Background:
Heart failure (HF) has a high rate of mortality. It would be useful to have, at the time of a HF admission, a method of predicting post-discharge short-term mortality risk.
Methods:
We studied 947 Veterans hospitalized for HF at our VA medical center from January 2004 to December 2008, survived to discharge, had an ICD-9 code for HF as the primary discharge diagnosis, and had complete data for comorbidities, labs, medications and ECG findings. Mortality at 30-days post-discharge was determined and multivariable analyses identified independent predictors of mortality by logistic regression. These were used to develop a mortality risk score.
Results:
Mortality at 30-days occurred in 3.9% (37/947). Independent predictors (p-value <0.05) were substance abuse (OR = 3.10), abnormal admission serum sodium (OR=2.56), abnormal admission troponin (OR = 2.15), absence of dyslipidemia (OR = 2.04), not on a calcium channel blocker at admission (OR = 3.33), and not on an oral anticoagulant at admission (OR = 5.26). Each independent predictor was assigned 1-point and the resulting figure shows the rates of 30-day mortality to be 0.49%, 5.33% and 19.15% for those with risk scores of 0 to 2, 3 or 4, and ≥5, respectively. This represents a 39-fold difference in risk between the low-risk and high-risk patients. The c-statistic for the model was good (c = 0.796).
Conclusions:
Risk for post-discharge short-term mortality can be accurately predicted at the time of initial HF admission by a risk score employing admission labs, co-morbidities and medications. Patients identified as high risk at the time of admission might benefit from more intense inpatient evaluation and closer outpatient care.
P1488A simplified risk score inspired by cardshock score to predict short-term mortality in patients with cardiogenic shock