A user-friendly risk-score for predicting in-hospital cardiac arrest among patients admitted with suspected non ST-elevation acute coronary syndrome – the SAFER-score

Aim: To develop a simple risk-score model for predicting in-hospital cardiac arrest (CA) among patients hospitalized with suspected non-ST elevation acute coronary syndrome (NSTE-ACS). Methods: Using the Swedish Web-system for Enhancement and Development of Evidence-based care in Heart disease Evaluated According to Recommended Therapies (SWEDEHEART), we identified patients (n = 242 303) admitted with suspected NSTE-ACS between 2008 and 2014. Logistic regression was used to assess the association between 26 candidate variables and in-hospital CA. A risk-score model was developed and validated using a temporal cohort (n = 126 073) comprising patients from SWEDEHEART between 2005 and 2007 and an external cohort (n = 276 109) comprising patients from the Myocardial Ischaemia National Audit Project (MINAP) between 2008 and 2013. Results: The incidence of in-hospital CA for NSTE-ACS and non-ACS was lower in the SWEDEHEART-derivation cohort than in MINAP (1.3% and 0.5% vs. 2.3% and 2.3%). A seven point, five variable risk score (age ≥60 years (1 point), ST-T abnormalities (2 points), Killip Class >1 (1 point), heart rate <50 or ≥100 bpm (1 point), and systolic blood pressure <100 mmHg (2 points) was developed. Model discrimination was good in the derivation cohort (c-statistic 0.72) and temporal validation cohort (c-statistic 0.74), and calibration was reasonable with a tendency towards overestimation of risk with a higher sum of score points. External validation showed moderate discrimination (c-statistic 0.65) and calibration showed a general underestimation of predicted risk. Conclusions: A simple points score containing five variables readily available on admission predicts in-hospital CA for patients with suspected NSTE-ACS.