Background QRS duration is a well known predictor for survival in heart failure patients. However, the prognostic implication of QRS dispersion in severe heart failure patients is unclear. We investigated whether QRS dispersion during hospitalization has association with clinical outcome in severe heart failure patients. Method We consecutively enrolled 229 patients (60짹14 years, 64% men) with severe heart failure (left ventricular ejection fraction (LVEF) ���35%) who admitted for decompensated heart failure between January 2005 and December 2010. Twelve lead ECG during admission was used to calculate QRS duration. QRS dispersion was defined by difference between maximum QRS duration (QRSmax) and minimum QRS duration (QRSmin) during hospitalization. We analyzed the QRS dispersion with all caused mortality. Results Sixty-one patients (27%) died during 52짹19 months follow-up. There was no significant difference in clinical characteristics except for mean age (DG with 64짹13 vs. SG with 58짹13 years, p=0.003) between deceased group (DG, n=61) and survival group (SG, n=168). Baseline LVEF was also similar between two groups (27짹7 vs. 27짹7, p=0.695). However, DG had wider QRSmax (102짹21 vs. 111짹25 msec, p=0.008), QRSmin (119짹30 vs. 136짹34 msec, p=0.001) and QRS dispersion (17짹22 vs. 25짹31 msec, p=0.043) than SG. In univariate analysis, QRSmax, QRSmin and QRS dispersion were significant predictors of all-cause mortality (QRSmax, HR 1.001 (95% CI 1.005-1.017); QRSmin, HR 1.012 (95% CI 1.002-1.021); QRS dispersion, HR 1.010 (95% CI 1.003-1.018), all p<0.05). Dividing QRS dispersion by 10ms, all-cause mortality increased as QRS dispersion widened (HR, 1.34 (95% CI 1.124-1.608), p=0.001, figure). After adjustment with multiple variables, QRS dispersion remained significant (HR, 1.27, (95% CI, 1.03-1.55) p=0.02). Conclusion Not only QRS duration but also QRS dispersion during hospitalization showed independent predictor for all-cause mortality in severe heart failure patients.