Predicting complications in emergency department patients with acute coronary syndrome – Existing risk scores versus a new logistic regression model

Background: Patients with acute coronary syndrome (ACS) are often admitted to monitored wards due to the risk of complications. Several risk prediction scores exist, but their use in the emergency department (ED) is limited. We aimed to compare the ability of existing risk scores with a new logistic regression model in predicting complications in ACS patients. Methods: This was a secondary analysis of data from the ESC TROP trial (NCT03421873), including ACS patients from five EDs in Region Skåne, Sweden (2017-2018). Complications were identified via diagnosis and/or intervention codes and manual chart review. GRACE, GRACE FFE, TIMI, HEART, ACTION ICU, and CHA₂DS₂-VASc scores were calculated. A new logistic regression model was developed, and its predictive performance was assessed using the area under the ROC curve (AUROC) and a net reclassification improvement analysis (NRI). Results: Among 2223 ACS patients, 164 (7.4%) experienced complications. Independent predictors for complications included age, STEMI, troponin and lactate at arrival, shock index, Killip class, and new ECG changes. The logistic regression model's AUROC 0.84 (95% CI 0.80-0.88) outperformed all known risk scores: GRACE FFE 0.79 (0.75-0.84), ACTION ICU 0.77 (0.72-0.82), GRACE 0.76 (0.70-0.81), TIMI 0.74 (0.68-0.79), HEART 0.69 (0.64-0.74), and CHA₂DS₂-VASc 0.64 (0.59-0.69). Logistic regression improved reclassification of non-events, with a positive non-event NRI compared with all other scores. Conclusions: Serious complications occurred in 7% of ACS patients. A logistic regression model based on simple ED variables showed excellent predictive performance, surpassing existing risk scores. Improved risk stratification may optimize resource allocation while maintaining patient safety.