Types of analysis
How you analyse your evidence will depend on the scope of your review question, the types of studies included, and the nature of the data extracted.
- Narrative analysis: a structured summary, comparing the data, and identifying themes to demonstrate evidence of an effect.
- Meta-analysis: pooling and analysing the results across several studies using statistical methods.
- Integrative analysis: synthesising diverse datasets using both narrative and statistical analysis. Use this for studies of varied study designs or mixed-methods approaches.
Narrative analysis techniques
Technique
|
Advantages
|
Limitations
|
Text description: consistently annotate individual studies with an organised summary of characteristics
|
Keeps details of each study together
|
Difficult to see themes
|
Tabulation: create a table of study elements and findings
|
Compares characteristics
|
Does not synthesise
|
Categorise: group or cluster studies around particular characteristic(s)
|
Visualise patterns
|
Difficult to see relationships across categories
|
Thematic analysis: analyse content with a rubric to identify themes and factors
|
Consistency
|
Time needed to develop a rubric
|
Describe relationships: identify relationships between primary study findings and characteristics to determine differences within and between studies.
|
Provides sub-analysis
|
Difficult with a large number of studies
|
(Booth, Papaioannau & Sutton, 2012)
Meta-analysis
A meta-analysis involves pooling and analysing the results across multiple studies using statistical methods.
You might conduct a meta-analysis as part of your systematic review to:
- increase the chance of detecting a statistically significant outcome or effect
- improve precision
- answer questions not posed by the individual studies
- settle controversies arising from apparently conflicting studies
- generate new hypotheses.
(Cochrane Handbook, Section 10.2 'Introduction to meta-analysis')
A meta-analysis should only be performed if you have a group of studies that are similar. The studies should be homogeneous in terms of participants, interventions and outcomes, so that the meta-analysis can draw a meaningful conclusion.
Types of systematic review that are not appropriate for meta-analysis include those based on:
- non-randomised studies
- studies with multiple outcomes
- studies with diverse methodological approaches.
In these instances, a systematic review can only be conducted using narrative analysis techniques.
Software for meta-analysis
- NVivo Qualitative data analysis software that can handle rich text-based information. It automates many manual tasks associated with analysis, like data classification and sorting.
- SPSS Software used for statistical meta-analysis. Available on campus desktop computers through the Citrix Workspace app.
- SAS Statistical software that allows you to perform data management, visualisation and prediction. You will need to accept the SAS Student Use Agreement online to obtain a copy of SAS.
- 'Practical Meta-analysis Effect Side Calculator' Calculates effect sizes for meta-analyses. This too can calculate the standardised mean difference, correlation coefficient, the odds-ration, and the risk ratio.