Framework analysis
Framework analysis uses a matrix, a table with cases as rows and categories as columns, to compare qualitative accounts while keeping their source context.
Fit
When to use it, and when not to
Use it when
- You have an applied question and need systematic comparisons within and across cases.
- Your accounts cover enough shared topics for a matrix, while allowing unexpected categories.
Not when
- You need numerical population prevalence. Use a suitable survey. Read: Survey and questionnaire research
- You seek patterns without a case-by-category matrix. Thematic analysis may fit. Read: Thematic analysis
Before you start
What you need
- Cases, sampling and the research questionDefine whether a case is a person, site or another unit, and preserve that unit in the matrix.
- An initial and revisable coding frameworkCombine question-led categories with categories developed from the material where appropriate.
- Source links and analytic notesKeep transcript references with summaries so the team can revisit context and exceptions.
- Cases share topics relevant to the question.How to check: Check whether the matrix would force unlike accounts into the same categories.
- The framework can include unexpected material.How to check: Pilot it on varied cases and revise it before applying it across the dataset.
- Charting preserves the meaning of source accounts.How to check: Review summaries against full passages and discuss conflicting interpretations.
Plain language
How it works
Familiarize yourself with the accounts and code a varied sample. Develop a working framework, apply it across the material, and revise it when the data require.
Chart summaries into a matrix with cases as rows and categories as columns. Compare within a case and across cases. Link each summary to source passages before interpreting patterns.
Qualitative inquiry is not one of the five research areas this hierarchy covers. That describes the hierarchy, not the quality of the analysis. Gale and colleagues describe the procedure. The cited primary studies illustrate applied qualitative comparisons; a policy framework alone is not proof of using this analytic method.
Worked example
A worked example
- Setting
- A team interviews patients about access to follow-up care.
- Question
- How do transport barriers vary across cases?
- Approach
- Rows represent interviewees and columns represent categories. Each summary includes a transcript reference, with space for unexpected material.
- Result
- There are 4 of 10 cases with no material about transport, or 40 in 100 within this sample.
- What it means
- Those empty cells mean transport was not discussed or recorded. They do not show that those people had no transport barriers.
In practice
Recent studies PubMed search · not vetted
Watch out
Common mistakes
The one to remember
Charted summaries are interpretations, not replacements for transcripts. Keep links to source material, retain unexpected findings, and do not treat an empty cell as evidence that an experience was absent.
- Forcing every passage into the initial framework.Instead: Revise categories and retain material that challenges them.
- Comparing short summaries without returning to accounts.Instead: Keep source references and examine contradictory cases in context.
Inside Mallard
How Mallard uses this
Plans do not link to this method yet. Mallard does not code qualitative data or chart interviews into a framework matrix.
analysis.framework_analysis, the id a plan cue records.Related Learn pages
Related designs and methods
Evidence
Sources & further reading
- Using the framework method for the analysis of qualitative data in multi-disciplinary health research (opens in a new tab) · BMC Med Res Methodol · doi 10.1186/1471-2288-13-117
- "I tell you, getting data for this is hell"-Exploring the use of evidence for noncommunicable disease policies in Ghana (opens in a new tab) · PLOS Glob Public Health · doi 10.1371/journal.pgph.0002308
- Ethical dilemmas in prioritizing patients for scarce radiotherapy resources (opens in a new tab) · BMC Med Ethics · doi 10.1186/s12910-024-01005-3