Background: Risk prediction models may potentially improve disposition decision-making for patients with heart failure (HF). However, few acute HF models have been prospectively validated or compared to physician-estimated risk (PER). We examined the performance of the Emergency Heart failure Mortality Risk Grade (EHMRG, https://ehmrg.ices.on.ca) for mortality prediction at 7 and 30 days.
Methods: We prospectively recruited HF patients from 9 emergency departments (EDs) in Ontario, Canada (Jul 2010-Mar 2015). The physician responsible for patient disposition in the ED recorded a PER of death within 7 days; with an allowable response range from 1 to 100%, based solely on clinical judgment. We linked prospectively-collected EHMRG score variables to the Registered Persons Database for vital status. The predictive ability of the EHMRG vs. PER was evaluated using the c-statistic and net reclassification improvement (NRI).
Results: A total 1983 HF patients (median 81 [IQR 71-87] yrs, 52% men) had data collected for calculation of 7-day score, and 1691 had 12-lead ECG interpretations allowing for calculation of the 30-day score. 1566 (79%) were admitted to the hospital. There were 39 deaths at 7 days and 138 deaths at 30 days. Mortality rates categorized according to previously established EHMRG score quintiles demonstrated no deaths in the lowest risk groups (Figure). At 7 days, c-statistics were 0.71 (95%CI; 0.64-0.78) for PER, 0.81 (95%CI; 0.75-0.87) for the EHMRG (p=0.022), and 0.82 (95%CI; 0.76-0.88) for PER combined with EHMRG (p=0.003). At 30 days, c-statistics were 0.63 (95%CI; 0.58-0.67) for PER, 0.77 for EHMRG (95%CI; 0.73-0.81), and 0.77 for PER combined with EHMRG (all p<0.001 vs. PER). Continuous NRI of the 7-day EHMRG was 76% compared to PER (p<0.001).
Conclusion: Standardized mortality risk modeling using 7-day and 30-day EHMRG scores was superior to PER based on clinical judgment alone. (ClinicalTrials.gov NCT 02634762)