To explore the risk factors for death in patients with advanced dementia and to establish a prediction model for death risk in these patients. Methods: Total 144 dementia patients were retrospectively analyzed and divided into discharge group (81 patients) and death group (63 patients) based on their survival. Death risk analysis model for advanced dementia patients was established by univariate and Logistic regression analysis, and its degree of calibration and discriminating power were tested. Results: Patients in the death group were older and had a later onset and shorter course of dementia (all P<0.05). Patients in the discharge group had a higher education level (P<0.05) and a higher proportion with hypertension (P<0.05). Univariate analysis found that patients in the discharge group had lower incidence of pulmonary infection, fever, leukocyte abnormality, electrolyte abnormality, blood sugar abnormality, liver and kidney function abnormality, and edema than those in the death group (P=0.000); levels of hemoglobin and albumin were significantly higher than those in the death group (P=0.000), and the total protein level was lower than that in the death group (P=0.012). The ability of eating, dressing, going to the toilet, and walking in the discharge group was higher than that in the death group (P<0.05). Logistic regression analysis showed that pulmonary infection, hemoglobin value, previous hypertension, walking ability, age at onset, and abnormal white blood cells were risk factors for death in patients with advanced dementia. The predictive model had excellent goodness of fit. Conclusion: The risk factors in the death risk model for patients with advanced dementia include pulmonary infection, low hemoglobin value, history of hypertension, walking ability, age of onset, and leukocyte abnormality; among these, history of hypertension and hemoglobin value are protective factors.