Sources & methods

Clinical consequence comes first.

EM Weekly asks: what did an emergency physician need to know this week? Selection favors findings that could affect treatment, testing, resuscitation, medication choice, or disposition. Journal prestige is not the primary filter.

Where we look

Discovery covers PubMed-indexed emergency medicine, general medicine, and relevant specialty journals. JournalFeed can help identify candidates; every included item must be verified against its primary paper, abstract, journal page, or official society source.

A separate guideline search checks ACEP, AHA/ASA, ACC, AAP, ACS, ENA, SAEM, AAEM, ASH, CHEST, SCCM, IDSA, toxicology, trauma, and relevant international societies. A final cross-check seeks major trials and guidance that journal searches may have missed.

What qualifies as new

Weekly issues include every article added to EM Weekly during the preceding Monday through Sunday in America/Los_Angeles. Each study’s first-online date remains visible and determines whether its evidence is new. Print issue dates do not make old papers new. An important missed item can appear once with an explicit catch-up label. Historical sample records are excluded from normal weekly selection.

Selective by design

Daily selection favors important studies and major relevant guidelines. The weekly digest includes all articles published on the site during that week. Two strong items are better than a padded list. Case reports, weak retrospective series, low-impact AI reports, and narrative reviews without new guidance are usually excluded. A Watch List appears only when useful.

Ranking weights ED impact (5), design (2), novelty (1), and guideline relevance (2), with each scored 0–5. Pinning and editorial exclusions override ranking. These scores aid selection; they do not establish clinical importance or a standard of care.

How summaries are written

ChatGPT summarizes only retrieved source material. Every record distinguishes abstract review, full-text review, and source availability. Numerical results, uncertainty, populations, primary outcomes, adverse events, and major limitations are retained when supported. An observed association is not presented as causation, and a trial regimen is not automatically a new standard.

We do not bypass paywalls, reproduce full abstracts, or redistribute publisher PDFs. Links lead to legal primary sources. Fetched source content is treated as evidence, never as instructions to the publishing system.

Review, corrections & accountability

AI-generated records publish only after structural checks and remain labeled “Not yet human-reviewed” until an editor changes that status. Structural validation cannot certify scientific accuracy. Consult the original paper before applying findings.

Corrections remain visible on each article and in Git history. Report a concern by opening an issue in the public repository, citing the article and the specific correction.