Objectives & Background Patient waiting times in Emergency Departments (ED) have risen in recent years, leading to increasing pressure on EDs to meet targets. As increased funding for EDs is an unlikely solution, better utilisation of current ED resources is required. An accurate method of predicting hospital admission at the point of triage could reduce wait times by facilitating patient streaming, bed planning and aiding staff decision making. Two tools that predict admission at triage have been created. They are the Glasgow admission prediction score (GAPS) and the ambulatory care score (AMBS). The primary aim of this study is to compare the the two tools' abilities to predict admission.

Methods Sampling was performed over ten days at the Sheffield Teaching Hospitals NHS Trust ED. Data was collected prospectively at triage for all consecutive patients who presented to the ED during sampling periods. GAPS and AMBS were calculated independent of patient clinical management and their disposition recorded. The accuracy of each tool was assessed using receiver operating characteristic curve (ROC) analysis.

Results Data was collected for 645 patients. GAPS had a higher area under ROC (AUC) of 0.815 (95%CI 0.781 to 0.849), compared to AMBS at 0.763 (95%CI 0.727 to 0.800). The difference in AUC was 0.0518 (95%CI 0.0135 to 0.0900) P=0.0080. GAPS had a sensitivity and specificity of 81.4% (95%CI 75.8% to 86.1%) and 67.5% (95%CI 62.7% to 72.0%) respectively, a PPV of 59.1% (95%CI 53.5% to 64.5%) and a NPV of 86.2% (95%CI 82.0% to 89.8%). Finally, GAPS correctly predicted the outcome of 72.6% of patients. AMBS had a sensitivity of 49.2% (95%CI 42.6% to 55.72%), a specificity of 81.4% (95%CI 77.3% to 85.1%), a PPV of 60.4% (95%CI 53.1% to 67.4%) and a NPV of 73.5% (95%CI 69.2% to 77.5%). Lastly, AMBS correctly predicted the outcome of 69.6% of patients.

Conclusion GAPS outperformed AMBS at predicting patient disposition. GAPS had a statistically significantly larger AUC than AMBS and as a result, better discriminative power. Additionally, GAPS had a more clinically favourable sensitivity, specificity, PPV, NPV and accuracy overall, when compared to AMBS. However, the results show that GAPS alone should not be used to determine patient outcomes, as it does not have sufficient accuracy, and may be best utilised alongside clinical judgment.