Do misconduct retractions have fewer verified authors?
Written against the design standard. Shares its case definitions and reason mapping with P17.
1. The question
An ORCID identifier that a publisher has authenticated ties a byline to a person who logged in to claim it. A paper mill selling authorship slots, or inventing authors, has less reason to produce that link. The share of a paper’s authors carrying an authenticated ORCID is therefore a candidate integrity signal that is cheap to compute for every article ever published.
2. Prior work, and the search behind the novelty claim
Log key P18.
corpus regex: ORCID.{0,80}(retract|misconduct|integrity|paper mill)
PubMed: ORCID[tiab] AND (retract*[tiab] OR "paper mill*"[tiab]
OR misconduct[tiab] OR "research integrity"[tiab])Seven PubMed records, none in the corpus. The closest proposes recording retractions on ORCID profiles (PMID 38815635); another discusses author name ambiguity (PMID 38704656). None measures ORCID coverage as a marker of retraction risk.
3. Hypothesis
Articles retracted for misconduct-class reasons have a smaller share of authors with an authenticated ORCID than matched controls; articles retracted for honest error do not.
Null. The odds ratio per 10-point increase in authenticated share is 1 for misconduct cases.
What would falsify it. No association for misconduct, or an equal association for honest error, which would point to journal-level practice rather than to fraud.
4. Design and unit
Matched case-control, run separately for the two reason classes, as in P17.
5. Sample
Cases and reason classes as defined in P17, from Retraction Watch with a DOI, published 2016 or later, when publisher ORCID deposit to Crossref became common. Controls: four per case, same journal and publication month, not in Retraction Watch (S9).
6. Exposure
The share of an article’s authors whose Crossref author record carries an ORCID with authenticated-orcid true. OpenAlex attaches ORCIDs by inference as well as from publisher deposit, and an inferred ORCID says nothing about whether the author claimed the paper, so OpenAlex is used only for a sensitivity analysis and never for the primary exposure (S3).
7. Outcome
Case or control status, by reason class.
8. Validation gate
The reason mapping is validated in P17. Here, the author checks for 100 articles that the Crossref record’s authenticated ORCIDs match the ORCIDs printed on the publisher page. Below 95 percent agreement, the exposure is recomputed from the source that agrees with the article.
9. Analysis
In R.
clogit(case ~ I(auth_share * 10) + log(n_authors) + strata(match_set), data = d)Fitted per reason class. Author count enters because a single-author paper has a share of 0 or 1 and large teams have intermediate shares for reasons unrelated to fraud.
10. Threats
| threat | handling | removed or disclosed | standard |
|---|---|---|---|
| Publishers differ in whether they deposit ORCIDs | Matching on journal and month holds the publisher constant | Removed | S9 |
| OpenAlex infers ORCIDs | Primary exposure uses only publisher-authenticated ORCIDs | Removed | S3 |
| Journals changed ORCID policy over time | Matching on month; window from 2016 | Disclosed | |
| Paper mills may adapt by registering ORCIDs | Would bias toward the null; the result is a snapshot of a moving target | Disclosed |
11. Probes
| probe | computes | goes to | what would change the design | status |
|---|---|---|---|---|
| PA | share of cases and of controls whose Crossref record carries any author ORCID, marginally | 5 | under 30 percent of articles with any deposit makes the exposure mostly zero | not run |
| PB | case counts per reason class from 2016 | 5 | fewer than 300 honest-error cases makes the contrast imprecise | not run |
| PC | the author’s 100 publisher-page checks | 8 | below 95 percent the exposure source changes | needs the author |
12. What this cannot show
- Whether requiring ORCIDs would prevent anything.
- Anything about fraud that has not been retracted.