How this list is built
The publications list is assembled by software and screened against one rule. This page states the rule, the sources, and the number of works at every stage, including the ones that could not be settled.
The rule
A work is listed if the research used an Axivity accelerometer to collect data, or evaluated one as an instrument. Validation studies are the strongest entries rather than edge cases. Where a paper is used to answer a question, the device that produced the data is what counts, not the field it belongs to.
Naming the device is not enough. A review that tabulates which accelerometer each study used, a methods section listing alternatives a reader might consider, and a related-work paragraph citing someone else's study all mention the device without having used one. They are screened out.
Where the works come from
Europe PMC is searched for the manufacturer's name in full text. OpenAlex is searched for a union of quoted device phrases, because a single-word search there returns 1.46 million works led by a 1983 paper on the invisible axion. Records are deduplicated on DOI and on normalised title, since the same work reaches us twice often enough to matter.
Screening
Counts from the most recent refresh. Candidate records and published works are different units: the same work can enter as two records, as a preprint and a version of record or under two different keys, and near-duplicates are merged before anything is published.
| Candidate records discovered | 3,052 |
|---|---|
| Screened out | 916 |
| Query noise: the search engine matched something that is not the device | 764 |
| Published before the device existed | 137 |
| Not a research output: peer-review reports, front matter | 15 |
| Unresolved, and therefore not published | 449 |
| The device is named in a paper we read, but use is unproven | 201 |
| No readable copy exists, so nothing could be checked | 248 |
| Admitted | 1,687 |
| Withheld as retracted | −1 |
| Merged into another record as a near-duplicate | −97 |
| Unique works published | 1,589 |
A further 67 pairs of records are similar enough to be the same work but not similar enough to merge automatically. Nobody has adjudicated them, so a small number of the 1,589 may be duplicates of each other.
What backs each entry
Entries do not all rest on the same strength of evidence, so the list says which. The strongest is a person's judgement. Next is a sentence we fetched and read ourselves, from Europe PMC full text or the publisher's open-access PDF. Weakest is a full-text index reporting that such a sentence exists in a document we cannot read; that is a real limit and it applies to roughly a third of the list.
| Curated by a person | 79 |
|---|---|
| A sentence we fetched and read | 1,026 |
| A full-text index reports the sentence; we cannot read the document | 484 |
| Published works | 1,589 |
What is not here
Screening leaves works it cannot settle, and they are not published. Some name the device in a paper we have read without showing that the study used it. The rest have no readable copy at all, so nothing can be checked.
They stay out on purpose. Everything on the list rests on a positive finding, and an unresolved work rests on the absence of one. Repeated sampling of that group has found classes of error nobody predicted, so it is a category we decline to publish rather than one we publish with a caveat. Coverage is sacrificed for correctness, and the size of the sacrifice is the number on this page.
Corrections
If a study belongs here and is missing, or is listed and should not be, tell us and it will be fixed by hand. A ruling made by a person is permanent and is never overwritten by a later automatic run. Get in touch.