The aim of this study was to determine the value of serial B-type natriuretic peptide (BNP) assay for predicting post-discharge outcome of patients admitted for decompensated congestive heart failure (CHF). Patients hospitalized for decompensated CHF are frequently re-admitted. Thus, identification of high-risk patients before their discharge is a major issue that remains challenging. B-type natriuretic peptide measurement could be useful. Serial BNP measurements were performed from admission to discharge in two samples of consecutive patients. Survivors were monitored for six months; the main end point combined death or first re-admission for CHF. Among the 105 survivors of the derivation study, all serial BNP values, percentage change in BNP levels, and predischarge Doppler mitral pattern correlated with the outcome. In contrast, clinical variables and left ventricular ejection fraction were poorly predictive. The predischarge BNP assay had the best discriminative power (area under the receiver operating characteristic [ROC] curve = 0.80) and remained the lone significant variable in multivariate analysis (hazard ratio [HR] = 1.14 [95% confidence interval [CI], 1.02 to 1.28], p = 0.027). Among the 97 survivors of the validation study, the predischarge BNP assay was also the most predictive parameter (area under the ROC curve = 0.83). The risk of death or re-admission increased in stepwise fashion across increasing predischarge BNP ranges (p < 0.0001). After adjustment for baseline covariables, the HRs were 5.1 [95% CI 2.8 to 9.1] for BNP levels between 350 and 700 ng/l and 15.2 [95% CI 8.5 to 27] for BNP levels >700 ng/l, compared with BNP <350 ng/l. High predischarge BNP assay is a strong, independent marker of death or re-admission after decompensated CHF, more relevant than common clinical or echocardiographic parameters and more relevant than changes in BNP levels during acute cares.