Pragmatic randomised trial of AI-triggered decision support for postoperative delirium finds no effect
In a pragmatic trial embedded in routine care at Mount Sinai, 7,294 surgical patients flagged as cognitively impaired by a natural-language-processing model, test scores or diagnosis codes were randomised to a non-interruptive decision-support nudge for anaesthesia teams or usual care, with no difference in postoperative delirium (35.1% against 32.7%, P = 0.37) or in adherence to 12 best practices, though fewer than one in five patients were ever assessed for delirium.
- Model
- Natural-language-processing model for cognitive impairment in health records, not named in the abstract
- Field
- Medicine
- Date
- 2026-10-09
- Human collaborators
- Danielle Scharp, Ira Hofer
What was found
Patients with preoperative cognitive impairment, identified immediately before surgery by a validated natural-language-processing algorithm over the electronic health record, Mini-Mental State Examination scores or diagnostic codes, were randomised between June 2023 and August 2024 either to a non-interruptive decision-support intervention in the record, prompting anaesthesia teams toward perioperative best practices, or to routine care. The primary outcome was delirium within seven days on the 4 A's Test. Only 1,255 of 7,294 patients (17.2%) had a documented assessment, of whom 34% developed delirium, with no difference between arms, and none of the 12 best-practice adherence measures differed either. The authors conclude that inconsistent routine delirium screening limits what the trial can say about whether the intervention works.
Novelty check
Europe PMC, delirium in the title with decision support, a randomised design and machine learning, natural language processing or artificial intelligence, to 2026-10-20: this trial, a 2024 cluster-randomised trial of nurse-led AI-assisted delirium prevention in intensive care (AI-AntiDelirium), and protocols and prediction studies. No earlier randomised trial of NLP-triggered perioperative decision support was found.
Caveats and known objections
A null result that cannot be read as a clean negative: delirium was assessed in only 17% of randomised patients, assessment depended on routine screening rather than the trial, and the primary outcome is missing for most of the sample and could be missing differently by arm. The registration (NCT05777187) lists the study as terminated, giving low enrolment as the reason, with 7,412 enrolled. Single centre in New York and pragmatic, with the National Institute on Aging as a collaborator per the registration; the investigators built the algorithm and ran the trial, and no commercial vendor is named. The AI component is the identification step; the intervention itself is a passive reminder, so this tests a nudge triggered by AI rather than an AI recommendation. Autonomy is ai-assisted: the model flagged patients and clinicians decided whether to act.
Trial registration
NCT05777187
That a study was registered, with the outcomes it declared and the date it declared them, so a published endpoint can be checked against the one planned. It does not certify the results, their analysis, or any clinical benefit; registry records are submitted by the sponsor and are not peer reviewed.
Nobody outside the lab has checked this 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
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
Entry history (1 event)
- 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
whataifound.org. (2026). Pragmatic randomised trial of AI-triggered decision support for postoperative delirium finds no effect. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2026-10-09-ai-cds-postoperative-delirium-rctBibTeX
@misc{whataifound-icahnschoolofmedicineatmountsinai-2026-rct,
title = {Pragmatic randomised trial of AI-triggered decision support for postoperative delirium finds no effect},
author = {{whataifound.org}},
year = {2026},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Icahn School of Medicine at Mount Sinai. Verification: Peer reviewed. Autonomy: AI-assisted.},
url = {https://whataifound.org/finding/2026-10-09-ai-cds-postoperative-delirium-rct}
}