Guides: Meta-Analysis of Prevalence

Single-Group Proportions Pooled Prevalence 60 minutes
Guides Objectives:
By the end of this session, users will be able to:
  • Understand the purpose and application of meta-analysis of prevalence (single-group proportions)
  • Configure prevalence meta-analysis in RevProHub with proper field mapping (sample size, case count)
  • Select appropriate model types (random-effects vs. fixed-effect) for proportion data
  • Interpret forest plots, heterogeneity statistics (I², Q-test), and pooled prevalence estimates
  • Perform subgroup analysis and meta-regression to explore sources of heterogeneity
  • Assess publication bias using funnel plots and Egger's test
  • Use sensitivity analysis to compare results across different methodological choices
  • Export results and add AI-generated interpretations to manuscript

Part 1: What is Meta-Analysis of Prevalence?

Meta-analysis of prevalence (also called single-group proportion meta-analysis) synthesizes data from studies reporting the frequency of a condition, event, or characteristic within a population when there is no comparison group.


Sample Size (N)
Total users per study

Case Count (n)
Number with the condition

Pooled Prevalence
Weighted average proportion

Heterogeneity I²
Variation between studies
📖 Common Applications: Disease prevalence (e.g., hypertension, diabetes, HIV), medication adherence rates, healthcare utilization proportions, diagnostic test accuracy (sensitivity/specificity with single-group data).

Part 2: Configuring Meta-Analysis of Prevalence

Navigate to: Reporting Phase → Data Synthesis → Configure Analysis → Prevalence Meta-Analysis

Step 1: Basic Information
- Analysis Name: e.g., "Hypertension Prevalence in HIV-Infected Adults"
- Select Data Extraction Form: Choose form containing sample_size and case_count fields
Step 2: Select Studies
- Choose studies marked as "Included" after full-text screening
- Only studies with completed prevalence data (N and n) appear
Step 3: Data Mapping & Settings
Map the following fields from your extraction form:
  • Total Sample Size (N): Number of users in the study
  • Number of Cases (n): Number with the outcome/condition
  • Model Type: Random-Effects (recommended for heterogeneous data) or Fixed-Effect
📊 Effect Measure for Prevalence: RevProHub uses the double-arcsine transformation (Freeman-Tukey) to stabilize variance for proportion data, then back-transforms for reporting.

Part 3: Example Data – Hypertension Prevalence in HIV-Infected Adults (SSA)

StudyCountrySettingSample Size (N)Cases (n)Prevalence (%)
Okoro C. (2021)NigeriaUrban45012026.7%
Uwamahoro D. (2021)RwandaMixed150077351.5%
Mbemba K. (2023)DRCMixed81018522.8%
Molefi S. (2022)BotswanaUrban3109530.6%
Abebe G. (2023)EthiopiaRural85530035.1%
Onyango P. (2021)KenyaRural2105827.6%
Mwamba Z. (2023)TanzaniaRural41022554.9%

Part 4: Forest Plot – Pooled Prevalence

Forest Plot: Hypertension Prevalence in HIV-Infected Adults (Subgroup by Setting) 
Study                              Prevalence [95% CI]     Weight%
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Mixed/Nationwide (k=3)             32.7% [16.5%, 54.6%]     33.1%
Urban Setting (k=9)                35.3% [31.2%, 39.6%]     43.9%
Rural Setting (k=3)                38.8% [25.3%, 54.3%]     23.0%
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
TOTAL (random)                     35.4% [30.7%, 40.4%]     100.0%
Model Summary: Random-Effects | Heterogeneity: Q=358.47 (df=14), p<0.001 | I² = 96.1%
📉 Interpreting Heterogeneity (I²): I² = 96.1% indicates considerable heterogeneity. This means 96.1% of observed variance is due to true differences between studies, not sampling error. Subgroup analysis or meta-regression should be conducted to identify moderators.

Part 5: Subgroup Analysis & Meta-Regression

Subgroup Analysis
By country, setting (urban/rural), study design, or population characteristics
Meta-Regression
Explore continuous moderators (publication year, sample size, mean age)
Publication Bias
Egger's test + funnel plot visual inspection
Subgroup Analysis Results (by Setting):
  • Mixed/Nationwide (k=3): Prevalence 32.7% [16.5%, 54.6%], I²=99.0%
  • Urban Setting (k=9): Prevalence 35.3% [31.2%, 39.6%], I²=90.7%
  • Rural Setting (k=3): Prevalence 38.8% [25.3%, 54.3%], I²=96.6%
Test for subgroup differences: Q = 0.28, df = 2, p = 0.869 (no significant difference between subgroups)

Part 6: Publication Bias – Funnel Plot & Egger's Test

Egger's Test Results: Intercept = -14.804, t = -3.88, p = 0.002 (df=13)
Interpretation: A significant Egger's test (p < 0.05) suggests funnel plot asymmetry, which may indicate publication bias, small-study effects, or true heterogeneity. Visually inspect the funnel plot – missing studies in the bottom-left corner would suggest selective reporting of positive results.

Part 7: AI Assistant – Interpretation & Methodological Feedback

Critical Interpretation:
The meta-analysis, based on 15 studies, estimates a pooled prevalence of 35.4% (95% CI: 30.7% to 40.4%). This suggests that approximately one-third of the population under study exhibits the condition of interest. However, the interpretation is severely undermined by the extreme level of statistical heterogeneity (I² = 96.1%). An I² of this magnitude indicates that 96.1% of the observed variance is attributable to true differences between studies rather than sampling error. Consequently, the 'average' prevalence of 35.4% should be viewed with extreme caution.
Methodological Feedback:
The extreme I² (96.1%) suggests significant clinical or methodological diversity that should be addressed through subgroup analyses or meta-regression to identify potential moderators. I recommend: (1) Reporting the Prediction Interval (PI) to illustrate the range for future studies; (2) Conducting leave-one-out sensitivity analysis; (3) Performing funnel plot analysis and Egger's test to assess publication bias.

Part 8: Sensitivity Analysis

Sensitivity Analysis Options:
  • Changed Zero-Cell Event Handling Method
  • Leave-one-out Analysis (exclude each study sequentially)
  • Changed Statistical Model (Fixed vs. Random effects)
  • Excluded Low-Quality Studies (based on QA scores)
  • Excluded Outlier Studies (using influence diagnostics)
  • Changed Effect Measure (proportion vs. logit vs. arcsine)
Comparison View: RevProHub displays base vs. sensitivity values side-by-side for pooled effect, CI, I², and Egger's p-value.

Part 9: Export Results & Add to Manuscript

Export as PDF
Full report with forest plot, tables
Export as DOCX
Editable for manuscript integration
Export as CSV
Raw data for external analysis
Add to Manuscript
Auto-inject into Manuscript Editor
🧪 Group Hands-On Activity: Prevalence Meta-Analysis (50 min)
  1. Access Data Synthesis: Reporting Phase → Data Synthesis → Configure Analysis.
  2. Select Analysis Type: Choose "Prevalence Meta-Analysis".
  3. Name Analysis: "Hypertension Prevalence in HIV-Infected Adults - Training".
  4. Select Extraction Form: Choose form with sample_size and case_count fields.
  5. Select Studies: Choose 10-15 included studies with completed prevalence data.
  6. Map Fields: Map Sample Size (N) → sample_size; Cases (n) → case_count.
  7. Set Model: Random-Effects (recommended for heterogeneous data).
  8. Run Analysis: Click "Run Analysis". Review forest plot, pooled prevalence, and I² heterogeneity.
  9. Add Subgroup Analysis: Clone the analysis → add subgroup by "country" or "setting" field. Compare subgroup prevalence and I² values.
  10. Meta-Regression: Run analysis with moderator "publication_year". Interpret p-value and R².
  11. Publication Bias: View funnel plot and interpret Egger's test p-value.
  12. Export & Add to Manuscript: Export results as PDF. Use "Add to Manuscript" to inject into manuscript.

Debrief Questions:

  • What is the difference between meta-analysis of prevalence and meta-analysis of effectiveness?
  • How do you interpret an I² of 90% or higher in a prevalence meta-analysis?
  • What transformation does RevProHub use for proportion data, and why?
  • How can subgroup analysis help explain high heterogeneity?
  • What does a significant Egger's test indicate about your prevalence meta-analysis?

Best Practices for Prevalence Meta-Analysis

Recommendations:
  • Use random-effects model when I² > 50% (heterogeneity expected in prevalence studies)
  • Report the Prediction Interval (PI) alongside CI to show the range for future studies
  • Perform sensitivity analysis (leave-one-out) to identify influential studies
  • Use double-arcsine transformation for better variance stabilization with proportions near 0 or 1
  • Assess publication bias using funnel plots AND Egger's test (not just visual inspection)
  • Report subgroup analyses even if not statistically significant – transparency about heterogeneity sources
Prevalence Meta-Analysis Completion Checklist:
✔ Analysis configured with correct field mapping (N and n)
✔ Random-effects model selected (for heterogeneous data)
✔ Forest plot reviewed with pooled prevalence and CIs
✔ Heterogeneity assessed (I², Q-test)
✔ Subgroup analysis performed (if applicable)
✔ Meta-regression explored (if applicable)
✔ Publication bias assessed (funnel plot + Egger's test)
✔ Results exported and/or added to manuscript
✔ AI interpretation reviewed and incorporated

Next: Meta-Analysis of Effectiveness (2x2 tables) for binary outcomes.

2026 RevProHub
Guides: Meta-Analysis of Prevalence | Single-Group Proportions | Forest Plot | Heterogeneity | Publication Bias | AI Interpretation