Background When acute coronary syndrome (ACS) cannot be ruled out, emergency department (ED) patients with chest pain are admitted for in-hospital observation because of the risk of complications such as arrhythmia and acute heart failure. A study was undertaken to compare the ability of three risk prediction models to identify patients at a very low risk of complications.

Methods 559 consecutive patients with chest pain presenting to the ED and admitted for a suspicion of ACS were prospectively included. Predefined in-hospital complications were recorded and the risk predictions of the Global Registry of Acute Coronary Events (GRACE) risk score, the Freedom-from-Events (FFE) risk score and the Goldman rule were compared using receiver operating characteristics (ROC) curves.

Results Of the 559 patients, 140 had ACS and 32 had at least one complication. The GRACE score was superior to the FFE score in predicting the risk of complications (area under ROC curve 0.76 (95% CI 0.68 to 0.85) vs 0.69 (95% CI 0.60 to 0.79), p=0.021) whereas the Goldman rule (area under ROC curve 0.60; 95% CI 0.49 to 0.72) was inferior to both the GRACE and FFE scores. With the GRACE score set to a negative predictive value of 100% (95% CI 96% to 100%), 108 patients (19.3%) at almost no risk of complications could have been correctly identified in the ED.

Conclusion The GRACE and FFE scores are able to predict low complication risks in patients with chest pain admitted for suspected ACS, but only the GRACE score may be able to identify a significant number of patients at almost no risk of complications. A larger multicentre study is needed to confirm the possibility of using the GRACE score to identify patients suitable for assessment without monitoring.