ACCORD and CREDES crosswalk
Transparent reporting lets readers judge whether a consensus result is credible and applicable. Delphi Studio preserves much of the evidence needed for that account, but no platform can write the complete methodological justification on behalf of the investigators.
This page maps two complementary sources of guidance to the records available in Delphi Studio:
- ACCORD — a 35-item reporting checklist for consensus methods used in biomedicine. It is applicable beyond Delphi and is not a conduct or quality-assessment standard.
- CREDES — on the abstract and the EQUATOR record only, guidance for conducting and reporting Delphi studies used to develop best-practice guidance in palliative care. This page does not quote the recommendation table.
For most biomedical Delphi manuscripts, this page's advice is to use ACCORD as the primary reporting checklist and CREDES as complementary Delphi-specific guidance. That advice is not a compliance result. Also apply any guideline required by the larger study design—for example, guidance for core outcome sets, clinical practice guidelines, qualitative research, or patient and public involvement.
:::caution Alignment is not compliance An export can supply dates, denominators, rules, and recorded decisions. Investigators must still explain why the method was appropriate, how evidence informed the item set, how the panel represented relevant perspectives, and what the findings can and cannot support. Choosing CREDES or ACCORD in a study's report settings records the choice. It does not make a downloaded file a compliance document. :::
Planning crosswalk
| Reporting topic | ACCORD/CREDES expectation | Delphi Studio evidence | Investigator must still supply |
|---|---|---|---|
| Objective and intended users | Explain the aim, output, and audience | The objective on Method, and the study name and principal investigator on Identity | Scientific rationale and intended decision context |
| Choice of method | Name and justify the consensus method | The method framework (Modified, Classical, RAND/UCLA or Custom) and its rule in one sentence, the round plan, the anonymity commitment in the Analysis plan, and the stopping advice | Why Delphi—and this modification—fit the question |
| Protocol and registration | Identify a protocol, registration, and planned deviations process where applicable | The protocol label, the document records on Ethics and data governance, the sections locked at launch, and the four recorded amendments, each with a reason | Registry identifier, public protocol citation, and deviations not represented in platform data |
| Steering group | Describe membership, responsibilities, and decision authority | Study team roles (principal investigator, steering committee, steering confirmer, coordinator, coder, adjudicator, biostatistics), and the study audit, which records each action under a pseudonymous actor | Selection process, relevant expertise, conflicts, and off-platform responsibilities |
| Initial topic or item generation | Describe sources, screening, piloting, and transformations | The Item bank: permanent statement IDs, recorded sources (DOI, PMID, web address or free-text reference), areas, imports, and revision evidence that cites the results a revision was based on | Search/review methods, interview or workshop methods, exclusion decisions, pilot findings |
| Panel eligibility | Define expertise and stakeholder criteria before recruitment | Inclusion and exclusion criteria and screening questions on Eligibility, locked at launch | Why each criterion and target was appropriate |
| Sample size | Explain the target and its rationale | Size & mix targets per stakeholder group, each marked core or context, and enrolled counts in the participation report | Methodological justification, anticipated attrition, and any power or precision reasoning used |
| Scale and response options | Describe constructs, anchors, abstention/unable-to-rate handling | Rating dimensions, the scale and its zones, the neutral anchors panelists read, the "Unable to rate" option, and the comment rule on Rating & consensus | Evidence or rationale supporting the measurement design |
| Consensus and stopping | Pre-specify definitions, thresholds, item handling, and stopping | Rating & consensus rules locked at launch, the Thresholds table on Review & launch, the rules fingerprint on each results version, the round plan, and the stopping advice. Stability is configured but not computed. | Justification and any rule not implemented directly by the platform |
| Feedback | State format, content, timing, and anonymization | The live feedback block can show the group median, the middle half of the ratings, zone shares, and the person's earlier rating. Panelists do not see a histogram. The recorded comment-sharing choice and the round approval, which binds the feedback settings, stay on the study. Feedback shows the previous round's current results and is not pinned when a round opens. | Why the chosen feedback was suitable and how any external feedback was handled |
Conduct and results crosswalk
| Reporting topic | Delphi Studio evidence | Reporting notes |
|---|---|---|
| Recruitment flow | Panelist statuses, invitation and consent records, and the participation report's recruitment counts | Report numbers identified, invited, enrolled, excluded, withdrawn, and completing each round; avoid publishing identities |
| Panel composition | Size & mix groups and the profile fields the study collected; groups below the reporting floor are pooled as "Other" | Report only fields collected with an appropriate purpose; explain missing perspectives and small groups |
| Chronology | The study audit CSV: when each round was approved, moved between stages, extended or amended. Report files give dates, not times. | Report the start and end date of every step, including material delays |
| Participation and attrition | The participation report per round: in the round, submitted, completed, not submitted, withdrew, answers removed, and response rate. Results state who the partial-responses rule counted. | Explain denominator definitions and how attrition may have changed the result |
| Feedback delivered | The Rounds & feedback settings, the comment-sharing choice for each round, and the preview of what panelists will see | Describe quantitative and qualitative content and whether feedback was anonymized |
| Item evolution | Permanent statement IDs, revisions with the results they cite, the decision log's Advance, Revise and Hold, and the round registers locked before each round | State what changed or was removed, why, and between which rounds |
| Consensus results | Each round's results version: zone shares, median, Q1, Q3, IQR, the disagreement index on 1–9, the status and its reason, and Kendall's W. The consensus CSV adds the rule outcome and the decision, and the final result on the last round. | Report every round—not only the final item list—and distinguish consensus from concordance and stability |
| Dissent and subgroup findings | Polarized and Group disagreement results, steering reviews, "Final: no consensus" statements, and comment coding by two blind coders and an adjudicator. Subgroup tables are not computed. | Preserve meaningful disagreement; do not describe non-consensus as panel failure |
| Overrides and deviations | Overrides with their reasons in the decision log, shown in the consensus report and linked from the consensus CSV; recorded amendments with their reasons; and superseding results versions with their reasons | Explain effects on interpretation rather than merely stating that an audit event exists |
| Stopping | The stopping advice, the response-rate notes, rounds planned against rounds run in the study summary, and any further round added as an amendment | State who made the final decision and why the study stopped when it did |
| Harms or unintended consequences | Withdrawals and answer removals, as counts | Add off-platform incidents, burdens, or consequences relevant to interpretation |
Manuscript-section checklist
Title and abstract
- Identify the work as a consensus exercise and name the method used.
- State the objective, participant groups, principal consensus definition, number of rounds, and main result.
- Avoid using “modified Delphi” as the only methodological description.
Introduction
- Explain why consensus was needed and why empirical evidence alone was insufficient.
- Identify the intended users and decisions the result is meant to support.
- State whether the goal was exploration, prioritization, appropriateness, definition, or validation.
Methods
- Describe the steering group and conflicts of interest.
- Describe protocol development, registration, ethics review, and piloting.
- Explain initial item sources and transformations.
- Report panel eligibility, recruitment routes, targets, and compensation.
- Define anonymity, feedback, scales, unable-to-rate handling, consensus, stability, stopping, and item-handling rules.
- Explain qualitative synthesis, AI use, human review gates, and any external tools.
Results
- Give dates for every round and analysis step.
- Provide participant flow and round-specific denominators.
- Report item additions, removals, revisions, and the reasons for them.
- Present distributions and item-level results for every round.
- Report feedback delivered, deviations, overrides, attrition, unresolved items, and meaningful dissent.
Discussion and declarations
- Explain applicability, missing perspectives, attrition, and other limitations.
- Distinguish panel consensus from empirical proof, population prevalence, and universal recommendation.
- Report funding, sponsor roles, conflicts, author contributions, and data/material availability.
The ACCORD explanation and elaboration provides detailed rationales and reporting examples for each checklist item.
Evidence sources in Delphi Studio
| Artifact | What it contributes | Important limitation |
|---|---|---|
| Locked Setup and the Thresholds table | The protocol as it stood at launch, and every research threshold with whether it changes an outcome by itself | Does not explain why investigators selected each choice |
| Round approval | The principal investigator's decision on a round's content, rules, consent version and schedule; any change makes it stale | Records that the round was approved, not why it was designed that way |
| Results versions | Sealed per-round results with the rules fingerprint and the responses they were computed from; a replacement keeps the earlier version and states why | Does not make the rule methodologically appropriate by itself |
| Decision log | Each round's confirmed plan, review decisions and overrides with reasons, append-only and hash-chained | A record of decisions, not the discussion behind them |
| Statement IDs and revision evidence | Wording changes, and the results each revision cites | Investigators must explain the substantive rationale |
| Study audit CSV | Who acted (pseudonymously), when, and on what, for the study | A system chronology is not a narrative account of conduct |
| Consensus, participation and responses CSVs | Item results with rule outcome and decision, per-round participation counts, and one row per answer by panelist number, with nothing below the reporting floor | No comment text; a data dictionary and manuscript-specific table design may still be needed |
| Study summary (.md) and outline (.csv) | The study record, its rounds, rounds planned against run, and current results as counts | Counts only: no statement wording and no responses |
Nothing in the platform drafts manuscript text. Off-platform incidents and deviations are not recorded by the platform and must be kept by the team.
Recommended workflow
- Before launch: create an ACCORD checklist working copy and map every applicable item to a platform field, protocol section, or named investigator-owned document.
- At launch: review the Thresholds table and the locked Setup sections, and save the protocol or registration citation outside the platform where appropriate.
- After each round: review denominators, feedback, item changes and overrides, and note any deviation or incident outside the platform while details are fresh.
- At study close: download the consensus, participation and responses files, the study summary and the study audit CSV; verify that unresolved disagreement and attrition are represented.
- During writing: complete the ACCORD checklist against manuscript page or line numbers. Use CREDES to review Delphi-specific conduct details.
- Before submission: have a researcher who did not draft the first version reconcile the manuscript against the frozen rules, round history, and exported evidence.