Study designs
A study design says how the research inside a work was done: a randomized controlled trial, an observational study, a systematic review. There are seven, all taken from PubMed’s own list of study types, and a work can carry more than one. Use them to find every randomized trial on a topic, or to see how the evidence in a field splits between trials, observational studies and reviews. A study design’s OpenAlex ID looks like https://openalex.org/study-designs/randomized-controlled-trial; list all seven at api.openalex.org/study-designs.
About
Study design is not work type
A work’s type says what kind of document it is. Its study design says how the research inside it was done. An editorial has a type and no study design. A preprint that reports a randomized trial has type preprint and study design Randomized Controlled Trial. A systematic review has type review and study design Systematic Review.
PubMed draws the same line. Its publication types come in two halves: publication formats (Editorial, Letter, Review, Guideline and so on) and study characteristics (trials, observational studies, case reports, systematic reviews, protocols). OpenAlex already uses the formats half to help set type. Study designs are the study-characteristics half. So if you want editorials, letters, reviews or guidelines, filter on type, not here.
Where the values come from
Every value comes from automated tagging of the work’s title and abstract. A decision model reads the title, the venue and the abstract and picks a design. Only confident answers become values, and a second model checks every Randomized Controlled Trial: where it finds the work is a digest, commentary or later analysis of a trial published elsewhere, the work gets neither Randomized Controlled Trial nor Clinical Trial.
Precision comes first. Measured against expert judgment on a sample weighted to match the whole index, each value is right at least 96% of the time, and Randomized Controlled Trial at least 99% of the time. When the tagging is unsure, it leaves the value off, so it misses some real trials and reviews rather than tagging works that aren’t. It only ever assigns the seven values below; OpenAlex adds nothing to PubMed’s vocabulary. Every benchmark, the code and the models are public: openalex-study-designs.
We do not copy PubMed’s own tags. On the same works, PubMed’s tags follow the National Library of Medicine’s indexing rules, which often tag a paper with the design of the study it came from: its Randomized Controlled Trial tag also covers secondary analyses, validation studies and methods papers written for a trial, and our benchmark found it right 63% of the time against the definitions below. The PubMed tags each value corresponds to are listed in the Values table.
What no value means
Automated tagging runs only on works that have an abstract and a type that can report research, so works without an abstract have no study design: article, review, preprint, conference-paper, book-chapter, dissertation, report and data-paper. An empty study_designs list means the work was not tagged, or has none of these seven designs. It never means the work is not a study. A lab experiment, an animal study or a computational paper is research, and has none of these designs.
This matters for negation. filter=study_designs.id:!randomized-controlled-trial returns every work not tagged as a trial, including trials the tagging missed.
More than one value
A work can carry several designs, and parents come along. Every Randomized Controlled Trial is also a Clinical Trial, and every Meta-Analysis is also a Systematic Review. So study_designs.id:clinical-trial returns randomized and non-randomized trials together. Automated tagging gives a work one design plus its parent.
Values
| ID | Display name | Definition | PubMed tags included |
|---|---|---|---|
randomized-controlled-trial | Randomized Controlled Trial | A trial that assigns human participants, groups of participants or treatment periods to interventions by explicit randomization and reports its results; secondary and post-hoc analyses of a trial’s data do not count. | Randomized Controlled Trial; Randomized Controlled Trial, Veterinary; Pragmatic Clinical Trial; Equivalence Trial; Adaptive Clinical Trial |
clinical-trial | Clinical Trial | A study that prospectively assigns human participants to one or more interventions and reports the results, randomized or not. | Clinical Trial; Controlled Clinical Trial; Clinical Trial, Phase I, II, III and IV; Clinical Study; Clinical Trial, Veterinary. Also every randomized controlled trial |
observational-study | Observational Study | A study of human participants in which the investigators do not assign an intervention, such as a cohort, case-control, cross-sectional, registry or survey study. | Observational Study; Observational Study, Veterinary; Twin Study |
case-report | Case Report | A description of one patient, or a small series of about ten or fewer described one by one, with no comparison group. | Case Reports |
systematic-review | Systematic Review | A review that reports a systematic search of the literature and explicit criteria for selecting the studies it brings together. | Systematic Review. Also every meta-analysis |
meta-analysis | Meta-Analysis | A study that statistically pools quantitative results from several independent studies. | Meta-Analysis; Network Meta-Analysis |
study-protocol | Study Protocol | The published plan for a study, setting out its aims, design and methods before any results exist. | Clinical Trial Protocol |
The definitions paraphrase the MeSH scope notes for the matching publication types, tightened where automated tagging needs a clear line (stated randomization for trials, human participants, about ten patients for a case report). Other PubMed study tags (Scoping Review, and modifiers such as Comparative Study, Multicenter Study and Validation Study) are not included.
The live list is at api.openalex.org/study-designs.
Attributes
The top-level fields on a study design object. Attributes shared with other entities (id, display_name, works_count, created_date, updated_date) are documented once on Common attributes.
id
String. The OpenAlex ID for this design, e.g. https://openalex.org/study-designs/randomized-controlled-trial. The last segment is the value you filter on.
display_name
String. The design’s name, e.g. Randomized Controlled Trial.
description
String. What the design means, in a sentence (the definitions in Values).
pubmed_publication_types
List of strings. The PubMed publication types mapped into this design, e.g. Meta-Analysis and Network Meta-Analysis for Meta-Analysis.
works_count
Integer. How many works carry this design. Parents are counted, so Clinical Trial includes every randomized controlled trial. See Common attributes.
works_api_url
String. A ready-made Works API URL for every work with this design (filter=study_designs.id:<ID>).
created_date
String. When the design was added to OpenAlex (YYYY-MM-DD). See Common attributes.
updated_date
String. When the record last changed. See Common attributes.
In the API
The Study designs endpoint is at api.openalex.org/study-designs. Fetch one by ID, /study-designs/randomized-controlled-trial, or list all seven.
Study designs are most useful from the works side. Every work carries a study_designs list. Filter with study_designs.id, which takes the short value (randomized-controlled-trial), the namespaced form (study-designs/randomized-controlled-trial) or the full ID. Group with group_by=study_designs.id. See Filtering for the syntax and the endpoints index for every endpoint.
# All seven study designs
https://api.openalex.org/study-designs
# Randomized controlled trials published in 2025
https://api.openalex.org/works?filter=study_designs.id:randomized-controlled-trial,publication_year:2025
# Clinical trials that are not randomized
https://api.openalex.org/works?filter=study_designs.id:clinical-trial,study_designs.id:!randomized-controlled-trial
# How a result set splits across designs
https://api.openalex.org/works?filter=publication_year:2025&group_by=study_designs.idOn openalex.org, the same thing is the study design filter on works (example).
Last updated
View as Markdown