To retrospectively analyze the early mortality and related risk factors in adult patients with maintenance hemodialysis (MHD). Adult MHD patients from 2008 to 2018 were enrolled and divided into training data group and validation data group. In training data group, multivariate logistic regression was used to analyze the risk factors of early death within 120 days after hemodialysis and establish a prediction model. The receiver operating characteristic (ROC) curve was applied to evaluate the prediction ability of the model. A total of 4 885 patients were included. The cumulative mortality within 120 days was 20.97/100 person years, and that within 365 days was 12.25/100 person years. A total of 3 603 patients in the training data group were analyzed. The following risk factors were correlated with early mortality (all <0.05), including age at start of dialysis over 60 years old (=1.792), non-chronic glomerulonephritis (=2.214), cardio-cerebrovascular disease (=2.695), plasma albumin less than 35 g/L (=1.358), platelet count less than 120×10/L (=2.194), serum creatinine less than 600 μmol/L (=1.652), blood urea nitrogen over 30 mmol/L (=1.887), blood phosphorus less than 1.13 mmol/L (=1.783), pulse pressure over 55 mmHg(1 mmHg=0.133 kPa) (=1.656), low density lipoprotein less than 1.5 mmol/L (=1.873), and blood calcium over 2.5 mmol/L (=1.876). Risk prediction model was established. The other 1 282 cases in the validation data group were verified. The area under ROC curve was 0.810, with sensitivity 85.7%, and specificity 62.5%. The mortality rate of adult MHD patients within 120 days after dialysis is high. The established prediction model can effectively predict the risk of early death.