Written and maintained by CASRAI Editorial Board
Last updated
An integrative review is a research synthesis method that combines findings from studies with diverse methodologies—experimental, quasi-experimental, non-experimental, qualitative, and theoretical—within a single review, rather than restricting inclusion to one study design. Whittemore and Knafl’s 2005 paper, “The integrative review: updated methodology” (Journal of Advanced Nursing, 52(5), 546–553, doi:10.1111/j.1365-2648.2005.03621.x), is the methodology most cited for doing this rigorously. It sets out five stages—problem identification, literature search, data evaluation, data analysis, and presentation—and was written specifically to address the rigor gaps in earlier integrative-review guidance, where the combination of experimental and non-experimental sources often meant unclear search strategies, no explicit quality evaluation, and superficial, narrative-only analysis.
An integrative review is not the same thing as a systematic review, a narrative review, or a scoping review—see CASRAI’s side-by-side comparison for the definitional distinctions between those three. This guide assumes an integrative review is the right tool for the question and walks through Whittemore and Knafl’s five stages in the order they’re meant to be applied.
When an integrative review is the right choice
Choose an integrative review when the goal is to synthesize a body of evidence that is genuinely heterogeneous in design—for example, combining randomized trials, observational studies, qualitative interview studies, and conceptual or theoretical papers on the same phenomenon. That heterogeneity is the method’s defining feature: a systematic review and meta-analysis typically pool comparable quantitative study designs against a single risk-of-bias framework, and a narrative review summarizes a literature without a predefined, reproducible search-and-appraisal protocol. An integrative review sits between them: it applies a structured, reportable process (unlike a narrative review) while still accommodating the mixed empirical-and-theoretical evidence base that a single-design systematic review’s inclusion criteria would exclude.
It’s a common choice in nursing and health-services research—the discipline the method originates from—particularly for questions about complex interventions, patient experience, or emerging concepts where the evidence base spans study types by necessity, not by a reviewer’s convenience. If the underlying studies are methodologically comparable enough to pool statistically, a meta-analysis is usually the stronger choice; if the goal is instead to map the extent and nature of a broad, emerging topic rather than answer a focused question, see CASRAI’s scoping review guide.
The five stages at a glance
| Stage | Core task |
|---|---|
| 1. Problem identification | Define the problem, purpose, and the variables or concepts of interest clearly enough to bound the review. |
| 2. Literature search | Run a comprehensive, documented multi-source search across the full range of relevant study designs. |
| 3. Data evaluation | Appraise quality across methodologically diverse sources without forcing them through one appraisal instrument. |
| 4. Data analysis | Reduce, display, and compare extracted data across studies to identify patterns, themes, and relationships. |
| 5. Presentation | Report the synthesis—and the method itself—with enough detail that another reviewer could reproduce it. |
Stage 1: Problem identification
The first stage defines the problem and purpose of the review clearly enough to determine what counts as relevant evidence. Whittemore and Knafl treat a vague or overly broad problem statement as one of the main sources of downstream rigor failures in integrative reviews: if the variables, concepts, and population of interest aren’t specified up front, the search strategy in Stage 2 has no real boundary, and inclusion decisions in Stage 3 end up made ad hoc rather than against a stated criterion. In practice, this stage produces the same output a systematic review protocol would—a defined question, the key variables or concepts, and the population or context—even though the review itself won’t be limited to a single study design.
Stage 2: Literature search
The search stage aims for a comprehensive, reproducible search across the full range of study designs the problem statement calls for—typically multiple bibliographic databases, hand-searching of key journals, and ancestry/citation searching from included studies, plus a deliberate decision about whether grey literature (theses, conference proceedings, unpublished reports) falls within scope. Because an integrative review’s sampling frame spans experimental, non-experimental, and theoretical literature, a single database or a single set of search terms tuned to one study design will systematically miss whole categories of eligible evidence. The search strategy, databases, terms, and date range should be documented in enough detail to report in the final write-up (Stage 5) and, ideally, to replicate.
Stage 3: Data evaluation
Data evaluation is where an integrative review most visibly departs from a single-design systematic review: there is no single validated risk-of-bias tool that fairly appraises a randomized trial, a qualitative interview study, and a theoretical paper on the same scale. Whittemore and Knafl’s methodology doesn’t prescribe one universal instrument; instead it calls for quality-evaluation criteria appropriate to each source type—often a separate scoring or appraisal approach for empirical versus theoretical sources—applied consistently and documented so a reader can see how quality was weighted in the eventual synthesis. Skipping this stage, or applying only a light editorial judgment instead of an explicit criterion, is one of the more common ways integrative reviews lose credibility relative to a systematic review’s more standardized appraisal process.
Stage 4: Data analysis
The analysis stage is the most methodologically distinctive part of Whittemore and Knafl’s framework. It adapts a constant-comparison approach—a technique with roots in qualitative analysis—to a mixed empirical-and-theoretical evidence base, working through four subprocesses:
- Data reduction—extracting and classifying data from primary sources into a manageable, comparable format (a matrix or coding scheme is typical).
- Data display—organizing the reduced data visually (tables, graphs, or matrices) so patterns across studies of different designs become visible side by side.
- Data comparison—iteratively comparing the displayed data to identify patterns, themes, relationships, and discrepancies across the full set of included sources.
- Conclusion drawing and verification—drawing conclusions from the patterns identified, then checking them back against the primary sources for accuracy before they’re reported.
Because the underlying sources aren’t methodologically uniform, this analysis is necessarily more interpretive than a meta-analysis’s statistical pooling—which is exactly why Stages 1–3 exist to constrain that interpretation with a defined problem, a documented search, and an explicit quality evaluation, rather than leaving the synthesis to unstructured narrative judgment.
Stage 5: Presentation
The final stage reports both the synthesis findings and the method itself—the search strategy, inclusion/exclusion criteria, quality-evaluation approach, and analysis process—with enough transparency that another reviewer could assess or reproduce it. Whittemore and Knafl recommend presenting the evidence through narrative summary supported by tables, models, or diagrams that make the pattern of evidence across study designs visible, rather than a purely prose-only summary. Because integrative reviews don’t have a dedicated PRISMA extension the way scoping reviews have PRISMA-ScR, reporting quality varies more across published integrative reviews than across published systematic reviews—which is part of why the search and evaluation stages need to be documented explicitly rather than summarized in a sentence.
Common methodological criticisms—and how the five stages address them
Whittemore and Knafl’s 2005 paper was itself a direct response to earlier integrative reviews being criticized for a specific, recurring set of problems: imprecise problem statements, unclear or unsystematic search strategies, no explicit data-evaluation step, and analysis that amounted to little more than narrative impression rather than a documented comparison process. Each of the five stages maps directly onto one of those criticisms—which is useful to keep in mind while planning a review, since a reviewer or editor evaluating the finished manuscript is likely to be checking for exactly these gaps. A review that skips or compresses Stage 3 (data evaluation) is the single most common weakness reviewers flag, since it’s the stage with no off-the-shelf checklist to lean on.
Frequently asked questions
What is the difference between an integrative review and a systematic review?
A systematic review typically restricts inclusion to a single study design (often randomized trials) and appraises quality against one standardized risk-of-bias tool, following a PRISMA-reportable protocol. An integrative review deliberately includes methodologically diverse sources—experimental, non-experimental, qualitative, and theoretical—in the same synthesis, which requires the more flexible, source-appropriate quality-evaluation and constant-comparison analysis Whittemore and Knafl’s method sets out, rather than a single appraisal instrument.
Can an integrative review combine qualitative and quantitative studies in one synthesis?
Yes—combining qualitative, quantitative, and theoretical sources in a single synthesis is the defining purpose of the method, not an edge case. That’s also why Stage 3 (data evaluation) and Stage 4 (data analysis) are the most demanding parts of the process: they have to accommodate sources that can’t be appraised or pooled on the same scale.
Does an integrative review need a PRISMA flow diagram?
There is no integrative-review-specific PRISMA extension the way scoping reviews have PRISMA-ScR. Many published integrative reviews still include a PRISMA flow diagram to report the search-and-selection process transparently, since it’s a well-understood reporting convention, but it isn’t a required or standardized element of Whittemore and Knafl’s methodology the way it is for a systematic review.
Can an integrative review be registered on PROSPERO?
PROSPERO’s scope is systematic reviews with a health-related outcome, and it does not accept integrative-review protocols. A protocol can still be published ahead of the review as a standalone paper, or registered on a general-purpose platform such as the Open Science Framework (OSF), if establishing a public, time-stamped protocol is a priority.
How is data quality evaluated when the included studies use completely different methods?
Whittemore and Knafl’s Stage 3 does not use one universal appraisal tool. Instead, the reviewer applies quality criteria suited to each source type—commonly a separate scoring or evaluation approach for empirical (quantitative and qualitative) sources versus theoretical or conceptual sources—and documents how that evaluation fed into how much weight each source was given in the Stage 4 synthesis.
Is the integrative review method specific to nursing research?
The method originates in nursing science, where Whittemore and Knafl’s 2005 paper remains the most cited methodological reference, and it’s still most common there and in adjacent health-services research. Nothing about the five-stage structure is nursing-specific, though, and the method is used in other fields that need to synthesize a genuinely mixed empirical-and-theoretical evidence base on a single question.
Further reading: How to Conduct a Scoping Review and Narrative Synthesis and SWiM cover two other approaches to combining evidence that a single risk-of-bias framework or meta-analysis can’t handle. See also CASRAI’s scholarly publishing hub for the site’s broader coverage of review and evidence-synthesis methodology.
