Integrating the interpretable machine learning Score For Emergency Risk Prediction (SERP) with emergency department triage to better predict 30-Day mortality

Comprehensive Summary

Wong Qi Feng et al. investigated whether integrating a machine learning-based Score for Emergency Risk Prediction (SERP) into the Patient Acuity Category Scale (PACS) could improve ED triage accuracy. Using data from over 190,000 ED visits, two enhanced models were developed: one that adjusted triage in both directions: upgrading and downgrading patients based on 30-day mortality risk, and one that exclusively upgraded high-risk patients. Both models outperformed standard PACS in predicting 30-day mortality, with higher AUC values reflecting superior discriminative accuracy. The bidirectional model demonstrated the greatest overall clinical benefit. These findings suggest that augmenting traditional triage systems with ML has the potential to improve risk stratification and support more accurate, data-driven clinical decision-making in emergency care.

Outcomes and Implications

Integrating ML tools such as SERP into ED triage has the potential to improve patient outcomes by more accurately identifying high-risk individuals and prioritizing their care accordingly. Such models may also alleviate overcrowding by safely downgrading lower-risk patients, thereby optimizing resource allocation. Furthermore, ML-assisted triage supports more consistent, empirically driven clinical decision-making. Nevertheless, careful implementation is essential to ensure these models perform reliably across diverse patient populations and remain a complement to, rather than a replacement for, physician judgment.