From Triage to Discharge: A Survey of NLP Tasks, Methods, and Open Challenges in the Emergency Department

Computer Science > Computation and Language
[Submitted on 23 Aug 2026]
Title:From Triage to Discharge: A Survey of NLP Tasks, Methods, and Open Challenges in the Emergency Department
View PDF HTML (experimental)Abstract:Emergency departments (EDs) operate under time pressure, generating multimodal data such as clinical conversations, triage notes, and discharge documents. Recent advances in natural language processing (NLP), particularly pretrained transformers and large language models, have created new opportunities to support language and time-intensive stages of emergency care. Yet existing surveys map clinical NLP across the broader hospital workflow or focus on specific tasks. This survey analyses 46 papers spanning the three phases of ED: triage, diagnosis, and disposition, covering tasks such as triage classification, clinical summarisation, automatic diagnosis, report generation, and discharge documentation. We examine modelling paradigms, evaluation practices, and emerging benchmarks and shared tasks. Across tasks, we identify common trends, including a shift from task-specific neural architectures to pretrained language models, growing interest in interactive clinical systems, and increasing attention to clinically grounded evaluation. Finally, we detail open challenges such as limited generalisability, noisy clinical inputs, and workflow constraints that inform future ED-NLP research.
References & Citations
Loading...
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Facts Only

The survey covers 46 papers related to NLP tasks in the Emergency Department (ED).
Tasks analyzed include triage classification, clinical summarization, automatic diagnosis, report generation, and discharge documentation.
The analysis spans the three phases of ED: triage, diagnosis, and disposition.
Common trends observed include a shift from task-specific neural architectures to pretrained language models.
There is growing interest in interactive clinical systems.
Evaluation practices and emerging benchmarks are examined.
Open challenges identified are limited generalisability, noisy clinical inputs, and workflow constraints.

Executive Summary

Emerging advancements in Natural Language Processing, specifically pretrained transformers and large language models, are being applied to support time-intensive stages of emergency care within the Emergency Department (ED). This survey analyzes 46 research papers spanning the three phases of ED operations: triage, diagnosis, and disposition. The covered tasks include triage classification, clinical summarization, automatic diagnosis, report generation, and discharge documentation. The research examines different modeling paradigms, evaluation practices, and emerging benchmarks in this domain. A key trend identified is a shift from task-specific neural architectures toward the use of pretrained language models. Furthermore, there is growing interest in interactive clinical systems and methods that focus on clinically grounded evaluation. Open challenges discussed include limitations related to generalisability, handling noisy clinical inputs, and adhering to workflow constraints within the ED environment.

Full Take

The trajectory described reflects a fundamental tension between the scalable power of large language models and the high-stakes, constrained environment of emergency medicine. The move towards pretrained models suggests a recognition that complex clinical reasoning can be abstracted from task-specific training toward broader linguistic understanding, yet the explicit focus on workflow constraints and noisy data indicates that purely model-centric approaches risk divorcing computational efficiency from real-world clinical utility. The identified pattern of shifting modeling paradigms implies a necessary pivot from optimizing algorithmic performance in isolation to integrating operational context directly into the learning process. The challenge lies in ensuring that advancements in representation learning do not introduce opaque, unmanageable error modes when dealing with the inherent uncertainty and variability of patient-generated or recorded clinical language. Future research must address how to build systems that are not only accurate but also intrinsically sensitive to temporal dependencies and procedural constraints to bridge the gap between linguistic fluency and dependable clinical decision support.

Sentinel — Human

Confidence

This text appears to be a typical, well-structured academic survey abstract, exhibiting the logical coherence and density expected from human scholarly contribution.

Signals Detected

Moderate sentence length variance; the tone is academic and direct, consistent with a survey paper.

High internal logical flow; the argument progresses smoothly from problem statement to scope to findings/challenges.

Standard academic citation structure and abstract framing, which follows established patterns for survey papers.

No obvious markers of LLM confabulation or overly polished narrative; the content reads like a genuine research summary.

Human Indicators

The specific framing of a literature survey, citing a defined set of papers (46 papers) and delineating clear research challenges (limited generalisability, noisy inputs), suggests specialized domain expertise.

The structure is characteristic of academic writing rather than generalized synthetic prose.