Integrative Review: synthesising across study designs

A systematic review of effectiveness admits comparative designs; a meta-synthesis admits qualitative studies. An integrative review admits both, plus theoretical literature — which makes it the standard choice in nursing and practice disciplines for questions about concepts (what is “care rationing”?), practices with mixed evidence bases, and fields where the theory papers carry as much weight as the data.

The cost of admission is analytic: every included source must be reduced to comparable units before anything can be integrated. That reduction step — not the searching — is what distinguishes strong integrative reviews from weak ones.

The method’s modern form comes from Whittemore and Knafl’s The integrative review: updated methodology, which fixed the five stages — problem identification, literature search, data evaluation, data analysis, presentation — and, crucially, imported constant-comparison techniques into the analysis stage.

Their data-analysis prescription is specific: data reduction (extract and code each source into a matrix), data display, data comparison, and conclusion drawing with verification — a sequence adapted from qualitative analysis precisely because it can absorb heterogeneous inputs.

For evaluating wildly different source types, the framework’s advice is pragmatic: rate methodological quality within source type (a trial against trial standards, a qualitative study against qualitative standards, a theory paper on argumentative coherence) and carry the ratings into the synthesis rather than using them as gates.

A review team asked what is known about “missed nursing care” — a concept spanning measurement studies, qualitative accounts, and theory papers. The search returned 54 includable sources: 31 quantitative (mostly survey-based prevalence and predictor studies), 17 qualitative, 6 theoretical.

Each source was appraised within genre and reduced into a matrix coded for definition used, antecedents, the care most often missed, and consequences. Comparison across genres produced the review’s two real findings. First, the surveys and the qualitative studies systematically disagreed on what gets missed most — surveys said ambulation and oral care; nurses’ accounts said emotional support and teaching, which the dominant survey instrument barely measures. Second, all six theory papers located the cause in unit-level resourcing, while the qualitative accounts kept describing implicit rationing decisions by individual nurses — a mechanism the theories had no slot for.

The review concluded with a revised conceptual model and a measurement critique, neither of which could have come from any single-genre review of the same literature.

• Parallel summaries instead of integration. A quantitative section plus a qualitative section is two mini-reviews, not an integrative one.
• One appraisal rubric for every genre. A theory paper scored on sample size fails a test it never sat.
• No matrix. Without a common reduction structure, comparison across 54 heterogeneous sources happens in the reviewer’s head, unverifiably.
• Concept drift. When the included sources define the concept differently, the review must analyse that variation — adopting one definition silently discards half the evidence.
• Overclaiming effect sizes. Integrative reviews describe and theorise; precise effect claims belong to the meta-analysis the design deliberately is not.

The method’s breadth multiplies its subjectivity surfaces: what to admit, how to appraise across genres, how to weight a strong ethnography against a weak trial. The matrix and the within-genre appraisal bound the discretion; they do not remove it.

Integrative conclusions are correspondingly soft: patterns, models, and gaps rather than pooled estimates. Commissioners wanting a number should be pointed to a different review type before the work starts, not after.

And the analysis stage does not scale gracefully — a matrix over 150 sources across three genres is months of reduction work, which is why integrative reviews suit concepts and practices more than sprawling intervention literatures.

A systematic literature review is the stricter, single-question sibling; prefer it when the question is effectiveness.

Meta-synthesis handles the qualitative-only case with interpretive translation rather than matrix reduction.

A scoping review maps the same broad territory without attempting integration — the right stop when the field is too immature to synthesise.

The matrix is the method’s engine, and filling it is codebook work: defined dimensions applied across dozens of heterogeneous PDFs. Evidano applies those dimensions as a codebook and returns each cell with its source quote, so genre-spanning comparison starts from verified extractions instead of memory — and the cross-genre disagreements that make integrative reviews worth doing become visible as side-by-side coded evidence.

What software cannot do is the genre-sensitive appraisal or the model-building; those are the reviewer’s contribution, and the matrix only makes them defensible.