JMIR Publications

ABSTRACT

Background:

Globally, Emergency Departments (EDs) use triage to ensure that the most acute patients are seen quickly and safely. This process is typically nurse-led and results in documentation that can be collected in large databases. Free text narratives generated as a part of this record have been previously studied for specific conditions but never reviewed in a comprehensive manner.

Objective:

The objective of this paper is to identify and map the academic literature that examines triage narratives. It describes the types of research conducted, identifies gaps in the research, and determines where additional review may be warranted.

Methods:

This paper uses a scoping review methodology. It maps the literature and describes what the data have been used for, what information is available on the form and structure of narratives, and highlights where there are similarities and opportunities for future research.

Results:

We screened 18,074 studies published between 1990 and 2022 in CINAHL, MEDLINE, Embase, Cochrane, and ProQuest Central. We identified 96 studies that directly examined the use of triage nurse narratives. There were over 12 million ED visits, drawn from over 63 million health records, generated by 2438 EDs included in the review. More than 80% (n=79) of these studies were performed in the United States (n=43), Australia (n=31), or Canada (n=5). Thirty-nine studies (41%), most of which were published after 2017, used machine learning to incorporate triage narratives into research. Triage narratives were grouped as being used in three ways: for case identification, as input variables for predictive modeling, and for quality improvement. Of the 96 included studies, thirty (31%) described triage narratives in some fashion: twenty-seven (28%) used keywords and seven (7%) offered more comprehensive discussion. There was inconsistent data reporting across studies and only eight (8%) declared using a reporting guideline.

Conclusions:

The breadth of studies identified suggest there is widespread routine collection of triage data, and that a significant number of studies examine triage narratives. These data have been used successfully to identify and describe cases rare disease and to perform disease surveillance for more common presentations. They have also been used to inform quality improvement and most recently as inputs for predictive modelling. Analyses of triage data have changed over tie and in 2017 machine learning became the prevalent method for examining triage narratives. Despite the common use of triage narratives as source data in studies these narratives, and the nurses who generate them, are poorly described in the literature and data reporting is inconsistent. Additional research is needed to describe the structure of triage narratives; to determine whether the use of triage narratives in improves the sensitivity of studies looking to describe disease prevalence, and to either develop triage specific data reporting guidelines, or improve the use of existing guidelines. While machine learning methods have been gaining popularity and have recently been incorporated into triage decision support tools, further research is required to determine how these models operate and whether they can be used prospectively to support clinical care.