Jump to text
On this page
Copy and source
Published source confirmed at last check

Source changed 2026-11-18 21:11:39 UTC · snapshot created 2026-11-18 21:12:35 UTC · last check 2026-11-18 22:25:20 UTC

RMP.7 - Synthesize a Research Evidence Corpus

Status: Stable

Type: Research Method Practice practitioner method pattern

RMP.7:1 - Problem Frame

Use this pattern to work out what several research reports or interpretations establish together, or what an existing synthesis already answers. State the question and identify the material the answer covers. Reuse a synthesis that answers it under the conditions at issue.

Four reports can cover two independent tests and material about a different outcome. Several histories can repeat one memoir. Studies can also appear to disagree because their questions or settings differ. Identify these relations before treating agreement or disagreement as evidence for the same claim.

The resulting EvidenceSynthesis states what the corpus warrants, how its sources contribute to that answer and which limits matter to criticism or reuse. A short answer citing an existing review may suffice. A new analysis may need a study map, a comparison or a calculation.

Start from the question, synthesis design, operations and record adequate for this use. They may come from RMP.1–RMP.4 or from work already available. When historical or qualitative material warrants refining the question, explain what changed and which evidence and criticism apply to the revised question.

When finding or comparing sources answers the question, use that result. RMP.5 examines an inference from particular material. Use RMP.7 when the question depends on how a corpus was appraised or its contributions combined. If a needed method, source or question is missing, say which conclusion this prevents. When that conclusion is still required, obtain the missing contribution from a suitable source where possible.

RMP.7:2 - Problem

Several accounts can look like independent corroboration even when they draw on one study. A corpus may contain only favorable results. A pooled number can conceal different target quantities, and a common theme can conceal a consequential difference in setting or meaning.

Disagreement or incomplete coverage may still permit a numerical comparison, an interpretation or a precise conclusion about what the corpus cannot establish. If a previous review already answers the question, giving that review a new date adds no evidential support. A newly obtained observation may add support of its own.

Choose the form of the answer from the research question, the relations among the sources and the criticism that affects the conclusion.

RMP.7:3 - Forces

ForceTension
Joint supportDistinct contributions can strengthen an inference. Several reports of one result can add detail while counting as one study.
A concise answerA synthesis reduces reading effort, but an aggregate can hide a contrast needed to interpret the finding.
Different traditionsQuantitative, interpretive and historical inquiries can inform one question, while their transformations and confidence claims differ.
CoverageRelevant omissions can change a finding, while a complete search may exceed what a bounded question needs or what access permits.
Reuse and changeAn existing synthesis can supply the answer. A change to a source, the question or the intended use can affect the conclusion or its scope.
CompletionA qualified answer may settle the corpus question. Commissioning a further inquiry depends on what it could add and what it would cost.

RMP.7:4 - Solution

Start with the question and the material available. Establish what each distinct contribution supports and how it depends on other sources. Use the synthesis the question and material warrant. Give the answer, the transformations that affect it, its uncertainty and its scope.

RMP.7:4.1 - State the question and test the supplied answer

Specify the subject, the context and period that matter, and the conclusion sought. A mean difference needs a contrast. An account of participants’ experiences, a historical sequence and an explanation each ask for different evidence, even about the same subject.

For a supplied synthesis, check its question, coverage, method, findings and limits against the present use. Return its applicable findings with their source and qualifications. If a different population, changed premise or uncovered question matters to the task, reconsider the affected part or identify the limit.

If further synthesis is needed, establish what met the inclusion criteria, what was sought and obtained, and what was excluded or unavailable. Explain how this affects the answer. A claim of systematic coverage needs evidence of the relevant search and selection. A specified set of documents may be sufficient for a question about that set.

When further work requires a missing contribution, use RMP.1 for a changed question, RMP.2 for the synthesis design, RMP.3 for the operations and RMP.4 for what material was obtained or can be examined. Use adequate available inputs directly.

RMP.7:4.2 - Recover distinct contributions before counting support

Identify reports, the underlying studies and material, the authors’ interpretations and the finding you propose to draw across sources. One report can contain several studies, and one study can generate several reports.

Source relationConsequence for synthesis
Several reports concern one studyUse details and follow-up information from the reports, and resolve discrepancies where possible. Count this study once when estimating independent study contributions.
One report contains several studiesIdentify the studies and establish whether their samples, controls or data are independent for the inference in question.
Samples, controls, records or data overlapAccount for the overlap in the analysis, group the contributions, or justify selecting among them. Establish independence from the underlying material.
Later accounts copy or interpret one earlier sourceIdentify their shared source and the interpretations each adds. Evidence of copying can establish transmission. Independent corroboration of the reported event needs evidence independent of that source.
A source adds context or a different outcomeUse it for the question it answers. Include it in an estimate of the target quantity only when its data support that estimate.

Preserve contrary findings and disagreements between reports. If dependence remains uncertain, explain which combinations or narrower comparisons it permits and which conclusions it prevents.

RMP.7:4.3 - Appraise support for the proposed findings

Apply the appraisal criteria for the research approach and the finding. Consider limitations within a study, dependence across contributions, which results became available and whether the evidence applies to the question. Explain which findings each condition affects.

For consequential omissions, distinguish material never sought, an eligible result left unreported, a lost record and a source you cannot access. State which conclusion the gap limits. To infer absence of an event, effect or experience, establish how the available evidence bears on that absence.

Additional publications give no independent replication when they rely on the same observations. Appraise the connection between each source and the inference. Material too weak for one conclusion can still provide context or challenge a proposed interpretation.

Use RMP.5 when an individual inference needs reconsideration. RMP.6 can help interpret a credibility examination already performed or decide whether a further examination could improve the synthesis enough to justify its cost. For an empirical test result, establish what the test found.

RMP.7:4.4 - Perform the synthesis the question permits

Use a disciplinary method appropriate to the question. Explain any grouping, transformation, weighting or translation on which the finding depends so a reader can trace the contribution of the sources.

Question and materialPositive moveWhat the result may claim
Compatible quantitative estimatesEstablish the target quantity, units, contrast, timing, dependence and uncertainty. Fit the justified statistical model and examine influential assumptions.An estimate for that target with its warranted uncertainty and scope. A common-effect and a distributional-effect analysis have different assumptions and interpretations.
Quantitative findings without a justified common aggregateGroup and tabulate the relevant results and compare what the evidence permits. Identify the missing relation when it prevents a requested aggregate.A narrower quantitative comparison or the conclusion that the corpus cannot establish the pooled magnitude. Significance counts do not estimate that magnitude.
Qualitative or interpretive findingsCompare, translate or develop interpretations under the selected tradition, retaining context, source concepts, disconfirming material and the synthesist’s contribution.A descriptive synthesis or explicitly developed explanatory construct with its grounding and applicable confidence account.
Historical sources and accountsExamine provenance, dates, purposes, perspectives, dependence, gaps and rival interpretations that bear on the chronology or explanation.The sequence or explanation the sources warrant, with any consequential alternatives still open.
Mixed evidence about one questionIf a transformation makes the contributions comparable for this question, justify it. Keep the original values, meanings and context needed to examine what the transformation loses.An integrated answer within the scope of the justified transformation.
Mixed evidence about different aspectsSynthesize the aspects on their own terms. Where the question requires their relation, compare the syntheses or use one finding to guide examination of the other.An account of how the findings agree, complement one another or diverge, including any unresolved difference.

Variation among studies can be compatible with a pooled estimate under a justified model. A random-effects model still needs a defensible relationship among the target quantities. A synthesis of effect directions alone does not estimate their magnitude.

For a qualitative synthesis, identify what the participants reported, how the source authors interpreted it and what interpretation the synthesis adds. When GRADE-CERQual fits, assess confidence in each finding from methodological limitations, coherence, adequacy and relevance. The resulting confidence judgement is qualitative and specific to that finding and approach.

If a required method or decisive premise is unavailable, identify the conclusion that this prevents and give any answer the available material supports.

RMP.7:4.5 - Return the useful corpus answer

Give the answer to the corpus question first. Identify the material and method used, source dependencies, transformations, uncertainty, omissions and transfer limits insofar as they affect interpretation.

Use existing notes, analysis output or a concise linked explanation when they suffice. Provide the evidence needed for the intended criticism within the applicable access conditions. If a qualified inspection of restricted material has established a finding, identify what others can conclude from it and what they can inspect for themselves.

Return the adequate existing synthesis, the new finding, a qualified answer, or the reason the requested result cannot be established. Identify a missing input if obtaining it matters to the work. A broader question may remain open. Select further inquiry only when the attainable contribution justifies the effort, cost and delay.

Use RMP.8 when reconciling the answer with a wider claim or model, or when its return to a receiving practice needs further judgement about reliance. Return an adequate corpus answer directly.

RMP.7:5 - Archetypal Grounding

These examples use constructed studies and records.

RMP.7:5.1 - Two compatible tests among four reports

A seal engineer receives four reports. A and A2 describe one randomized accelerated test. A2 provides fuller outcome information. B reports an independent test of the same materials under compatible stress and measurement conditions. C describes in-service replacement decisions under different conditions.

The engineer wants the difference in mean time to a defined laboratory failure between the new seal and its comparator under specified accelerated conditions. Suppose all lifetimes were observed, the study estimates are approximately normal and independent, and a common underlying effect is justified for these conditions. A2 informs the first study estimate. C does not measure this laboratory contrast.

The two estimates are 14 hours for A/A2 (standard error 4 hours) and 8 hours for B (standard error 3 hours). The inverse-variance weights are 1/16 and 1/9. The weighted estimate is 10.16 hours, with a standard error of 2.4 hours and an approximate normal 95% interval from 5.46 to 14.86 hours.

Under the stated assumptions, the new seals have a longer estimated mean time to laboratory failure. Report the two study estimates as well as their combined estimate. Transfer to service and a change in the maintenance interval need separate grounds. C can inform a question about service use. If the common-effect premise fails, reconsider the interpretation of the combined estimate even when the arithmetic holds.

RMP.7:5.2 - A refusal with useful content

Another supplied corpus reports leakage after one hour, failure time under another stress and technicians’ replacement decisions. The requested answer is a common percentage reduction in 500-hour failure risk.

Report the observed leakage, failure time and replacement choices with their contrasts, conditions and uncertainty. The supplied material gives no way to convert them to a common 500-hour failure risk. It supports the individual comparisons, but does not establish the requested percentage or show an absence of effect. This answers the question about what this corpus can establish.

RMP.7:5.3 - Situated participation and a mixed question

A resident group asks what impedes participation in water monitoring. One interview study links distrust to withheld feedback. Another interpretive study relates continued participation to community authority over the question. A survey reports stated reasons within its own sampled group.

For a mixed synthesis addressing this question, preserve the survey counts and denominator, the interview contexts and the authors’ interpretations. The synthesist proposes that control over the question and the information returned may both matter. The account supplies no population prevalence for that explanation. Continued participation in one case despite withheld feedback qualifies the proposed connection.

To estimate the intervention’s effect on water quality, use studies that measure that effect. If the question also concerns how participation relates to water quality, synthesize each kind of evidence appropriately and examine the connection.

RMP.7:5.4 - Shared memoir and contrary chronology

Several histories attribute a factory’s decision to close to a public protest. They cite one retrospective memoir. An authenticated record dates the decision before the protest.

The histories share a source, though their interpretations may differ. The dated record rules out the later protest as the cause of the earlier decision. The decision’s cause and the protest’s possible effect on implementation remain open. The historian can finish the chronology without counting the histories as independent corroboration.

RMP.7:6 - Bias-Annotation

In the seal example the target quantities are compatible, independence is established and the common-effect premise is supplied. Other corpora may require domain expertise, further source recovery or a narrower answer. Historical chronology and accounts of participation illustrate different appraisal criteria.

Publication and access can favor certain institutions, languages and surviving sources. Interpret the coverage and silences of a corpus in that setting. Reuse a convenient synthesis to the extent that its question and qualifications apply.

RMP.7:7 - Conformance Checklist

When examining a synthesis, check the following.

  • Can the reader identify the question and determine what the existing work already answers or leaves open?
  • Can the reader distinguish reports, underlying studies and material, source interpretations and synthesis findings, and identify dependence that affects the answer?
  • Do inclusion criteria, exclusions, omissions and access conditions justify the stated scope of the corpus?
  • Are the method and interpretation of each consequential transformation appropriate to the question it serves?
  • Does each affected finding account for contrary evidence, uncertainty and limits on transfer?
  • Does the answer give what the available material establishes, including a specific limit when it cannot support the requested finding? Is any proposed further work distinguished from evidence already obtained?

Use the evidence for the findings and source relationships to judge whether the synthesis answers its question.

RMP.7:8 - Common Anti-Patterns and How to Avoid Them

FailureWhat changes in practice
Reports are counted as independent studies.Identify the underlying material, account for dependence and retain any supplementary information.
Numerical results are pooled without a common target.Establish their targets and any justified transformation. Give the narrower comparison or the specific limit if a common magnitude cannot be estimated.
Agreement is judged by counting significant results.Compare the relevant estimates or other findings under a justified method, including uncertainty and missing evidence.
A theme is reported as a population percentage.Preserve its interpretive basis. A prevalence estimate needs appropriate sampling and measurement.
A single confidence label is applied to every finding.Appraise each finding under criteria suited to the research approach and the question.
Disagreement leads the reviewer to discard the corpus.Give the distinctions the sources support or explain precisely which conclusion they cannot establish and why.

RMP.7:9 - Consequences

The resulting answer preserves the differences between sources that matter to the question. Compatible findings may strengthen an estimate. A dependent report may add follow-up detail, and a result about another outcome may help explain the scope of the comparison.

Appraisal and transformations take work. Reusing adequate earlier syntheses and documenting the decisions that affect interpretation can reduce that cost. The strength of the evidence still depends on what the sources establish.

RMP.7:10 - Rationale

A synthesis draws an inference about how several contributions relate. Its design and criticism must support that inference. Use an established individual finding directly when it answers the question. Use RMP.5 when the inference from particular material still needs examination.

A finding that the requested conclusion is unsupported can settle the inquiry when it identifies the missing relation and shows what, if anything, the material does establish.

RMP.7:11 - SoTA-Echoing

What counts as independent support? Cochrane chapter 5 explains how to identify several reports of one study and combine the information they contribute. Counting repeat reports as additional studies inflates the number of independent studies. One report can also describe several studies. Discarding all repeated reports can lose follow-up observations. Section 4.2 preserves A2’s fuller outcomes in §5.1 and counts A/A2 as one study. Recover underlying relationships where they could change an inference. New evidence of overlap, provenance or additional outcomes can change it.

Which numerical answer can this corpus support? Cochrane chapter 10 guides combination for a defined quantity under an appropriate dependence model. chapter 12 describes synthesis without meta-analysis. Sections 4.4 and 5.1–5.2 use the target and model to distinguish a justified combined estimate from a narrower comparison. Pooling all numerical results assumes they measure compatible quantities. Refusing every heterogeneous comparison would discard some useful estimates. Choosing groups and models takes judgement. Reconsider those choices if the target, dependence or a defensible alternative model changes. These chapters address intervention reviews.

How should missing contributions affect a finding? Cochrane chapter 13 examines missing evidence in a synthesis and its potential effect on a finding. Its meta-analysis tools have that scope. Sections 4.1 and 4.3 distinguish material unavailable, unreported, never sought or lost. Assuming completeness can conceal selective reporting. Rejecting every incomplete corpus would also discard answers supported within a narrower scope. Reconsider the affected finding when new material, a selective omission or evidence of its effect becomes known.

How can situated and mixed findings be related? For a review of intervention effectiveness, Harden and colleagues (2018) recommend synthesizing qualitative and quantitative evidence separately before integration. The archived Cochrane Handbook v6.0, §21.13 gives that recommendation in its original setting. In Sutcliffe and colleagues, Cochrane–Campbell Handbook, chapter 14 (February 2024 draft), the authors describe transformation for a shared question, comparison of separate syntheses and using one synthesis to guide another. JBI §8.3 distinguishes an integrated approach for evidence that answers the same question from separate syntheses of different aspects. Sections 4.4 and 5.3 use the question to choose among these operations, preserve context and show what a transformation loses. GRADE-CERQual assesses confidence in individual qualitative findings. Reconsider the choice if the question, research approach or evidence about those losses changes.

What can agreement between historical accounts establish? The AHA professional standards address the integrity and interpretation of source evidence. Sections 4.2–4.4 and 5.4 examine dependence, dates, purpose, contrary material and rival explanations. Counting the histories is quick, but they repeat one memoir. Source criticism establishes the chronology the dated decision record supports. It leaves the cause open. Reconsider that inference if new provenance, dating or a rival explanation affects its grounds.

These sources offer different ways to synthesize evidence. Choose the method appropriate to the research question and the relationships among the sources.

RMP.7:12 - Relations

RMP.10 can distinguish candidate accounts of how a practice was performed when evidence supports different variants. RMP.11 compares explanations of an identified method’s history. The synthesis can establish how their sources contribute or depend on one another. Assess the explanations against the evidence that distinguishes them.

  • RMP.1–RMP.4 can supply the question, design, operations and record of actual work. Use existing inputs that answer those questions. Revisit a missing or changed premise when it limits the required synthesis.
  • RMP.5 examines individual source claims, including one that needs revision. RMP.6 can interpret an existing credibility test or help decide whether another is worth obtaining.
  • Use RMP.8 when a wider claim or model needs reconsideration, or when a receiving practice needs a judgement about reliance. Return an adequate answer about this corpus directly.
  • FPF F.1, F.0.1 and A.10, with a qualified F.0.2 contribution where applicable, help find, select and compare sources and establish how evidence bears on a claim. For a research synthesis, also establish what the contributions warrant together.
  • Use the relevant domain, access or receiving-practice method for a disciplinary criterion, permission or decision when it is needed. Use RMP.9 and the applicable FPF methods when the question is how to improve a research Method.

RMP.7:End

Referenced in the corpus

34 literal mentions in other sections. Read their context to establish the relation.