The Office for Health Improvement and Disparities, in its GOV.UK guidance on mixed methods studies, defines the approach as one that “combines quantitative and qualitative data collection and analysis in one study.” The important word is combines: the data are not merely collected side by side. They form parts of one research design, with each component contributing to the study’s central question.
Mixed methods should therefore be treated as a choice between designs, not as a sign of methodological completeness. The case for combining must come from the question itself.
What Combining Quantitative and Qualitative Data Actually Buys You
The GOV.UK guidance identifies several possible benefits. Combining the approaches “can balance out the limitations of each method,” provide “stronger evidence and more confidence in your findings,” and produce “more granular results than each individual method.”
Each benefit depends on the design being used. Balanced evidence does not mean that one method automatically corrects every weakness of the other. It means that the study gives each a defined role in answering the question.
A quantitative result may identify a change or pattern, while a qualitative component can investigate the issues surrounding that result. The GOV.UK examples make this relationship concrete. A randomised controlled trial can assess the effect of a mindfulness meditation app on establishing a regular meditation routine. Focus groups with trial participants can then help explain why some users stopped using the app after one month. The measured effect and the reported experience address different aspects of the same problem.
Greater granularity should likewise not be equated with adding material for its own sake. The additional component is useful only when it exposes a feature of the question that the other component would leave unexamined.
When Combining Quantitative and Qualitative Data Works
Combination is most defensible when the research question contains distinct jobs that can be assigned to different components. One component may establish a pattern or effect; another may investigate experiences or issues that help interpret it. A survey with open-ended questions, for example, can gather quantitative data while also collecting qualitative responses within the same study.
The relevant question is not “Can this study include both methods?” but “What work will each method do, and how will those contributions be considered together?” If the answer remains unclear, adding a second method may increase workload without improving the evidence.
Concurrent Mixed Methods Designs: Collecting Both at the Same Time
In a concurrent design, the quantitative and qualitative components are collected at the same time. Neither phase has to shape the later collection of the other.
The GOV.UK guidance gives a straightforward example: a survey can gather quantitative data while open-ended questions collect qualitative data. This design applies when both kinds of response are needed within the same research encounter and the open questions do not depend on findings produced later.
Its appeal is directness. Structured responses and open comments are gathered without waiting for an initial result to determine what should be asked next. The study still needs to explain how the two forms of evidence will be collected and analysed as parts of the same investigation.
Each component requires its own bias assessment. The Cochrane Handbook, chapters 7 and 25, explains that study designs carry their own risk-of-bias domains. A concurrent label does not remove the need to assess the quantitative survey strand and the qualitative open-response strand separately.
Sequential Mixed Methods Designs: Letting One Phase Inform the Next
A sequential design gives the components an order: one phase produces something that the next phase uses. It applies when the second phase is needed to investigate, interpret or explain an issue arising from the first.
The mindfulness-app example in the GOV.UK guidance is sequential. The randomised controlled trial comes first and assesses the app’s effect. Focus groups with RCT participants then examine why some users stopped using the app after one month. The second method is not simply additional commentary; it pursues a question prompted by the trial.
The guidance also describes a before-and-after study followed by interviews. The first design assesses whether a product can be effective for different populations. The interviews then investigate issues experienced by a potential user group. Here, the qualitative phase follows a result and asks what the measured assessment does not explain.
The design should state the connection explicitly. What result from the first phase creates the need for the second, and what will the second phase investigate? Without that link, sequential collection may amount to two related studies rather than one mixed methods design.
Each phase still needs a separate bias assessment. The Cochrane Handbook’s design-specific approach means that connecting a qualitative phase to an RCT or a before-and-after study does not transfer the first component’s risk-of-bias assessment to the second.
Embedded or Nested Mixed Methods Designs: Giving One Component a Different Role
In an embedded or nested design, one design sits inside the other and performs a different role. Its defining feature is the assigned function within the wider study, rather than simply the time at which data are collected.
This design applies when a central design carries the main inquiry and another component has a bounded purpose within it. That purpose might be to investigate an issue that the host design cannot address directly, but the proposal must specify it. “Interviews were also conducted” does not explain why the interviews are embedded, what question they address, or how their findings relate to the wider analysis.
Embedding should therefore be justified against the research question, not selected because it appears comprehensive. The Cochrane Handbook, chapter 3, states that included study designs should be justified by their appropriateness to the question and their potential for bias. Applied to method choice, that principle requires the embedded component to have a reason to be there.
The host design and the embedded component remain distinct evidence streams. Each needs its own bias assessment, including consideration of the domains relevant to its design.
How to Decide Whether a Mixed Design Is Right for Your Question
Begin with the problem, not with a preferred method. Write down what the study needs to establish, describe or investigate. Then ask what each proposed component contributes that the other cannot address on its own.
A useful test is to remove one component temporarily. If the central question remains fully addressed, why is that component necessary? If its purpose cannot be stated without vague terms such as “add depth” or “increase richness,” the design is not yet justified.
The GOV.UK guidance also makes the cost explicit. Mixed methods “can be more complex to carry out.” They “may require more expertise to collect and analyse data, and to interpret the results” than a study using one method. Combining methods “requires extra resources, such as time and money.”
These costs should appear in the design decision rather than after data collection has begun. The question must justify the additional collection, analysis, interpretation and coordination. The Cochrane Handbook’s rule is decisive here: method combination should be appropriate to the question and assessed for bias, not adopted for completeness.
A defensible proposal can therefore be checked by asking:
- What specific job does the quantitative component perform?
- What specific job does the qualitative component perform?
- Are the components concurrent, sequential, or embedded?
- What connects their collection and analysis?
- Which bias domains apply to each component?
- What expertise, time and money does the combination require?
If the proposed components lack distinct roles, a single-method design may be more appropriate. If each has a necessary function, the next task is to make their relationship explicit.
Deciding Which Component Comes First
Order should follow dependency. A useful drafting prompt is: “The first component will provide __; the second component will use that to ____.”
If the initial phase must produce an effect or before-and-after result that later interviews investigate, the evidence supports a quantitative-first sequential design. This is the logic of the GOV.UK examples involving an RCT followed by focus groups and a before-and-after study followed by interviews.
If neither component depends on findings from the other, a concurrent design may fit better. If one design occupies a distinct supporting role within a wider design, an embedded or nested structure may be more appropriate.
Before comparing programmes or further academic directions, convert the idea into a short design note: the research object, the role of each component, their relationship, the plan for considering both forms of evidence, the component-specific bias checks and the resources required. That note turns “I might mix methods” into a question that can be examined, compared and revised.