8.4 Post-Publication Metrics and Correcting the Record
Key Takeaways
Citation metrics, journal metrics, attention metrics, usage, policy citations, and practice indicators answer different questions and should not be treated as interchangeable measures of quality or patient impact.
Publishers use correction terminology differently; “erratum” and “corrigendum” do not have one universal publisher-versus-author definition, so follow the journal’s policy and describe the error and effect clearly.
Minor errors that do not change the record may need no notice, while errors that affect interpretation, attribution, or reproducibility generally require a transparent, linked correction under the venue’s terminology.
An expression of concern alerts readers when serious doubt exists but evidence or an investigation is incomplete; it is not a finding of misconduct.
Retraction is appropriate when findings are unreliable, publication is redundant without proper justification, plagiarism is substantial, unethical research is reported, or other COPE grounds apply; the notice should remain linked and explain the basis.
Post-Publication Metrics and Correcting the Record
Publication work continues after release. Teams may evaluate reach, answer reader questions, propagate corrections across versions, and respond to concerns about data or attribution. The goal is not reputation management; it is an accurate, usable scholarly record.
Metrics answer different questions
Citations indicate that later scholarly works referenced an item, not whether the citing work agreed with it. Counts vary by database, discipline, article age, document type, and citation practice. The Journal Impact Factor is a journal-level average over a defined window and should not be used as the quality score of an article or author. An h-index combines an individual's output and citations but is strongly affected by career length, field, authorship practices, and database coverage.
Article-level and field-normalized indicators may help comparison, but their methods and reference sets must be understood. Usage measures such as views and downloads show access events, not comprehension or practice change. Altmetric Attention Score and related indicators aggregate selected online attention—news, social posts, blogs, policy sources, and other channels—with platform-specific weighting and coverage. Attention can be favorable, critical, automated, or driven by controversy.
Build a measurement plan around the question:
- Discovery: indexing, search impressions, link referrals.
- Use: abstract views, full-text views, downloads, video completion.
- Scholarly uptake: citations and citation context.
- Public or professional attention: news, social, guideline, or policy mentions.
- Educational outcome: audience-appropriate knowledge or behavior measures, when a valid evaluation exists.
Do not claim patient benefit, prescribing change, or scientific quality from a high attention score. Report source, definition, time window, denominator where relevant, and limitations.
Triage a possible error
When an error is reported, preserve the evidence and assess it promptly. Ask:
- What version and exact content are affected?
- Is the issue typographic, factual, analytical, interpretive, attributional, ethical, or related to data integrity?
- Does it change any numbers, figures, conclusions, safety information, authorship, or reader action?
- Which authors, institution, sponsor, editor, or data owner must be involved?
- Are related abstracts, repositories, preprints, plain-language summaries, digital assets, or derivative materials affected?
Do not silently replace the version of record. Contact the journal with a factual description, corrected material, supporting verification, and an assessment of impact. The editor or publisher determines the publication mechanism under its policy.
Correction terminology varies
Some publishers use erratum for a publisher-introduced error and corrigendum for an author error; others use the words differently or publish all changes as a correction. COPE guidance focuses on transparent correction of the literature rather than enforcing a universal vocabulary.
A transposed table column that leaves the primary conclusions intact will usually call for a correction, but the journal decides its label and format. The notice should identify what changed, preserve the original record as appropriate, and be linked bidirectionally to the article. A harmless spelling error may not warrant a formal notice; a wrong unit, denominator, conflict disclosure, author list, or analysis may.
An omitted qualifying author is not always resolved by a document specifically called a “corrigendum.” The journal normally investigates the requested authorship change, obtains explanations and agreement as its policy requires, may involve institutions in a dispute, and publishes the appropriate linked notice if the byline changes.
Expressions of concern
An expression of concern can alert readers when serious questions exist but a conclusive determination is delayed or unavailable—for example, an institutional inquiry is ongoing, evidence is inconclusive, or an investigation appears procedurally inadequate. It should state the concern and be linked to the article. It is not a declaration that misconduct occurred and should be updated when the matter is resolved.
Retraction
Retraction protects the literature when an article's findings are unreliable because of major error or misconduct, or when other serious grounds apply, such as substantial plagiarism, unjustified redundant publication, unethical research, or compromised peer review. A retraction is not limited to intentional misconduct; an honest analysis error can make central findings unreliable.
A retraction notice should be clearly identified, freely accessible where possible, linked to the retracted article, published promptly, and state who is retracting and why. The article generally remains available but marked as retracted so the history is not erased. Authors may disagree with a retraction; the editor can still act on sufficient evidence.
A failed replication alone does not automatically require retraction. Differences in population, methods, uncertainty, or chance may explain it. Likewise, scientific disagreement belongs in correspondence or further research unless evidence shows that the original record is unreliable.
Propagate the correction
After an editorial action, update the publication tracker and notify owners of repositories, preprints, congress materials, PLSPs, digital assets, internal evidence summaries, and data citations as appropriate. Do not edit independent third-party posts, but provide the authoritative notice. Monitor bibliographic databases for linkage and preserve the audit trail.
Important
Choose the mechanism based on the error's effect and journal policy, not on a memorized erratum/corrigendum label. Escalate serious integrity questions and correct the record transparently.
A table contains transposed baseline percentages, but verified primary efficacy and safety results are unchanged. What should the team do?
Quietly replace the PDF without a notice.
Notify the journal with the verified correction and impact assessment so it can issue and link the appropriate correction under its policy.
Retract the article automatically because every numerical error invalidates the study.
Wait until the next manuscript to publish the right table.
Which finding most clearly supports retraction under COPE principles?
A reader disagrees with interpretation of a secondary endpoint.
An independent trial estimates a smaller effect.
Verified fabrication in central datasets renders the article’s main findings unreliable.
A typographic error appears in an author affiliation.
How does an online attention score differ from Journal Impact Factor or an author h-index?
It proves the article changed clinical practice.
It counts only favorable social-media comments.
It is a fixed measure of methodological quality.
It aggregates selected online attention to an item, while JIF is a journal-level citation metric and h-index combines an individual’s output and citations; none alone proves quality or impact on care.
Sections you finish are checked off in the contents.