EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Predictive Modeling in Triage Using Ordinal Logistic Regression: A Study Based on the MIMIC-IV-ED Dataset. "Ordinal Logistic Model for Emergency Triage"

Daniel Becerra Pabón, Danny Martínez Lobo, Pedro Araujo Zárate

Revista Ciencias Biomédicas · 15.10.2025 · 14 (4) · 216-227

DOI: 10.32997/rcb-2025-5945 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Revista Ciencias Biomédicas, online erschienen am 15.10.2025. Verfasst von Daniel Becerra Pabón und 2 weiteren Autor:innen. Die Arbeit nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Introduction

Accurate triage is essential in emergency departments operating under high demand and constrained resources.

Objective

To develop and evaluate an ordinal logistic regression model to predict five-level triage acuity using variables available at arrival, including symptom clusters derived from chief complaint text.

Methods

We analyzed the MIMIC-IV-ED database. After cleaning and variable selection, records with missing predictors were removed, yielding 380,052 complete ED visits. Predictors comprised demographics, arrival vital signs, and mode of arrival; chief complaint free text was represented by 300 k-means clusters. A proportional odds model was fitted and evaluated on a held-out dataset. Performance metrics were accuracy with 95% confidence interval, the No Information Rate (NIR) with p-value, Cohen’s kappa, and McFadden’s R².

Results

The model achieved 65.76% accuracy (95% CI, 65.48%–66.03%), exceeding the NIR of 56.29% (p < 2.2×10⁻¹⁶). Agreement beyond chance was fair (kappa = 0.3258) and fit was acceptable (McFadden’s R² = 0.1812). The most informative predictors were heart rate, oxygen saturation, mode of arrival, and symptom clusters capturing recurrent clinical presentations associated with higher or lower urgency.

Conclusion

Ordinal logistic regression proved to be a useful tool for triage prediction that respects the ordered nature of acuity categories. Incorporating unsupervised symptom clustering from free-text chief complaints enriches clinical pattern representation and may support clinicians during initial assessment and resource prioritization. Future work should address class imbalance in extreme acuity levels and assess generalizability across hospitals and clinical standards through prospective validation studies.

Zitieren

APA

Pabón, D. B., Lobo, D. M., & Zárate, P. A. (2025). Predictive Modeling in Triage Using Ordinal Logistic Regression: A Study Based on the MIMIC-IV-ED Dataset. "Ordinal Logistic Model for Emergency Triage". Revista Ciencias Biomédicas, 14(4), 216-227. https://doi.org/10.32997/rcb-2025-5945

Vancouver

Pabón DB, Lobo DM, Zárate PA. Predictive Modeling in Triage Using Ordinal Logistic Regression: A Study Based on the MIMIC-IV-ED Dataset. "Ordinal Logistic Model for Emergency Triage". Revista Ciencias Biomédicas. 2025;14(4):216-227. doi:10.32997/rcb-2025-5945

BibTeX
@article{Pabn2025,
  author = {Pabón, Daniel Becerra and Lobo, Danny Martínez and Zárate, Pedro Araujo},
  title = {Predictive Modeling in Triage Using Ordinal Logistic Regression: A Study Based on the MIMIC-IV-ED Dataset. "Ordinal Logistic Model for Emergency Triage"},
  journal = {Revista Ciencias Biomédicas},
  year = {2025},
  volume = {14},
  number = {4},
  pages = {216--227},
  doi = {10.32997/rcb-2025-5945}
}
Kostenlos starten →