Heart failure prognosis: comorbidities matter.

BACKGROUND Prior risk prediction models have included a selective group of broad comorbidities in scoring prognosis of heart failure (HF) patients. OBJECTIVE We examined whether scoring a comprehensive set of comorbidities separately, could improve the performance and accuracy of predicting HF prognosis. METHODS This is a cross-validated, longitudinal, retrospective, observational study. Data were collected on 602,050 unique HF patients, who received care through the Veterans Administration (VA) between October 1, 2006 and September 31, 2011. The dependent variable was mortality in six months. The independent variables were all International Classification of Disease (ICD) comorbidities, without grouping into broad disease categories. RESULTS The area under the receiver-operating curve (AROC) for the multimorbidity (MM) index was 0.784 (95% confidence interval [CI]: 0.781-0.786). The MM index was significantly (alpha <0.05) more accurate than the Quan variant of the Charlson Index (AROC=0.656), the comorbidity categories within the Care Assessment of Need (CAN) Index (AROC=0.677), the von Walraven variant of the Elixhauser Index (AROC=0.639), chronological age (AROC=0.649), or ejection fraction (EF) (AROC=0.533). CONCLUSION Inclusion of additional comorbidities improves the accuracy of HF prognostic indices. Future studies are needed to drive HF prognostic indices with the MM index as a component.