03 · What You Need to Know
Why Integration Has to Be Designed, Not Added at the End
Mixed Methods Is More Than Two Methods in One Study
The NIH Office of Behavioral and Social Sciences Research describes mixed-methods research as involving the collection, analysis, and integration of quantitative and qualitative data to provide a more comprehensive understanding of a research problem than either approach might provide alone.
That final element matters. A study containing a survey and interviews may be multimethod without being meaningfully integrated. If the quantitative and qualitative components address unrelated questions, use unrelated samples, and are interpreted independently, their coexistence does not by itself explain what is gained by combining them.
A mixed-methods plan should therefore identify the purpose of integration. What can be understood by bringing the strands together that would be difficult to understand from either one alone?
Start With a Mixed-Methods Question or Integrative Purpose
A study may have quantitative research questions and qualitative research questions, but it should also be clear why their answers need to be related.
For example, a quantitative strand may estimate how common a behavior is, while qualitative interviews investigate why participants engage in it. An intervention study may estimate whether outcomes changed while qualitative inquiry examines how participants experienced the intervention or why implementation varied.
NIH guidance recommends that mixed-methods aims explicitly call for integration when integration is necessary to address the overall study purpose. This makes integration part of the scientific question rather than an analytical decoration.
The broader principle that each research question needs a planned analytical path therefore extends to the mixed-methods question itself.
Choose a Design That Determines How the Strands Relate
The mixed-methods design provides the architecture for integration.
NIH guidance identifies common basic designs including exploratory sequential, explanatory sequential, and convergent approaches. These labels are useful because they describe fundamentally different relationships between quantitative and qualitative work.
| Design |
Typical Relationship |
Planning Implication |
| Exploratory sequential |
Qualitative work occurs first and informs a subsequent quantitative phase. |
The plan must explain how qualitative findings will help build, select, or shape what happens quantitatively. |
| Explanatory sequential |
Quantitative work occurs first and qualitative inquiry follows to explain or elaborate selected findings. |
The plan must explain how quantitative results will guide qualitative sampling, questions, or areas of investigation. |
| Convergent |
Quantitative and qualitative strands are conducted in a broadly parallel period and then brought together. |
The strands should be designed so that their findings can be meaningfully compared, related, or merged. |
More complex mixed-methods designs also exist. The important issue is not choosing a fashionable label. It is knowing how the strands are supposed to interact.
Identify the Point or Points of Integration
Integration can occur at more than one stage.
NIH mixed-methods guidance refers to the “point of interface” where mixing occurs, which may be during data collection, analysis, or interpretation. Methodological literature further describes integration at the design, methods, and interpretation or reporting levels.
Before collection, identify where the study intends to bring the strands together. A project may have one main point of integration or several.
If you cannot identify any point at which the quantitative and qualitative components actually interact, reconsider whether the project has a mixed-methods design or simply contains separate methods.
Connecting Uses One Strand to Inform Sampling for Another
One form of integration is connecting. Here, information from one component influences who participates in another.
For example, researchers may administer a survey and then purposefully select interview participants based on survey responses. They might seek participants with very high and very low scores, cases showing unexpected combinations of responses, or individuals representing particular quantitative profiles.
This requires advance planning because the quantitative dataset must contain identifiers or ethically appropriate linkage mechanisms that permit the intended sampling, and consent procedures may need to anticipate recontact.
If the survey is collected completely anonymously and all linkage is destroyed, researchers cannot later decide that they would like to interview particular respondents unless another ethically appropriate recruitment mechanism was built into the design.
Building Uses One Strand to Shape the Next
In sequential designs, one phase may build the data collection or analysis procedures of another.
Qualitative findings might inform questionnaire items, intervention components, constructs, or hypotheses in a later quantitative phase. Conversely, quantitative findings may identify surprising results that require deeper qualitative exploration.
The integration plan should explain the decision process without pretending the first phase's findings are already known. For example, it can state that interview participants will be selected to illuminate statistically important, unexpected, or contrasting quantitative patterns without specifying which patterns will occur.
Merging Brings Two Sets of Results or Data Together
In a convergent design, researchers may analyze quantitative and qualitative data separately and then bring the results together for comparison or interpretation.
Fetters, Curry, and Creswell describe merging as bringing the two databases together for analysis and comparison. They also emphasize that researchers should ideally plan collection so that the datasets are conducive to being merged.
This may require conceptual alignment. If the survey measures institutional support but the interview protocol never explores anything related to institutional support, comparing the two strands on that issue may be impossible.
Integration planning therefore influences data collection even when the actual merging happens later.
Embedding Integrates a Secondary Strand Within a Larger Design
Some studies embed one type of data within another design. A randomized trial, for example, may include qualitative interviews to understand implementation, participant experiences, mechanisms, or contextual factors.
The embedded component should have a defined purpose. It should not exist merely because adding interviews makes the project appear more comprehensive.
Planning should establish when the embedded data will be collected, from whom, how those participants relate to the larger sample, and how the resulting qualitative findings will inform interpretation of the primary quantitative study.
Plan Whether the Strands Need to Address Parallel Concepts
Integration is easier when researchers know which constructs or phenomena need to be brought together.
This does not mean forcing qualitative interviews to reproduce a survey questionnaire. Qualitative inquiry should retain the openness and depth appropriate to its purpose. But if the study intends to compare qualitative experiences with quantitative measures of the same broad phenomenon, the two components need enough conceptual correspondence for that comparison to make sense.
Planning may therefore involve mapping quantitative constructs against qualitative topics while preserving questions unique to each strand.
Plan How Integration Will Occur During Analysis or Interpretation
Do not stop the integration plan at “results will be compared.” Specify how.
Fetters and colleagues describe several strategies at the interpretation and reporting level, including narrative integration, data transformation, and joint displays. A joint display places quantitative and qualitative information together in a table, figure, matrix, or other structured representation so that relationships can be examined directly.
Narrative integration can also weave quantitative and qualitative findings together by concept or theme rather than reporting one complete results section followed by another with little interaction.
The appropriate technique depends on the question and design. The key is that the planned integration should produce an analytical insight, not merely visual proximity.
Decide How You Will Interpret Agreement and Disagreement
Researchers sometimes assume that successful mixed-methods integration means the quantitative and qualitative findings should agree. That expectation is too narrow.
Findings may converge, complement one another, address different dimensions, or appear to conflict. Fetters and colleagues use the concept of “fit” to describe how findings relate, including confirmation, expansion, and discordance.
Disagreement is not necessarily a methodological failure. It can reveal measurement differences, subgroup variation, contextual influences, different levels of analysis, or assumptions that deserve reconsideration.
The integration plan should therefore allow the research team to investigate discordance rather than forcing one strand to validate the other.
Watch Out
Do not automatically treat qualitative findings as anecdotes whose purpose is to explain or decorate the “real” quantitative result. Nor should a few quotations be used to claim confirmation of a population-level numerical pattern. Each strand has its own evidentiary logic, and integration should respect what each can and cannot establish.
Plan How Integrated Conclusions Will Be Developed
A mixed-methods study should eventually produce conclusions that reflect what is learned by combining the strands.
These are sometimes called meta-inferences: interpretations generated through consideration of quantitative and qualitative evidence together. They should go beyond repeating the separate findings.
For example, a survey may show that institutional support is strongly associated with faculty AI adoption, while interviews reveal that “support” operates differently across departments through access to training, local leadership, and perceived permission to experiment. The integrated conclusion can therefore be more specific than either result alone.
If the final paper could remove either strand without changing its principal conclusions, researchers should ask whether meaningful integration actually occurred.
Integration May Require Practical Infrastructure
Planning also involves mundane but consequential details.
Will participants need identifiers linking survey and interview data? Who will have access to the linkage key? Can quantitative results be available early enough to guide qualitative sampling? Does the project timeline allow transcription and qualitative analysis before the next phase begins? Do team members have expertise across quantitative, qualitative, and mixed-methods integration?
NIH guidance specifically recommends planning sufficient time for qualitative analysis and for integration. These requirements can affect staffing, sequencing, ethics procedures, data management, and project timelines.
Some Integration Decisions Can Legitimately Remain Open
A sequential design cannot specify exactly which quantitative finding will require qualitative explanation before the quantitative analysis occurs. An exploratory sequential study cannot know exactly which qualitative concept will become a survey item before qualitative analysis.
The plan can instead specify the decision logic. It might define what kinds of findings would trigger follow-up, how participants will be selected, what criteria will guide instrument development, or how unexpected divergence will be investigated.
This is a useful example of why an analysis plan can contain deliberate and legitimate flexibility without becoming methodologically vague.
