A guide to mixed-methods systematic reviews: how to combine quantitative and qualitative evidence, the three integration designs, appraisal with MMAT, and common pitfalls.
Some questions have two halves. Does this intervention work, and why do people abandon it? Is the program effective, and how do participants actually experience it? Answer only the first and you get a number without an explanation. Answer only the second and you get an explanation without knowing whether the thing works at all.
A mixed-methods systematic review answers both. It brings quantitative and qualitative evidence into one review and, crucially, integrates them rather than simply parking them side by side. That integration is the whole method, and it is also where most mixed-methods reviews fall short.
A mixed-methods systematic review synthesizes both quantitative and qualitative evidence to answer a question that needs both. The quantitative strand tells you what the effect is; the qualitative strand tells you how and why.
The defining feature is integration: the two strands must be brought together to produce insights neither could yield alone. A review that reports quantitative findings, then qualitative findings, and never connects them is not a mixed-methods review. It is two reviews stapled together.
The test is whether your question genuinely has two components that depend on each other.
Good candidates include: an intervention whose effectiveness varies widely across settings, where the qualitative evidence might explain why. A program with strong trial results and poor real-world uptake, where the qualitative evidence explains the gap. A complex intervention where "does it work" is meaningless without "under what conditions and how."
If your question is purely about effect size, you want a standard systematic review with a meta-analysis. If it is purely about experience, you want a qualitative synthesis. The mixed-methods review is for questions where the two illuminate each other, and it costs considerably more effort, so it should be a deliberate choice rather than a default.
How you combine the strands is the central design decision, and there are three established approaches.
Convergent (parallel) design. The quantitative and qualitative strands are searched, appraised, and synthesized separately, then integrated at the end. The two syntheses are brought together and compared, looking for agreement, disagreement, and complementarity. This is the most common approach and the most straightforward to execute.
Sequential design. One strand is synthesized first, and its findings shape the second. For example, a quantitative synthesis identifies that an intervention's effect varies by setting, and a subsequent qualitative synthesis explores why. Or a qualitative synthesis surfaces the outcomes participants actually care about, which then guide the quantitative synthesis. The strands inform each other in order.
Convergent integrated design. The two types of data are transformed so they can be synthesized together. This usually means "qualitizing" quantitative data (converting numerical findings into narrative statements) so both strands can be synthesized as text. JBI describes this approach in its guidance. It is the most genuinely integrated of the three and also the most demanding.
Pick one and state it explicitly. A review that has not decided on its integration design usually ends up with the two strands never quite meeting.
| Design | How it works | Best for |
|---|---|---|
| Convergent (parallel) | Synthesize separately, integrate at the end | Most reviews; simplest to execute |
| Sequential | One strand informs the other | When findings from one should shape the other |
| Convergent integrated | Transform data so both are synthesized together | Deepest integration; most demanding |
The method follows a systematic review, with adjustments at each stage to accommodate two evidence types.
Frame the question so both components are explicit. You will often need a framework for each strand, since PICO suits the quantitative half and something like PICo or SPIDER suits the qualitative half.
Write and register a protocol, stating your integration design upfront. Deciding how to integrate after you have seen the findings is the mixed-methods equivalent of setting criteria after searching.
Search for both types of evidence. This is harder than a single-strand search, because qualitative studies are less consistently indexed and the search strategies for each strand differ. Expect to run more than one search.
Screen and select against criteria that accommodate both study types, in duplicate.
Appraise each study with a tool matched to its design. The Mixed Methods Appraisal Tool (MMAT) is designed exactly for this, allowing appraisal of qualitative, quantitative, and mixed-methods studies within one framework. Alternatively, use design-specific tools for each strand.
Synthesize each strand with an appropriate method: statistical or narrative synthesis for the quantitative evidence, and a qualitative synthesis approach for the qualitative evidence.
Integrate. This is the step that defines the review. Bring the strands together according to your chosen design, and interrogate the relationship between them. Where do they agree? Where does one contradict the other? Where does the qualitative evidence explain a pattern in the quantitative results?
Report. State your integration design clearly, present both strands, and make the integration visible rather than implied.
This deserves emphasis because it is the single most common weakness in published mixed-methods reviews.
It is easy to produce a document with a quantitative section and a qualitative section, a conclusion that mentions both, and no actual integration anywhere. The reader learns what the trials found and what the interview studies found, but nothing that required combining them. That is not a mixed-methods synthesis. It is a collection.
Real integration produces claims that neither strand could support alone. The trials show the intervention works on average but with high variability; the qualitative evidence shows that implementation quality differs sharply between settings; together they suggest the variability is an implementation problem rather than an efficacy one. That conclusion belongs to the integration, not to either strand.
Ask yourself, of your final review: what does this say that a separate quantitative review and a separate qualitative review would not have said? If the honest answer is "nothing," the integration did not happen.
No real integration. Reporting the strands separately and never connecting them. This is the defining failure of the method.
Choosing the design late. Decide your integration approach in the protocol, not after you see the findings.
Using one search for both strands. Qualitative and quantitative studies are found differently. A single search optimized for trials will under-retrieve qualitative work.
Appraising everything with one quantitative tool. Use MMAT or design-appropriate tools. Judging a qualitative study by trial standards is meaningless.
Doing it by default. Mixed-methods reviews cost substantially more effort. Choose one because the question demands both strands, not because more evidence sounds better.
What is a mixed-methods systematic review? A review that synthesizes both quantitative and qualitative evidence and integrates them to answer a question that needs both, typically what the effect is and how or why it occurs.
What are the integration designs? Convergent (synthesize separately, integrate at the end), sequential (one strand informs the other), and convergent integrated (transform the data so both are synthesized together).
What appraisal tool should I use? The Mixed Methods Appraisal Tool (MMAT) is designed to appraise qualitative, quantitative, and mixed-methods studies within one framework. Design-specific tools for each strand are the alternative.
How is this different from doing two separate reviews? The integration. If the two strands never meet and produce no combined insight, you have written two reviews rather than one mixed-methods review.
When should I not do a mixed-methods review? When your question is purely about effect size, or purely about experience. The added effort is only justified when the two strands genuinely illuminate each other.
A mixed-methods systematic review combines quantitative and qualitative evidence to answer questions that need both halves. Choose your integration design in the protocol, search for both strands properly, appraise each with an appropriate tool, and synthesize each on its own terms.
Then do the part that actually defines the method: integrate. If your final review says nothing that a separate quantitative and a separate qualitative review would not have said, the effort was wasted. The insight lives in the join.
Managing two strands of evidence in one review? Verflux handles the search, deduplication, screening, and appraisal underneath both.
All analytical features included in the free trial. No credit card, no installation, no R or Python.
Create free account