Abstract: Objective：To explore the effects of glucose variability on mortality risks of severe acute stroke patients 3 months after admission. Methods：Totally, 162 severe acute stroke patients (from January 2016 to January 2018) in ICU were retrospectively analyzed. According to survival conditions 3 months after admission, patients were assigned to survived group and dead group. The initial glucose levels, average glucose levels, glucose standard deviations, and variable coefficients were recorded to measure the glucose variability. National Institutes of Health Stroke scale (NIHSS) scores at baseline and Acute Physiology and Chronic Health Evaluation Ⅱ (APACHEⅡ) scores between groups were compared. The correlation of glucose variability with patient’s prognosis was analyzed. Results：Three months after admission, the survey results showed 132 survived cases and 30 dead cases; it was significantly different in the APACHE Ⅱ and NIHSS scores between the groups (P＜0.05); and the initial glucose levels, average glucose levels, glucose standard deviations, and variable coefficients were statistically different between the two groups (P＜0.05). According to variable coefficients of average glucose levels, patients were assigned to <30% group, 30% to 50% group, and >50% group. The proportion of patients with glucose variability over 50% in dead group (60%) was found to be significantly higher than that of patients with glucose variability ＜30% and 30% to 50% (16.67%, 23.33%). The proportion of patients with glucose variability ＜30% in survived group (49.24%) was found to be significantly higher than that of patients with glucose variability 30% to 50% and ＞50% (35.60%, 15.15%); the difference was statistically significant (P＜0.05). Logistic regression analysis indicated that the independent predictive factors of mortality risks 3 months after admission included baseline APACHE Ⅱ scores, NIHSS scores, initial glucose levels, average glucose levels, glucose standard deviations, and variable coefficients. Conclusions：The glucose variability is proved to have high predictive value on mortality risks 3 months after admission. Therefore, it can be served as an important predictive indicator for patient’s prognosis.