From Audit Trail to Manuscript

If your methods section is written from memory, the study record is already doing too little.

A sound design can still be difficult to evaluate when its execution is incompletely reported: round dates are missing, denominators do not reconcile, item wording changes between tables, or the participant flow cannot be reconstructed. The goal of an audit trail is not to make the manuscript longer. It is to make every important claim traceable to the study record.

Write against a checklist from the start

For health-related consensus studies, ACCORD is a general reporting guideline that covers Delphi and other consensus methods. CREDES remains useful Delphi-specific guidance developed in palliative care. A checklist is a reporting aid, not a substitute for methodological judgment, but using one during protocol design helps ensure the required information will exist later.

The overlapping core includes:

• The definition of consensus, stated a priori, with scale, statistic, cut point, denominator, and missing-response policy.
• Panel eligibility, candidate sources, selection process, intended size, final composition, and relevant strata.
• The number of rounds, planned stopping rule, and any departures from the plan.
• Participation and attrition by round, with counts as well as percentages.
• The anonymity arrangement and content of feedback between rounds.
• Item origin, additions, exclusions, and wording changes across rounds.
• The result and disposition of every item, including unresolved disagreement.
• Investigator, facilitator, and steering-committee roles in proposing, revising, voting, or adjudicating items.
• Ethics, consent, conflicts of interest, funding, protocol availability, and registration where applicable.

Treat these as a data dictionary during protocol development. If every element has a home before Round 1 opens, manuscript drafting becomes reporting from a record rather than reconstructing one.

Report non-consensus as a finding

Items the panel could not agree on can identify weak evidence, competing values, or legitimate practice variation. Omitting them makes the final instrument look cleaner while hiding the limits of the consensus process. Report unresolved items and their distributions, and discuss what the disagreement means rather than treating it as procedural debris.

Publish a supplement another team can use

Subject to consent, ethics, licensing, and confidentiality constraints, a useful supplement may include:

• The instrument as presented in each round.
• Per-round item results with valid denominators and missingness.
• The feedback template and a description of any personalization.
• Participant flow and attrition by relevant stratum.
• Item version history with revision rationales.
• A de-identified qualitative codebook or synthesis method when comments were analyzed.

Do not publish free text merely because it has been stripped of names. Rare roles, locations, cases, and writing details can still identify a panelist. State what was withheld and why when privacy limits the shareable record.

Protocols, registration, and amendments

A prospectively available protocol—and registration where an appropriate registry exists—strengthens the distinction between planned methods and decisions made after observing the data. ACCORD asks authors to identify registration or protocol access when available and to describe deviations. Amendments are not inherently a flaw: record what changed, when, why, and which results were affected. Never make the current protocol overwrite the version under which earlier rounds ran.

What the platform produces

Delphi Studio's reporting surfaces draw on the same study objects used during execution: per-round distributions and denominators, frozen consensus-rule snapshots, response and attrition data, item-version lineage, revision rationales, feedback-packet snapshots, protocol amendments, and the hash-chained audit log. The checksummed trace bundle preserves these records with an integrity manifest, while publication exports provide a readable reporting layer. See Reporting and exports and the ACCORD–CREDES crosswalk.

That was the design goal from the beginning. A Delphi study should be defensible because of how it was run, and the record should make that visible without anyone having to reconstruct it eighteen months later.

This concludes our five-part series on Delphi methodology. Continue with the documentation for study setup, panel management, and analysis configuration.