Truace
Subscribe

Artificial Intelligence-Driven Clinical Decision Support for Postoperative Delirium Prevention: A Pragmatic Randomized Controlled Trial

Background Postoperative delirium affects up to 50% of older adults with cognitive impairment, yet preoperative identification of at-risk patients remains limited. Artificial intelligence methods, such as natural language processing, can extract cognitive impairment signals from electronic health records and enable targeted clinical decision support interventions to promote adherence to perioperative best practices and reduce postoperative delirium. We aimed to evaluate the effectiveness of an artificial intelli…

By the Truvace desk2 min read1 sourcePeer-reviewedTRV-2026-1664· cite & verify
Artificial Intelligence-Driven Clinical Decision Support for Postoperative Delirium Prevention: A Pragmatic Randomized Controlled Trial
USAF medical team provides healthcare to Peruvian patients (8467770) by U.S. Air Force photo by Tech. Sgt. Devin Boyer. Public domain

In brief

Researchers at The Mount Sinai Hospital in New York City ran a pragmatic randomized trial in routine care. Surgical patients flagged as cognitively impaired just before surgery were assigned either to an AI-driven alert in the electronic health record for anesthesia teams or to usual perioperative care. Impairment was identified by a natural language processing algorithm, Mini-Mental State Examination scores, or diagnostic codes.

The result matters because postoperative delirium is common in older adults with cognitive impairment, and AI-guided prompts are a proposed way to improve prevention. In this trial the tool did not shift outcomes or clinician behavior measurably. Yet the evidence is inconclusive about whether the approach works, because most patients had no documented delirium assessment, and how to use such tools effectively remains open.

Main points

  1. The trial randomized 7294 patients with preoperative cognitive impairment: 3624 to the intervention and 3670 to control.
  2. Only 1255 patients (17.2%) had documented postoperative delirium assessments, and 426 of them (34%) developed delirium.
  3. Delirium incidence was 35.1% (225/641) in the intervention group and 32.7% (201/614) in the control group, P = .37.
  4. Adherence to 12 perioperative best practices showed no significant differences between groups (all P > .05).

The problem

An AI-triggered, non-interruptive alert in the health record for surgical patients with cognitive impairment showed no significant effect on delirium incidence or on anesthesia teams' adherence to best practices.

The rundown

The study ran from June 20, 2023, to August 31, 2024, and was embedded in routine clinical care. The intervention was non-interruptive, meaning it was presented in the electronic health record rather than as a disruptive alert, and it was aimed at anesthesia teams.

The primary outcome was delirium within seven days after surgery, assessed with the 4 A's Test. Secondary outcomes covered adherence to 12 perioperative best practices, and analyses used chi-square tests.

The authors attribute the sparse outcome data to variability in routine delirium screening practices. They conclude that further work is needed to determine how best to use these tools effectively.

Sources

  1. Peer-reviewedAnesthesia & Analgesia2026-10-09

The debate