Objective: To investigate the prognostic value of N-terminal B-type natriuretic peptide (NT-proBNP) on all-cause mortality in heart failure patients with preserved ejection fraction (HFpEF) at real world scenarios. Methods: Patients who met the diagnostic criteria of HFpEF in the China National Heart Failure Registration Study (CN-HF) were divided into death and survival groups. The demographic data, physical examination, results of the first echocardiography, laboratory results at admission, complications, drug use and clinical outcomes were obtained from CN-HF. The univariate Cox proportional hazard model was used to screen the variates that might predict prognosis, and then the covariates with statistical significance were included in the multivariate Cox regression model to analyze the predictive value of baseline NT-proBNP on all-cause death. Spearman correlation analysis was used to evaluate the relationship between NT-proBNP and estimated glomerular filtration rate (eGFR), so as to further explore the predictive value of the interaction between renal dysfunction and NT-proBNP on death. Since NT-proBNP did not obey the binary normal distribution, it was expressed by the natural logarithm of NT-proBNP (LnNT-proBNP). Results: A total of 1 846 HFpEF patients were enrolled in this study, with an average age of 71.5 years, 1 017 males(55.1%), median NT-proBNP 860 ng/L, and median eGFR 73.9 ml·min^-1·1.73m^-2. After a median follow-up of 34 months, 213 (11.5%) patients died. Patients in the death group were older, with higher NYHA classification Ⅲ-Ⅳ ratio, longer hospital stay, higher serum potassium and NT-proBNP level, prevalence of complications of diabetes mellitus, arrhythmia and atrial fibrillation, use of angiotensin receptor antagonist(ARB), mineralocorticoid receptor antagonists (MRA), diuretic and digoxin was significantly higher in death group than in survival group. Body mass index (BMI), diastolic blood pressure, left ventricular ejection fraction (LVEF), hemoglobin, serum cholesterol(TC), serum triglycerides (TG) and eGFR, and use of angiotensin converting enzyme inhibitors (ACEI), statins and aspirin were lower in death group than in survival group. Univariate Cox regression analysis showed that NT-proBNP was a predictor of all-cause death in HFpEF patients (HR=2.522, 95%CI 2.040-3.119, P<0.001). Multivariate Cox regression analysis showed that the elevated NT-proBNP remains as the independent predictor of all-cause death in patients with HFpEF (HR=1.230, 95%CI 1.049-1.442, P=0.011) after adjusting for age, BMI, diastolic blood pressure, LVEF, hemoglobin, serum potassium, serum sodium, TC, serum high-density lipoprotein cholesterol (HDL-C), TG, eGFR, atrial fibrillation, as well as the treatment of ACEI/ARB, MRA, diuretics and digoxin. Spearman correlation analysis showed that LnNT-proBNP was negatively correlated with eGFR (r=-0.361, P<0.001), but there was no interaction between NT-proBNP and renal dysfunction in predicting death in HFpEF patients (P>0.05). Conclusion The elevated level of NT-proBNP at admission is an independent predictor of all-cause mortality in HFpEF patients.