MEESSI-AHF risk score performance to predict multiple post-index event and post-discharge short-term outcomes

Background: The multiple estimation of risk based on the emergency department Spanish score in patients with acute heart failure (MEESSI-AHF) is a risk score designed to predict 30-day mortality in acute heart failure patients admitted to the emergency department. Using a derivation cohort, we evaluated the performance of the MEESSI-AHF risk score to predict 11 different short-term outcomes. Methods: Patients with acute heart failure from 41 Spanish emergency departments (n=7755) were recruited consecutively in two time periods (2014 and 2016). Logistic regression models based on the MEESSI-AHF risk score were used to obtain c-statistics for 11 outcomes: three with follow-up from emergency department admission (inhospital, 7-day and 30-day mortality) and eight with follow-up from discharge (7-day mortality, emergency department revisit and their combination; and 30-day mortality, hospital admission, emergency department revisit and their two combinations with mortality). Results: The MEESSI-AHF risk score strongly predicted mortality outcomes with follow-up starting at emergency department admission (c-statistic 0.83 for 30-day mortality; 0.82 for inhospital death,P=0.121; and 0.85 for 7-day mortality,P=0.001). Overall, mortality outcomes with follow-up starting at hospital discharge predicted slightly less well (c-statistic 0.80 for 7-day mortality,P=0.011; and 0.75 for 30-day mortality,P<0.001). In contrast, the MEESSI-AHF score predicted poorly outcomes involving emergency department revisit or hospital admission alone or combined with mortality (c-statistics 0.54 to 0.62). Conclusions: The MEESSI-AHF risk score strongly predicts mortality outcomes in acute heart failure patients admitted to the emergency department, but the model performs poorly for outcomes involving hospital admission or emergency department revisit. There is a need to optimise this risk score to predict non-fatal events more effectively.