Predicting admission at triage: are nurses better than a simple objective score?

Area of Science:

• Emergency Medicine
• Clinical Decision Support

Background:

• Predicting hospital admission from Emergency Department (ED) triage is crucial for patient flow and resource allocation.
• Current methods rely on subjective assessments by triage nurses, which can be variable.

Purpose of the Study:

• To compare the predictive accuracy of the Glasgow Admission Prediction Score (GAPS) against triage nurses' estimations for hospital admission.
• To evaluate the performance of GAPS and nurse estimations in different clinical certainty scenarios.

Main Methods:

• A single-centre prospective study involving 1829 ED attendances.
• Triage nurses estimated admission probability using a visual analogue scale (VAS).
• GAPS scores were automatically generated from triage data; calibration, discrimination (AUC), and accuracy were compared.

Main Results:

• GAPS predicted admissions (41.0%) comparably to actual admissions (40.7%), while nurses overestimated (47.3%).
• GAPS and nurses showed similar discrimination (AUC 0.876 vs 0.875) and binary prediction accuracy (80.6% vs 79.0%).
• GAPS significantly outperformed nurses in calibration, discrimination, and accuracy when nurses were not highly certain (≥95%); a combined approach yielded the best results (AUC 0.891, accuracy 82.5%).

Conclusions:

• The Glasgow Admission Prediction Score (GAPS) is a more reliable predictor of hospital admission than general nurse estimations.
• Triage nurses' clinical judgment remains valuable when they express high certainty (≥95%) about the admission outcome.
• A hybrid approach, utilizing GAPS and respecting high-certainty nurse judgments, optimizes prediction accuracy.