Can Manchester Triage Predict Hospitalisation?

ABSTRACT INTRODUCTION Emergency Departments (EDs) face increasing patient volumes with limited resources. The Manchester Triage System (MTS) is designed to prioritisecare based on acuity (1). This study aimed to assess whether MTS can independently predict hospitalisation at triage. METHODS A retrospective observational study was conducted at the University Clinical Center Maribor ED (Jan 2021–Apr 2024). All adult patients triaged using MTS were included. The primary outcome was hospitalisation. Statistical methods included Chi-square testing, relative risk (RR) estimation, logistic regression, and Precision-Recall analysis due to class imbalance. RESULTS Among 324,170 patients, 27,446 (8%) were hospitalised. Hospitalisation rates increased with urgency: red (53%), orange (22%), yellow (12%), green (5%), blue (0.4%) (p<0.001). Compared to green, RR for hospitalisation was 18.6 (red), 11.2 (orange), 5.3 (yellow). Logistic regression yielded a 95% OR [0.2472, 0.2560], with p<0,000. Due to class imbalance, the Precision-Recall AUC was 0.29 with an optimal threshold of 0.04 (F1 score: 0.30), whilst AUC-ROC was 0.76. CONCLUSIONS While MTS is effective for triage, it has limited standalone predictive value for hospitalisation (PR-AUC: 0.29). Performance is better than random but insufficient for clinical prediction. Compared to Brutschin et al. (AUC-ROC: 0.639) on MTS “unwell adult” flowchart (2), our full-cohort model showed slightly higher AUC. Future predictive models may improve by incorporating additional clinical scores (e.g., NEWS2, Sepsis-6).