A Multivariable Prediction Model for Mortality in Individuals Admitted for Heart Failure

OBJECTIVES: To derive and validate a 30-day mortality clinical prediction rule for heart failure based on admission data and prior healthcare usage. A secondary objective was to determine the discriminatory function for mortality at 1 and 2 years. DESIGN: Observational cohort. SETTING: Veterans Affairs inpatient medical centers (n = 124). PARTICIPANTS: The derivation (2010-12; n = 36,021) and validation (2013-15; n = 30,364) cohorts included randomly selected veterans admitted for HF exacerbation (mean age 7111; 98% male). MEASUREMENTS: The primary outcome was 30-day mortality. Secondary outcomes were 1- and 2-year mortality. Candidate variables were drawn from electronic medical records. Discriminatory function was measured as the area under the receiver operating characteristic curve. RESULTS: Thirteen risk factors were identified: age, ejection fraction, mean arterial pressure, pulse, brain natriuretic peptide, blood urea nitrogen, sodium, potassium, more than 7 inpatient days in the past year, metastatic disease, and prior palliative care. The model stratified participants into low- (1%), intermediate- (2%), high- (5%), and very high- (15%) mortality risk groups (C-statistic = 0.72, 95% confidence interval (CI) = 0.71-0.74). These findings were confirmed in the validation cohort (C-statistic = 0.70, 95% CI = 0.68-0.71). Subgroup analysis of age strata confirmed model discrimination. CONCLUSION: This simple prediction rule allows clinicians to risk-stratify individuals on admission for HF using characteristics captured in electronic medical record systems. The identification of high-risk groups allows individuals to be targeted for discussion of goals and treatment.