@inproceedings{chen-etal-2026-self,
title = "Why Do Self-Harm Prediction Models Struggle to Generalise? {--} Lexical and Semantic Variations in Emergency Department Triage Notes",
author = "Chen, Liuliu and
Conway, Mike and
Robinson, Jo and
Rozova, Vlada",
editor = "Zirikly, Aya and
Bar, Kfir and
MacAvaney, Sean and
Ireland, Molly and
Ophir, Yaakov and
Atzil-Slonim, Dana and
Varadarajan, Vasudha and
Bedrick, Steven and
Desmet, Bart",
booktitle = "Proceedings of the 10th Workshop on Computational Linguistics and Clinical Psychology ({CLP}sych 2026)",
month = jul,
year = "2026",
address = "San Diego, California, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.clpsych-1.31/",
pages = "383--388",
ISBN = "979-8-89176-421-7",
abstract = "Self-harm presentations to emergency departments (EDs) are strongly associated with higher suicide risk. NLP models have shown strong performance in detecting self-harm from triage notes within single hospitals, yet performance often declines across institutions. To examine potential causes, we compare ED triage notes from two hospitals by analyzing lexical characteristics, highly associated predictive features, and salient topics. Our results reveal variation in lexical expression and feature importance related to self-harm across hospitals, despite consistent core themes such as self-poisoning and self-injury. These documentation differences are associated with reduced cross-site performance. These findings provide insight into how institutional variation affects the identification of self-harm in clinical text and highlight potential methods to improve model generalisability."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chen-etal-2026-self">
<titleInfo>
<title>Why Do Self-Harm Prediction Models Struggle to Generalise? – Lexical and Semantic Variations in Emergency Department Triage Notes</title>
</titleInfo>
<name type="personal">
<namePart type="given">Liuliu</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mike</namePart>
<namePart type="family">Conway</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jo</namePart>
<namePart type="family">Robinson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vlada</namePart>
<namePart type="family">Rozova</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 10th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Aya</namePart>
<namePart type="family">Zirikly</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kfir</namePart>
<namePart type="family">Bar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sean</namePart>
<namePart type="family">MacAvaney</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Molly</namePart>
<namePart type="family">Ireland</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yaakov</namePart>
<namePart type="family">Ophir</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dana</namePart>
<namePart type="family">Atzil-Slonim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vasudha</namePart>
<namePart type="family">Varadarajan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Steven</namePart>
<namePart type="family">Bedrick</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bart</namePart>
<namePart type="family">Desmet</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-421-7</identifier>
</relatedItem>
<abstract>Self-harm presentations to emergency departments (EDs) are strongly associated with higher suicide risk. NLP models have shown strong performance in detecting self-harm from triage notes within single hospitals, yet performance often declines across institutions. To examine potential causes, we compare ED triage notes from two hospitals by analyzing lexical characteristics, highly associated predictive features, and salient topics. Our results reveal variation in lexical expression and feature importance related to self-harm across hospitals, despite consistent core themes such as self-poisoning and self-injury. These documentation differences are associated with reduced cross-site performance. These findings provide insight into how institutional variation affects the identification of self-harm in clinical text and highlight potential methods to improve model generalisability.</abstract>
<identifier type="citekey">chen-etal-2026-self</identifier>
<location>
<url>https://aclanthology.org/2026.clpsych-1.31/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>383</start>
<end>388</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Why Do Self-Harm Prediction Models Struggle to Generalise? – Lexical and Semantic Variations in Emergency Department Triage Notes
%A Chen, Liuliu
%A Conway, Mike
%A Robinson, Jo
%A Rozova, Vlada
%Y Zirikly, Aya
%Y Bar, Kfir
%Y MacAvaney, Sean
%Y Ireland, Molly
%Y Ophir, Yaakov
%Y Atzil-Slonim, Dana
%Y Varadarajan, Vasudha
%Y Bedrick, Steven
%Y Desmet, Bart
%S Proceedings of the 10th Workshop on Computational Linguistics and Clinical Psychology (CLPsych 2026)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, USA
%@ 979-8-89176-421-7
%F chen-etal-2026-self
%X Self-harm presentations to emergency departments (EDs) are strongly associated with higher suicide risk. NLP models have shown strong performance in detecting self-harm from triage notes within single hospitals, yet performance often declines across institutions. To examine potential causes, we compare ED triage notes from two hospitals by analyzing lexical characteristics, highly associated predictive features, and salient topics. Our results reveal variation in lexical expression and feature importance related to self-harm across hospitals, despite consistent core themes such as self-poisoning and self-injury. These documentation differences are associated with reduced cross-site performance. These findings provide insight into how institutional variation affects the identification of self-harm in clinical text and highlight potential methods to improve model generalisability.
%U https://aclanthology.org/2026.clpsych-1.31/
%P 383-388
Markdown (Informal)
[Why Do Self-Harm Prediction Models Struggle to Generalise? – Lexical and Semantic Variations in Emergency Department Triage Notes](https://aclanthology.org/2026.clpsych-1.31/) (Chen et al., CLPsych 2026)
ACL