Research Design for Qualitative Study: Choosing a Design Reviewers Accept

Look at the verb in your question. "Explore experiences of" is vague enough to fit anything, which is why reviewers read it as a sign the design was never decided. Sharpen it until it can only be answered one way.

Here is the move in practice. "Explore nurses' experiences of night shifts" could be any of the six rows. "Describe what nurses identify as the barriers to safe handover on night shifts" is qualitative description. "Explain how nurses develop their own handover routines when no protocol exists" is grounded theory. "Interpret what it means to nurses to be the only senior clinician on the ward at 3 a.m." is phenomenology. Same topic, three studies, three different designs.

If you cannot decide between two rows, the question is still too broad. Fix that before you write the protocol.

Each design implies how you find participants and what you collect.

• Phenomenology needs people who have lived the experience in question, usually a small, specific group interviewed in depth.
• Grounded theory samples in rounds: early analysis tells you whom to interview next. A paper that recruited everyone at once and then claims grounded theory has a visible gap.
• Ethnography is built on time in the setting. Interviews alone rarely carry an ethnographic claim.
• Case study draws on several sources about one bounded case (documents, interviews, observation, records) and says where the case's boundary is.
• Narrative research keeps each person's account intact long enough to analyze its structure, so it usually involves fewer participants and longer accounts.
• Qualitative description can sample for variety (roles, sites, experiences) because it aims to cover the range of views.

Write down the sampling logic in one sentence before recruitment. If that sentence would not change when you swap the design name, the design is not doing any work yet.

This is where most qualitative manuscripts lose reviewers. The analysis section borrows procedures from a different tradition than the one named in the methods.

Two mismatches come up often. A study names reflexive thematic analysis, then reports inter-rater reliability statistics as proof of rigor; the procedure answers a question that approach does not ask. The reverse also happens: a codebook or coding-reliability approach is described as if a single researcher's interpretation were enough. Neither is wrong in every case. What reviewers object to is a paper whose stated approach and described steps do not agree.

When we read a qualitative manuscript, we judge rigor inside the approach the authors say they used. A single coder is not a defect in reflexive thematic analysis, and we do not ask for consensus coding as a universal fix. Where the claimed approach and the actual procedure conflict, the right request is for clarification, not for a different paradigm.

For the methods write-up itself, our guide to writing a methods section covers the order and level of detail editors expect.

There is no correct number of interviews. There is a justification that fits the design.

Reviewers tend to accept an account that ties the number to the question's breadth, how specific the sample is, and how much each interview or observation yields. They tend to challenge "data saturation was reached" when the paper never says how saturation was judged or when analysis started. If you used saturation, say what you looked for and when you stopped. If your approach does not use the idea, do not borrow the phrase.

The quantitative logic of power calculations does not transfer here; our sample size calculation guide covers that side if your study has a quantitative strand.

In our pre-submission review work, five design problems account for most of the reviewer pushback on qualitative studies. Each one shows up in a specific part of the manuscript, which is where to look for it.

1. A design label without design logic. The methods say "phenomenological approach" in one sentence and never mention it again. Nothing in the sampling, interview guide or analysis reflects the choice. The fix is one paragraph early in the methods that says why this design answers this question, followed by sampling and analysis subsections that refer back to it.

1. Procedures from another paradigm. Inter-rater statistics attached to reflexive analysis, or interpretive claims resting on a fixed codebook, with no explanation. Read your analysis subsection against the source you cite for the approach. If a step is not part of that approach, either drop it or say in one sentence why you added it.

1. Themes that are topic headings. The results headings repeat the interview questions ("Experiences of training," "Views on support"). A reviewer reads that as summary, not analysis. Rename each theme as a claim ("Training ended at the ward door") and check that the results text argues for it.

1. Quotes doing the analysis. Each theme rests on one vivid quote from one participant, with no sense of how widely the pattern held or where it broke down. In the results, give each theme your interpretation first, then two or three quotes from different participants, and say where the theme did not hold.

1. Claims the design cannot make. "Most clinicians believe" written into the abstract or discussion as if it were a survey prevalence. Qualitative designs support depth and explanation, not population estimates. Replace counts with what the data actually show, and move any generalization into the limitations.

None of these needs new data. They are writing and alignment problems, and they are cheaper to fix now than in a response to reviewers.