Natural Language Processing Techniques to Detect Delirium in Hospitalized Patients from Clinical Notes: A Systematic Review
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Delirium is a serious and common condition in hospitalized patients, associated with increased morbidity, mortality, and healthcare costs. Early detection and management of delirium is crucial for improving patient outcomes. Clinical notes contain valuable information about a patient’s mental status that may not be captured by structured data alone. Natural language processing (NLP) techniques have the potential to automatically detect signs and symptoms of delirium from free-text clinical notes, which could aid in early identification and prompt treatment.
Objective
The objective of this systematic review is to summarize and critically appraise the existing research on NLP methods for detecting delirium in hospitalized patients from clinical notes.
Methods
The review will follow the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and use the PRISMA-P framework for protocol development. A comprehensive search of PubMed, Web of Science, Embase, CINAHL, MEDLINE, The Cochrane Library, PsycINFO, and Scopus will be conducted from each database’s inception to February 2025. The search strategy will include terms related to delirium, natural language processing, machine learning, and clinical notes. Two independent reviewers will screen titles, abstracts, and full texts for eligibility using Covidence software and extract data using a standardized form. Risk of bias will be assessed using the PROBAST tool for prediction studies and the TRIPOD checklist for reporting quality. A narrative synthesis of the findings will be provided.
Discussion
This review aims to consolidate the current evidence on NLP approaches for delirium detection from clinical notes, identify promising methods and limitations, and highlight areas for future research and development. The findings may guide the development of automated tools to enhance the early recognition and treatment of delirium in hospitalized patients. The protocol follows best practices for systematic reviews and will be openly disseminated.
PROSPERO Registration
CRD42025634871
