AI predicts PTSD risk based on stories about what brought patients to the emergency room
· Medical Xpressedited by Gaby Clark, reviewed by Robert Egan
Gaby Clark
Scientific Editor
Meet our editorial team
Behind our editorial process
Robert Egan
Senior Editor
Meet our editorial team
Behind our editorial process Editors' notes
This article has been reviewed according to Science X's editorial process and policies. Editors have highlighted the following attributes while ensuring the content's credibility:
fact-checked
peer-reviewed publication
trusted source
proofread
The GIST Add to Preferred Sources
An AI tool can analyze patients' narratives about injuries that sent them to the emergency room to predict who will later develop posttraumatic stress disorder (PTSD), a new study found.
About 43 million patients are admitted to a U.S. emergency department (ED) each year with physical injuries (trauma), and more than a quarter of them develop PTSD, according to past studies. That makes the ED a checkpoint for catching the condition early, but clinicians face time constraints severe enough to rule out the use of standard screening questionnaires.
Patient narratives reveal early warning signs
Led by NYU Langone Health researchers, the new study found that the team's set of AI models—each analyzing the trauma narratives of 145 ED patients—correctly identified more than 95.4% of at-risk patients. In a head-to-head test, the AI tool, called PREVIA, flagged 100% of 30 cases from the same patient group that went on to develop PTSD, while an ED clinician specializing in PTSD correctly identified 89% of these cases.
"This is the first study to show that a patient's spontaneous answers to a single, routine care question—'What brought you to the ED?'—carry enough of a psychological signal to forecast their PTSD risk a month later," said senior study author Katharina Schultebraucks, Ph.D., associate professor in the departments of psychiatry and population health at NYU Langone. "PREVIA can serve as an AI screener that identifies at-risk patients and explains why, without adding any time to intake in busy emergency rooms."
Published online Nov. 9 in Nature Communications, the new work examined the capabilities of generative AI, which predicts options for the next word in a sentence based on how billions of people used that word in context. This sensitivity to language lets the models analyze a person's description of what happened to them to find patterns linked to their risk of developing PTSD, the researchers said. The patients' answers were recorded and transcribed for analysis.
A narrow window for early screening
"The emergency department is often the only point of contact a trauma survivor has with the health care system before symptoms become much harder to treat," added Schultebraucks, who is also co-director of computational psychiatry at NYU Grossman School of Medicine. "By identifying at-risk patients at that first encounter, PREVIA could open a window for timely care approaches, when preventive intervention is most effective."
The first month after trauma is a key window for finding those at risk for PTSD because many patients are no longer in contact with the hospital after that, the researchers said. Current questionnaires—such as the Peritraumatic Dissociative Experiences Questionnaire and the Posttraumatic Adjustment Scale—effectively forecast PTSD development, yet only about 28% of EDs routinely screen for PTSD at all, and only two-thirds of those use validated tools.
This likely reflects overcrowding and limited clinician time, typically about 15–20 minutes per patient, which forces clinicians to prioritize immediate medical concerns over lengthy mental health assessments, the study authors said.
Layered models weigh risk and protection
PREVIA was designed as a pipeline of steps that use different AI tools based on their strengths (e.g., the Llama model tends to emphasize signs of PTSD risk, while Mistral highlights protective and contextual factors that could buffer against it).
Specifically, the tool's first layer scans the language of a patient's narrative for symptoms that happen during the initial trauma, like loss of control or amnesia, as well as early symptoms that show up while a person is in the ED, such as feeling disconnected from the body. Both types of symptoms have been linked to later PTSD development.
The second layer of the team's AI tool also weighs risk factors such as preexisting mental health problems or lower socioeconomic status, alongside protective factors such as coping skills or strong social support. A third AI model synthesizes these assessments into a structured summary, reducing the chance that any single model dominates the results. A final layer combines that summary with demographic information (e.g., age, sex), trauma type and known PTSD prevalence rates to produce a risk score with a plain-language explanation.
Experiments confirmed that this three-layer design consistently outperformed simpler configurations across most performance metrics. Unlike AI that succeeds or fails for reasons unknown, PREVIA explains its reasoning and grounds predictions in established PTSD risk factors. Results with the tool were also validated by comparing them to diagnoses made for the same cases using the Clinician-Administered PTSD Scale for DSM-5 (CAPS-5), the gold standard clinical interview for diagnosing PTSD.
"We built PREVIA on open-source models with no training or fine-tuning required," said Schultebraucks. "That means other hospitals can test it independently, without access to proprietary software or our patient data, which is essential for widespread adoption."
Along with Schultebraucks, authors of the study from NYU Langone were Victoria Mueller, Jyoti Yadav, Annie Farnsworth, Charlotte Hilberdink, Sapir Gershov and Arieh Shalev in the Department of Psychiatry, Eric Oermann in the Department of Neurosurgery, and Stephen Wall in the Ronald O. Perelman Department of Emergency Medicine.
Publication details
PREVIA: A Clinically-Informed LLM Ensemble for Early PTSD Risk Prediction from ED Trauma Narratives, Nature Communications (2026).
Journal information: Nature Communications
Key medical concepts
Post-Traumatic Stress Disorder
Clinical categories
PsychiatryPsychology & Mental healthEmergency medicine Provided by NYU Langone Health Who's behind this story?
Gaby Clark
MA in English, copy editor since 2021 with experience in higher education and health content. Dedicated to trustworthy science news. Full profile →
Robert Egan
Bachelor's in mathematical biology, Master's in creative writing. Well-traveled with unique perspectives on science and language. Full profile →
Citation: AI predicts PTSD risk based on stories about what brought patients to the emergency room (2026, November 9) retrieved 9 November 2026 from https://medicalxpress.com/news/2026-11-ai-ptsd-based-stories-brought.html This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.