What we got wrong

Fifteen entries carried titles that could not be checked

An early sourcing pass recorded topics and approximate titles rather than verified citations, and never returned to complete them. A full audit of every entry typed as a paper that carried neither a DOI nor a source URL — fifteen rows — found four distinct defects.

Two titles had drifted from the published record. One was recorded as a study of automated external defibrillators “at large public venues”; the real paper studies their use by a U.S. airline. That is not a cosmetic difference — anything citing that row as evidence for venue AED provision was leaning on it harder than the study supports. Both titles have been corrected to the published record with their real identifiers attached.

One paper was attributed to the wrong author. A 2011 study of alcohol-related presentations at collegiate football games was credited to an author with no publications on the subject. The real work is Shook & Hiestand, Journal of American College Health 2011;59(5):388–92. This is the same signature as the fabricated citation this log already records: one work’s title carrying another’s authorship. Corrected, with the DOI attached.

Nine entries were retired. Two named authors whose complete publication records leave no room for them. Two were placeholders — topic descriptions with parenthetical glosses, never citations at all. Five could not be matched to any record in PubMed or Crossref. Retirement is reversible: each row remains visible and will be restored if its source is identified.

One was not a defect at all. A CDC MMWR report flagged by the audit is entirely genuine; it carries no DOI because MMWR does not issue them. A link has been added. It is recorded here because it is the reason this audit was done row by row rather than by rule — a blanket sweep would have wrongly impugned it. One further entry was a municipal regulation miscategorised as a journal article, and has been retyped.

Root cause. The publication gate checked verification status and physician review, but never checked that an entry typed as a paper carried any resolvable identifier. A row with no DOI and no URL could sit in the library indefinitely without tripping anything. Every defect in this audit passed through that gap, and so did the fabricated citation recorded above it.

Closed. Schema migration 023 blocks publication of any paper lacking a DOI or source URL, enforced in the database rather than by convention, and exposes remaining unresolvable rows through a public view. The gate was tested in both directions — a publish attempt without an identifier aborts; the same attempt with one succeeds — before this entry was written. The count of unresolvable papers went from nine to zero. Three genuine papers surfaced by the audit were added to the library.

A funding statute cited as authority for a clinical requirement

The published District of Columbia row cited D.C. Code § 1-325.81 — a FEMS special-events fee fund — as the authority for a medical requirement, and at the same time understated what the District requires, omitting its physician-staffing thresholds entirely. Both errors came from the same cause: the row's cells were machine-extracted from the MSETG PDF and never read by a person. The pending flag was correct and was not acted on before the row was published.

How found: the row was read against three primary sources — the DC Health HSPA special-events service page, the 2026 MSETG Special Events Planning Guide (p.31), and the D.C. Law Library. Action: the fee-fund citation removed; the physician thresholds added (physician-staffed aid station required at ≥15,001 attendees for athletic events and ≥50,001 for concerts, parades and conventions); the row is now verified against its official hosts. The former special-events licensing statute, D.C. Code § 47-2826, was found repealed effective 1 October 2025.

Author lists left empty while the registry record held them

Thirteen entries stored no authors at all despite carrying a complete registry record that listed them. Eleven were additionally slugged anon-, presenting as anonymous works whose authorship the library already possessed — including a consensus statement attributed to no one when its authors were on file.

Action: author lists restored from the stored registry records. Slugs retained to preserve existing links; display names corrected.

Wrong years, wrong first authors, wrong journals

A cluster of entries carried publication years off by one to three years, first authors who were not the first author, or journals belonging to a different article. Each was identifiable against material already held in the library.

Action: corrected individually, each against its evidence.

Ninety-two verification flags with no recorded basis

Ninety-two entries were flagged as identifier-verified, but the table that records when and how each verification occurred was empty. Ninety of the ninety-two hold a registry record that corroborates the flag; two do not. A verification you cannot audit is a claim, not a verification.

Action: the two uncorroborated flags cleared pending re-verification; the audit trail is now written on every verification, and re-verification of the full set is queued for when registry access is restored.

A non-existent identifier was carried on a published-adjacent row

This entry stored the DOI 10.3109/10903127.2015.1051680, which does not resolve: the Crossref API returns "Resource not found", and a bibliographic search on the full title surfaces only sibling NAEMSP resource documents, never this paper. The identifier had been carried on a draft row one review flag away from publication.

The paper itself is real — Schwartz B, Nafziger S, Milsten A, Luk J, Yancey A 2nd, Mass Gathering Medical Care: Resource Document for the NAEMSP Position Statement, Prehosp Emerg Care 2015;19(4):559-68, confirmed in PubMed as PMID 26270473 — and PubMed lists no DOI for it either. So the defect was a wrong identifier, not a fabricated citation: the entry is now url-only, not wrong.

How found: Crossref verification pass, 2026-08-11. Action: the unresolvable DOI was cleared rather than kept, and the entry pinned to its PubMed record; it is now url-only until a real DOI is confirmed with the publisher. No replacement DOI was invented.

Two previously reported defects did not survive scrutiny

An earlier pass reported these entries as carrying identifiers from the wrong publishing platform. Full comparison against their stored registry records corroborated every field, and the resource URLs in question are ordinary publisher deposits rather than evidence of a mismatched record. No foreign content had been ingested into either entry.

Recorded here because a corrections log that only ever confirms suspicions is not an audit. Findings that fail verification are findings.