Meta-ethnography: synthesising qualitative studies interpretively

The organising idea is translation: taking the concepts of one study and expressing them in the terms of another, then checking whether the translation holds. If study A describes patients "rationing" their disclosure to clinicians and study B describes them "editing" what they report, the analytic question is whether these are the same concept differently named, overlapping concepts, or genuinely different things that the surface similarity conceals.

The synthesis works with second-order constructs — the original authors' interpretations — rather than with raw participant quotes, which are first-order. What the synthesis produces is a third-order construct: a new interpretation, built across studies, belonging to neither.

That layering is what distinguishes meta-ethnography from other qualitative syntheses. The output is not "these six studies all found stigma"; it is a conceptual account explaining how the phenomena the six studies described relate to one another, at a level of abstraction none of them reached.

Meta-ethnography was set out by George Noblit and Dwight Hare in Meta-Ethnography: Synthesizing Qualitative Studies in 1988, developed in educational research and later adopted heavily in health.

Its adoption outran its methodology. France and colleagues' methodological systematic review of what is wrong with meta-ethnography reporting found published meta-ethnographies frequently omitted which studies contributed which concepts, how translation was actually carried out, and what distinguished the synthesis from the inputs.

That led to a reporting standard: the eMERGe reporting guidance, developed through a formal consensus process described in its protocol paper. It is now expected by many journals, and following it is the single most effective thing an author can do to get a meta-ethnography accepted.

A meta-ethnography of eleven qualitative studies of adherence in long-term conditions sampled purposively for conceptual variety across conditions rather than exhaustively within one.

Concept extraction produced second-order constructs including "strategic non-compliance", "drug holidays", "testing the boundaries", "negotiated adherence" and "resistance to medicalisation".

Reciprocal translation established that the first four were substantially the same concept: patients varying their medication deliberately, on the basis of self-observation, to establish what they actually needed. The vocabularies differed by discipline — the sociological studies framed it as resistance, the health-psychology studies as self-regulation — and the underlying construct was shared.

Refutational translation was the more interesting half. Two studies contradicted the rest: patients described adherence as a way of asserting control rather than departing from the regimen as a way of doing so. Rather than treating this as noise, the synthesis characterised the opposition and looked for its conditions. Both dissenting studies concerned conditions with immediate, unambiguous feedback from a missed dose. Where the body answers quickly, adherence itself becomes the site of agency; where feedback is slow or absent, variation is what generates the information.

The line-of-argument synthesis was that construct: patients are conducting experiments on themselves to obtain the evidence their condition does not otherwise supply, and the form the experiment takes depends on the feedback the condition provides. That reading appears in none of the eleven studies. It is what a synthesis is supposed to produce, and it came out of the refutational work rather than the reciprocal work.

• Counting themes across studies. The commonest failure. Reporting that seven of eleven studies mentioned stigma is aggregation, not translation.
• Synthesising participant quotes. Working from first-order data ignores the interpretive layer the method is built on.
• Skipping refutation. Contradiction is analytically productive and is usually discarded.
• No line of argument. A set of reciprocal translations is a legitimate partial output; presenting it as a third-order synthesis is not.
• No provenance. Not showing which studies fed which construct removes the reader's ability to check anything.
• Chasing exhaustive retrieval. Conceptual richness beats completeness here, and an enormous thin sample is harder to translate, not easier.
• Synthesising incommensurable studies. Papers from very different theoretical traditions may not be translatable, and forcing it produces a construct that means nothing.

Follow eMERGe. Beyond that, the questions that matter are whether translation is actually demonstrated rather than asserted, whether refutational cases were sought and reported, whether each construct's provenance is traceable to specific studies, and whether the synthesis contains anything that was not already in one of the inputs.

A reasonable test: if every finding could be attributed to a single included study, the synthesis has not synthesised.

Meta-ethnography depends on the interpretive quality of the studies it includes. Papers that report thin descriptive themes offer little to translate, and no amount of synthetic effort recovers what the primary analysis did not do.

It also does not scale. The method assumes close reading of every paper, and beyond roughly forty studies the translation work becomes impractical — which is why large qualitative evidence bases are usually handled with framework synthesis or thematic synthesis instead.

Reproducibility is limited by design. Two teams synthesising the same eleven papers may produce different third-order constructs, both defensible. The method's answer is transparency about how the translation was done, which is why eMERGe exists and why compliance with it matters more here than reporting guidelines usually do.

Meta-synthesis is often used as an umbrella term; meta-ethnography is one specific approach within it.

Thematic synthesis is more descriptive and scales better, at the cost of the interpretive layer.

Qualitative evidence synthesis covers the whole family, including framework synthesis and critical interpretive synthesis.

Narrative synthesis is the choice where included studies are too heterogeneous in method to translate.

The material is a set of published papers, and the operations are extraction, juxtaposition and provenance tracking: pulling second-order constructs with their page references, laying concept sets side by side, and maintaining the link from every third-order construct back to the studies that produced it.

That last one is what eMERGe compliance turns on, and it is what breaks down when the work lives in a word processor across a team. Handling the papers as a document corpus with coded constructs makes the audit trail a by-product rather than a separate chore. Evidano supports meta-ethnography as a named methodology on the document-analysis side. Translation itself — deciding that two concepts are the same concept, or that they genuinely conflict — is the interpretive work the method consists of.