The use of artificial intelligence models for interpretation and triage of clinical referral letters to acute and specialist care pathways: Protocol for a systematic review

James Foley, Enda Hession, Etimbuk Umana and Fiona Boland

PLOS ONE, 2026, vol. 21, issue 8, 1-10

Abstract: Background: Clinical referral letters direct patients into acute, urgent, and specialist care pathways, but their triage is typically manual, heterogeneous, and resource intensive, contributing to waiting-list pressure and potential patient harm. Artificial intelligence (AI) methods, including natural language processing, machine learning, and large language models, offer potential to automate or augment referral triage, yet the performance, safety, and implementation characteristics of AI applied to referral documentation have not been systematically synthesised. A February 2026 PROSPERO search identified no completed or ongoing systematic reviews addressing this question. Objectives: To evaluate the prioritisation performance of AI models used to triage clinical referral documentation for acute and specialist care pathways, and to map how this emerging field defines and evaluates AI-assisted referral triage, including model types, reference standards, validation practices, safety outcomes, and implementation-related outcomes. Methods: This protocol is reported in accordance with PRISMA-P. Systematic searches will be undertaken in MEDLINE, EMBASE, Web of Science, Scopus, CINAHL, and the Cochrane Central Register of Controlled Trials, covering January 2016 to the 5 May 2026, with no language restrictions. Eligible studies will include diagnostic accuracy, model development and validation, and comparative studies that apply an AI model to referral documentation and report at least one triage-performance outcome. Two reviewers will independently screen, extract data, and assess risk of bias using PROBAST-AI as appropriate. A structured narrative synthesis and evidence map using the SWiM reporting guideline will be the primary synthesis output, stratified by AI model class and clinical pathway. Meta-analysis will be treated as conditional and exploratory, and will be undertaken only where studies report comparable outcomes using a similar model class, clinical setting, reference standard, outcome definition, and threshold structure. Expected outcomes and significance: This review will produce the first systematic synthesis and evidence map of AI-assisted referral triage across acute and specialist care pathways, characterising how the field defines and evaluates this task as well as what is currently known about prioritisation performance. Findings will inform clinicians, service planners, and policymakers considering AI adoption in referral workflows, and will identify methodological priorities for future research.

Date: 2026
References: Add references at CitEc
Citations:

Downloads: (external link)
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0355476 (text/html)
https://journals.plos.org/plosone/article/file?id= ... 55476&type=printable (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:plo:pone00:0355476

DOI: 10.1371/journal.pone.0355476

Access Statistics for this article

More articles in PLOS ONE from Public Library of Science
Bibliographic data for series maintained by plosone ().