The Impact of Prior Heart Failure Hospitalizations on Long-term Mortality Differs by Baseline Risk of Death

Abstract

Background Hospitalizations for decompensated heart failure (HF) are thought to increase long-term mortality. However, previous reports focus on newly hospitalized HF patients or clinical trial populations and do not always adjust for baseline mortality risk. We hypothesized that the number of HF hospitalizations within the prior 12 months would improve overall mortality risk stratification, particularly in otherwise “low-risk” HF inpatients. Methods We studied 2221 HF patients admitted to 14 Michigan community hospitals during 2002-2004. We estimated 1-year mortality using the multivariable (Enhanced Feedback For Effective Cardiac Treatment [EFFECT]) model and classified patients as low (EFFECT <90), moderate (90-120), and high risk (>120). We used logistic regression and stratified Cox proportional hazard modeling to explore the overall EFFECT model performance and the influence of HF hospitalizations within the prior 12 months on mortality risk. Results The EFFECT model adequately predicted and stratified for 1-year mortality (odds ratio 1.35 [95% confidence interval (CI), 1.30-1.40] per 10 points, P <.001, C-statistic 0.698), with low-, moderate-, and high-risk group mortality 18%, 35%, and 58%, respectively. The number of prior HF hospitalizations only modestly improved overall discrimination (C-statistic 0.704, P = .04). However, in low-risk patients the number of prior HF hospitalizations progressively increased the hazard for 1-year mortality (none: mortality 13%; 1: mortality 20%, hazard ratio [HR] 1.50 (95% CI, 0.86-2.60), P = .15; 2 or 3: mortality 27%, HR 2.24 (95% CI, 1.39-3.60); P = .001; 4 or more: mortality 31%, HR 2.80 (95% CI, 1.70-4.63); P <.001; P <.001 for trend). There was no consistent relationship between prior HF hospitalizations and 1-year mortality in moderate- or high-risk HF patients. Conclusion In otherwise “low-risk” HF inpatients, a history of 2 or more HF hospitalizations within the prior 12 months markedly increases 1-year mortality risk. This easily obtained information could help allocate specialized HF resources to the subset of “low-risk” patients most likely to benefit.