Truace
Subscribe

Artificial intelligence-driven triage in emergency departments: a comparative study of machine learning approaches

Objective In emergency departments, accurate and timely triage plays a critical role in prioritizing patient care. Traditional triage often suffers from subjectivity and inconsistency, especially under time pressure and overcrowded conditions. The aim of this study was to evaluate the predictive potential of multiple supervised machine learning models for three-class emergency department triage classification and compare their performance using routinely available triage variables. Methods We conducted a compara…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-1841· cite & verify
Artificial intelligence-driven triage in emergency departments: a comparative study of machine learning approaches
Construction site tour of the Skilled Nursing Residence, in Edgartown, on the island of Martha's Vineyard, in Dukes County, Maryland on August 5, 2024 - 50 by USDAgov. Public domain

In brief

Researchers compared eight supervised machine learning models for sorting emergency department patients into three urgency levels. They used 510 retrospective triage records containing demographics, vital signs and chief complaints, and tested each model with 10-fold cross-validation.

The findings suggest routine triage data may be enough for algorithms to support urgency decisions, which matters where triage is subjective and departments are crowded. However, the evidence comes from retrospective records rather than live use, so effects on patient flow or care quality remain untested.

Main points

  1. Eight supervised machine learning models were compared on a three-class triage task.
  2. The dataset held 510 retrospective emergency department triage records.
  3. Inputs were demographic features, vital signs and chief complaints.
  4. Logistic regression, random forest, support vector machine and majority voting performed best, above 89% accuracy.
  5. Low-urgency cases were classified most accurately, while errors mostly occurred between high and medium urgency.

The gain

Several supervised models, including logistic regression, random forest, support vector machine and majority voting, classified three-level triage urgency with accuracy above 89% using routine triage variables.

The rundown

The study was motivated by concerns that traditional triage can be subjective and inconsistent, particularly under time pressure and overcrowding. It framed accurate and timely triage as central to prioritizing patient care.

Performance was judged with accuracy, precision, recall and F1-score under 10-fold cross-validation. Confusion matrix analysis showed low-urgency cases were handled best across all models.

The main weakness was distinguishing high-urgency from medium-urgency patients, where most misclassifications occurred. The authors conclude that such models appear well-suited to support triage decisions and could improve prioritization and clinical operations if integrated into workflows.

Sources

  1. Peer-reviewedRevista da Associação Médica Brasileira2026-10-05

The debate