extraction Workflow
Skim article → mark candidate data.
Cross-verify with co-reviewer.
Enter into table; use data-validation lists to maintain consistency (e.g., predefined design categories).
Consistency & Accuracy
Explicitly record scale units (5-pt vs 20-pt).
Double-data extraction to reduce transcription errors.
Using Tables for Analysis
Identify Patterns → e.g., majority of CBT studies report d\approx-0.60 (moderate effect).
Spot Discrepancies → Investigate outlier study with negligible effect; maybe short intervention or different anxiety subtype.
Generate Hypotheses → Effect stronger in group formats than individual sessions?
Prepare for Meta-Analysis → Table becomes source file for statistical software (RevMan, R \textit{meta} package).
Examples & Best Practices
Study published meta-analyses for formatting cues (e.g., Journal of Consulting & Clinical Psychology).
Use footnotes to define abbreviations; maintain consistent decimal places for effect sizes.
Employ reader-centric design – concise wording, white space, visual anchors.
Common Pitfalls
Overly dense tables (information overload).
Inconsistent terminology ("pre-test" vs "baseline").
Neglecting to update table when re-analysing or excluding a study.
Ethical, Philosophical & Practical Considerations
Ethical: Accurate synthesis prevents clinician misguidance; misreporting can harm patients.
Philosophical: Reflects epistemological humility – recognising knowledge is provisional and context-bound.
Practical: High-quality evidence tables speed up grant writing, guideline development, and replication planning.
Connections to Prior Course Content
Builds on Week 4’s systematic search strategies and Week 5’s critical appraisal skills.
Draws statistical concepts from earlier modules (confidence intervals, variance, weighting).
Reinforces research-integrity principles introduced in Week 1 (transparency, reproducibility, FAIR data).
Conclusion & Next Steps
Mastery of narrative, quantitative, and mixed-methods synthesis enables psychologists to craft robust, transparent, and actionable evidence reviews.
Evidence tables translate those syntheses into user-friendly artefacts for clinicians, policymakers, and researchers.
Practice Task: Create a synthesis & evidence table for 5 recent RCTs on mindfulness-based stress reduction; evaluate heterogeneity and draft a narrative interpretation.
Continue refining skills using PRISMA flowcharts, advanced meta-analytic software, and peer de-briefing for bias checking.