As treatment strategies for patients with chronic heart failure (HF) become more sophisticated, identifying patients at high risk of death and predicting mode of death is important. The aim of this study was to explore the potential utility of heart rate turbulence (HRT) to identify patients with HF at high risk of death.In a prospective study, 553 ambulant outpatients age 63±10 with symptoms of HF and evidence of cardiac dysfunction were recruited. All patients underwent 24-h Holter ECG recordings, which were analysed for arrhythmias, heart rate variability and HRT a measurement that is thought to quantify cardiac autonomic regulatory mechanisms. Baseline chest radiograph, biochemistry and 12-lead electrocardiograms were also obtained. In patients with HRT measurements at 5 years follow up, 146 patients had died, 59 due to decompensated HF. Independent predictors of death from decompensated HF at 5-year follow up (Cox proportional hazard model) were HRT slope (HR for 10% increment 0.84, 95% CI 0.77–0.91), serum sodium (HR for 10% increment 0.75, 95% CI 0.62–0.91) and serum creatinine (for 10% increment HR 1.14, 95% CI 1.08–1.19) all P<0.01. These 3 variables combined had excellent discrimination between patients dying of decompensated HF and other patients, C-statistic=0.82.In patients with mild-to-moderate HF, HRT slope is an independent predictor of death due to decompensated HF. HRT may have the potential to help tailor therapy in this patient group.