Peer reviewed AI-assisted

Randomised trial of a risk model triaging patients with a positive antinuclear antibody test

At one US academic health system, 2,385 people with a new positive antinuclear antibody test were randomised to a pathway in which an electronic health record risk model invited those it scored high risk to expedited rheumatology evaluation, and those in the intervention arm reached an autoimmune disease diagnosis faster, with a median of 43 days against 138.

Model
An interpretable electronic health record risk model for systemic autoimmune disease, developed by the same group in 2024
Field
Medicine
Date
2026-11-06
Human collaborators
April Barnado, Ryan P. Moore, Daniel W. Byrne

What was found

A positive antinuclear antibody test is common and mostly uninformative, and no evidence-based rule existed for deciding which of these patients should see a rheumatologist. The model was built in 2024 from a chart review of 2,000 positive-antinuclear-antibody patients using demographics, billing codes for autoimmune symptoms and laboratory values, reaching an area under the curve of 0.83 in training and 0.75 in validation, and was then embedded in the health record. In this two-year pragmatic study every patient with an incident positive test was randomised 1:1 independently of their risk score; in the intervention arm high scores triggered an invitation to expedited rheumatology evaluation, while in the control arm scores were computed and not acted on. Both arms received standard care and could be referred as usual, and chart review every six months recorded evaluations and diagnoses. Of 2,385 randomised, 954 intervention and 1,006 control patients were analysed after postrandomisation exclusions, and time from positive test to diagnosis favoured the intervention arm with a hazard ratio of 1.9 (95% CI 1.3 to 2.7, P<0.001P < 0.001).

Novelty check

Europe PMC, 2026-11-09: TITLE:("antinuclear" OR "ANA") AND ("prediction model" OR "machine learning" OR "risk model") AND (rheumatology OR "autoimmune disease") returned 16 records. The only prior work on this exact model is the same group's 2024 development and validation paper in Frontiers in Immunology, which was retrospective and proposed deployment; the remaining records are antinuclear antibody pattern classification from immunofluorescence images, referral-practice surveys and unrelated prognostic models, with no other randomised evaluation. ClinicalTrials.gov, 2026-11-09: searches on the trial name, on Barnado and on positive antinuclear antibody with randomisation returned no matching registration, and the paper gives none, so this entry carries no trial registration.

Caveats and known objections

A single academic health system in Nashville, pragmatic and unblinded, with the model's developers running the trial, so there is no external validation and the result is not yet independent in this registry's sense. The endpoint is time to diagnosis, a care-pathway measure rather than a hard patient outcome: nothing here shows that faster diagnosis changed disease course, organ damage or treatment response. The secondary comparison that would speak to whether the model identified the right patients, the share diagnosed among those evaluated by rheumatology within six months, was 23% against 16% and did not reach significance (P=0.07P = 0.07). 425 of 2,385 randomised patients, about 18%, were dropped after randomisation and the abstract does not break down why. The trial was funded by the National Institutes of Health and others. Neither the model nor the underlying record data is accessible outside the institution. The journal lists the paper ahead of print. Autonomy is ai-assisted: the model scored and invited, and rheumatologists made every diagnosis.

No independent check is recorded on this entry yet.

Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.

Or on GitHub: submit a check challenge the grade send a correction or send a pull request

Entry history (1 event)
  1. AddedEntered the registry graded Peer reviewed and AI-assisted.

Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.

Graded peer reviewed for verification and ai-assisted for autonomy. What these mean.

Cite this entry

Plain text
whataifound.org. (2026). Randomised trial of a risk model triaging patients with a positive antinuclear antibody test. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-11-06-aura-ana-autoimmune-rct
BibTeX
@misc{whataifound-vanderbiltuniversitymedicalcenter-2026-rct,
  title        = {Randomised trial of a risk model triaging patients with a positive antinuclear antibody test},
  author       = {{whataifound.org}},
  year         = {2026},
  howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
  note         = {Result by Vanderbilt University Medical Center. Verification: Peer reviewed. Autonomy: AI-assisted.},
  url          = {https://whataifound.org/finding/2026-11-06-aura-ana-autoimmune-rct}
}

Related findings

← All medicine findings in the registry