Three citation styles dominate medical and health-sciences writing: AMA (American Medical Association), NLM (National Library of Medicine, often called Vancouver style), and APA (American Psychological Association). AMA and NLM are both numbered systems built for biomedical publishing, while APA uses author-date parenthetical citations and shows up more often in psychology, nursing, and public health journals. The differences between them can look subtle on the surface, but a misplaced comma or a wrongly abbreviated journal title is enough to trigger a desk rejection at many journals, so the details matter more than you might expect.
The Two Numbered Styles and How They Differ
AMA and NLM are close relatives. Both assign a sequential number to each source the first time it appears, and both list references at the end of the paper in numbered order rather than alphabetically. The first edition of the AMA Manual of Style was published in 1962, and the guide has been revised repeatedly to keep pace with digital publishing and evolving standards.
1Science Editing. The AMA Manual of Style, 11th edition: the ultimate guide to scholarly publishing in the digital ageIn AMA style, in-text citations are superscript numbers placed after punctuation. A typical reference for a journal article looks like this: the author’s last name, followed by initials with no periods or spaces, then the article title in sentence case (no italics, no quotation marks), the abbreviated journal title in italics, the year, a semicolon, the volume number, the issue number in parentheses, a colon, and the page range. It is compact by design, squeezing a full citation into as little space as possible.
NLM style follows a similar skeleton but diverges in small ways. The NLM’s recommended format is based on the NISO Z39.29-2005 bibliographic reference standard, adapted specifically for NLM databases like PubMed and MEDLINE.
2PubMed Central. Samples of Formatted References for Authors of Journal ArticlesOne practical difference: NLM uses periods to separate major elements of a reference (authors. title. journal.), while AMA separates them with a mix of periods and punctuation that varies by element. NLM also italicizes journal abbreviations differently depending on the context, and its approach to capitalization of article titles can vary. If you have ever stared at two references that look almost identical and tried to figure out which style they follow, this is usually where the answer hides.
A survey of over 1,000 biomedical journals found that only about 7% use the current AMA or NLM/Vancouver format as written; roughly 39% use older versions or a variation, and another 40% have developed their own house style entirely.
3Learned Publishing. Citation rules through the eyes of biomedical journal editorsThat means even within medical publishing, you will encounter journals that loosely follow AMA or NLM but add their own tweaks. Always check the target journal’s author instructions before assuming a “standard” format will work.
When APA Shows Up in Medical Writing
APA style is the default in psychology, education, and much of the social sciences. It enters the medical world through fields that straddle both territories: nursing, public health, health psychology, occupational therapy, and speech-language pathology. If you are writing for a journal in one of those disciplines, there is a good chance it asks for APA rather than AMA or NLM.
The most obvious difference is the in-text citation format. Instead of a superscript number, APA uses the author’s last name and the year of publication in parentheses. For a journal article reference, APA lists authors (last name, then initials with periods), the publication year in parentheses, the article title in sentence case, the journal title in italics and title case, the volume in italics, the issue number in parentheses (not italicized), and the page range. APA also requires a DOI at the end when one is available, formatted as a hyperlink (https://doi.org/10.xxxx/xxxx).
Because APA embeds author names and years directly in the text, sentences read differently. A statement like “Smith et al. (2023) found that…” is standard APA. In AMA or NLM, the same sentence reads “Smith et al found that…14” with the superscript number doing the work. For readers skimming medical literature, the numbered style is denser on the page. An analysis of articles in PubMed Central found that roughly 86% of articles used numeric-style citations, while about 12% used author-date style, and numeric citations required far fewer characters per reference.
Formatting Author Names
Author formatting is where most people first stumble, because each style has different rules for punctuation, spacing, and when to cut off a long author list.
In AMA, you list up to six authors using the format “LastName AB” (last name followed by initials, no periods, no spaces between initials). If there are more than six, you list the first three and then write “et al.” For group authors or a combination of individual and group authors, the group name follows a semicolon after the individual authors.
NLM follows a similar initials-without-periods convention. The traditional NLM guidance was to list the first six authors and then add “et al.” when there were more.
4National Library of Medicine. Samples of Formatted References for Authors of Journal ArticlesHowever, the NLM now lists all authors in its own databases, and some journals following NLM style have adopted this practice. If you are submitting to a specific journal, check whether it uses the traditional six-author cutoff or the full-list approach.
APA, by contrast, lists up to 20 authors before truncating. Author names include periods after each initial (Smith, J. A., rather than Smith JA), and authors are separated by commas with an ampersand before the final name. The different punctuation might seem trivial, but automated reference checkers used by journal submission systems flag these mismatches, so getting the format right from the start saves you a revision cycle.
Journal Titles and Abbreviations
AMA and NLM both require abbreviated journal titles, and both follow the same abbreviation list maintained by the NLM (the one you can look up in the NLM Catalog or the PubMed journal database). So “The New England Journal of Medicine” becomes “N Engl J Med” in both styles, and “Journal of the American Medical Association” becomes “JAMA.” The abbreviations drop articles (“the,” “of”) and shorten common words (“Journal” to “J,” “Medicine” to “Med,” “American” to “Am”). If a journal has a single-word title, like “Nature” or “Science,” it is not abbreviated.
APA does not abbreviate journal titles at all. You write out the full name in title case and italicize it. This is one of the quickest ways to tell whether a reference list follows a medical style or APA: if you see abbreviated journal names, it is AMA or NLM. If every journal title is spelled out in full, it is APA (or one of a handful of other non-medical styles).
Handling DOIs and URLs
Digital object identifiers have become the standard way to link readers to the exact version of a paper you are citing. All three styles now require DOIs when available, but the formatting differs.
AMA’s 11th edition treats DOIs as the preferred identifier over URLs. The recommended format is “doi:10.xxxx/xxxx” without a period after it. If a source has both a DOI and a URL, you include only the DOI. The rule about not placing a period after a DOI or URL was a specific update in the 11th edition, designed to prevent the period from being mistakenly included in clickable links.
NLM style also uses DOIs but formats them slightly differently, typically as a full URL (https://doi.org/10.xxxx/xxxx) or in the “doi:” prefix format depending on the specific journal’s instructions. NLM references for online-only content often include an “Available from:” tag before the URL.
APA formats DOIs as clickable hyperlinks: https://doi.org/10.xxxx/xxxx. No “doi:” prefix, no “Retrieved from” language (that was the old APA 6th edition format). APA also requires a retrieval date only for sources that change over time, like a webpage that gets updated regularly. For stable sources like journal articles, the DOI alone is sufficient.
Preprints, Clinical Trials, and Other Edge Cases
Medical literature increasingly involves source types that did not exist when these citation styles were first created. Preprints, clinical trial registry entries, clinical practice guidelines, and datasets all require their own formatting, and each style handles them a bit differently.
For preprints (papers posted to servers like medRxiv or bioRxiv before peer review), the scholarly publishing community recommends including a clear “[Preprint]” label, the DOI if available, and standard bibliographic details so the reader knows they are looking at work that has not yet been peer-reviewed.
5Springer Nature Research Communities. Citing a PreprintIn AMA, you add the preprint server name and the word “Preprint” where the journal name would normally go. In APA, you list the preprint server as the source and include the DOI. The key in any style is transparency: your reader should be able to tell at a glance that the source was not peer-reviewed.
Clinical trial registry entries, such as records on ClinicalTrials.gov, come up frequently in medical writing. AMA style treats these like database entries, with the registry name and trial identifier number included in the reference. APA formats them similarly to reports, with the registry as the publisher.
Retracted papers present a particularly tricky citation scenario. You should generally avoid citing retracted work as if it were valid evidence, but sometimes you need to reference a retracted paper to discuss the retraction itself. In that case, all three styles expect you to include a note indicating the retraction. The stakes here are real: one study found that researchers identified over 700 instances of post-retraction citations that explicitly acknowledged a retraction, but plenty of others cited retracted work without any such acknowledgment.
6PubMed Central. Continued use of retracted papers: Temporal trends in citations and (lack of) awareness of retractions shown in citation contexts in biomedicineChecking the retraction status of sources before you cite them is a step that many authors skip but shouldn’t.
How Often References Go Wrong
If you have ever assumed that published papers in major journals must have flawless reference lists, the data will surprise you. Studies examining reference accuracy across scientific disciplines report error rates ranging from 25% to 54%.
7PubMed Central. Manuscript Referencing Errors and Their Impact on Shaping Current EvidenceThese are not just typos in page numbers. A review of reference accuracy studies found that about 15% of all analyzed citations contained quotation errors, meaning the cited source did not actually support the claim the author used it for. Roughly two-thirds of those content errors were classified as major, meaning they meaningfully changed the original source’s findings or conclusions.
8PubMed Central. Accuracy of cited “facts” in medical research articles: A review of study methodology and recalculation of quotation error rateA separate study that verified nearly 1,500 references found an overall error rate of about 21%, with author names and article titles being the most frequent types of mistakes.
9Journal of Librarianship and Information Science. Reference accuracy of primary studies about reference accuracy: A descriptive studySome of these errors cascade: if you copy a reference from another paper’s bibliography instead of verifying it against the original source, you inherit whatever mistakes the earlier author introduced. That is how “citation chains” form, where the same wrong volume number or misspelled author name propagates through dozens of papers over years.
The practical takeaway is unglamorous but important: verify every reference against the original source. Do not copy references from other papers’ bibliographies, even in the same journal. Pull up the article itself, confirm the authors, title, journal, year, volume, issue, and pages. It takes an extra few minutes per reference, but it is the single most effective way to avoid being part of the error statistics.
Reference Management Software
Formatting references manually is tedious, and most researchers use software to handle it. Tools like Zotero, Mendeley, EndNote, and RefWorks can store your references and automatically format them in AMA, NLM, APA, or hundreds of other styles. But “automatically” does not mean “perfectly.”
A study comparing the accuracy of references generated by four major reference managers (EndNote, Mendeley, RefWorks, and Zotero) against 15 biomedical journal styles and the NLM citation style found that Zotero produced the fewest errors for 11 of the journals and for NLM style overall. Mendeley came in second. EndNote had the most errors for 9 of the journals, and RefWorks had the most for the remaining 4. All four programs struggled with formatting URLs and access dates for online sources.
10The Journal of Academic Librarianship. Comparison of the Accuracy of Bibliographical References Generated for Medical Citation Styles by EndNote, Mendeley, RefWorks and ZoteroSome of the errors came from technical limitations of the software itself, while others resulted from incorrect settings in the citation style files. This means that even with a reference manager, you still need to review the output. A good workflow is to let the software generate your reference list, then manually check each entry against the original source and the target journal’s formatting instructions. Pay special attention to the elements that software handles worst: online source URLs, access dates, and the punctuation between reference elements.
Zotero has the added advantage of being free and open-source, with a large community maintaining its citation style repository. If a journal updates its formatting rules, the community-maintained style file for that journal usually gets updated within a few weeks. Mendeley is free for basic use but owned by Elsevier, which matters to some researchers who prefer not to have their library data tied to a publisher. EndNote is a paid product with strong institutional support but, based on the accuracy data, does not justify its cost through superior reference formatting.
Choosing the Right Style for Your Situation
If you are submitting to a specific journal, the choice is already made: use whatever the journal’s author guidelines specify. The question of which style to use only matters when you have flexibility, such as in a thesis, a class assignment, or a report for a clinical department.
For clinical medicine and basic biomedical science, AMA is the safest default. It is the style of JAMA and its network journals, and it is widely recognized across clinical specialties. For work that will interface with PubMed or the broader NLM ecosystem, NLM/Vancouver is a natural fit, especially if you are writing a systematic review or contributing to a database-indexed publication. For nursing, public health, health psychology, or any field where the literature draws heavily on social science methodology, APA is standard.
When in doubt, check the most prestigious journal in your target field and see what it uses. That will tell you the community expectation. And regardless of which style you end up using, the underlying principles are the same: give the reader enough information to find the exact source, format it consistently, and make sure every citation actually supports the claim you attached it to.
AI-Generated Citations and Emerging Concerns
A newer challenge that did not exist when the current editions of AMA, NLM, or APA were written is the use of AI tools to generate or format references. Large language models can produce citation text that looks plausible but points to papers that do not exist, a phenomenon sometimes called “hallucinated references.” The bibliographic details, including author names, journal titles, and DOIs, can look entirely realistic while being fabricated.
If you use an AI tool to help draft a paper or organize references, treat every reference it produces as unverified until you have confirmed the source exists and says what the AI claims it says. This is not a hypothetical risk: multiple journals have flagged submissions containing fabricated references traced to AI assistance. The same verify-against-the-original discipline that protects against citation chains applies here, but with even more urgency because the AI-generated reference may not correspond to any real publication at all.
Several journals have begun requiring authors to disclose AI tool use in their methods or acknowledgments sections. The AMA, NLM, and APA style organizations are still working out formal guidance on how to cite AI-generated content when it is used as a source (as opposed to using AI as a drafting tool). For now, the safest approach is to use AI for efficiency in organizing your references, but never to trust its output without verification, and to follow whatever disclosure policies your target journal has adopted.