Background: Children in third world countries suffer from severe acute malnutrition in an extent of public health important. Prevention, early identification and management of SAM are important to decrease associated mortality. Objective: The main aim was to assess survival status and to identify predictors of mortality among under- five SAM children admitted to TGSH. Method: A retrospective cohort study was conducted. Systematic random sampling method was used; a check list was developed from registration log book. Data analysis was done using SPSS version 20. Frequency, median survival curve, life table was used to present the data. To identify the associated factors, Cox-proportional hazard model was employed. Hazard ratio and 95% confidence intervals were applied to show the strength of association between the outcome variable and its predictors. Variables with p value < 0.05 are declared as statistical significant Results: A total of 396 children were enrolled into the study. A censored and death rate was 91.4% and 8.6 % respectively. The median follow up period was 10 days. The overall mean survival time was 25 (95% CI 23.50–27.10) days. The main predictors of hospital deaths were Age <24 months (AHR= 11.44, 95% CI (3.38-38.67), patient in shock (AHR=5.40, 95 % CI (1.47-19.86), impaired level of consciousness (AHR= 5.67, 95% CI (2.03-15.87) and patient having vomiting during admission (AHR=3.57, 95% CI (1.41-9.13). Conclusion and recommendation: The treatment outcomes (improvement rate, death rate and average length of stay in the hospital) were in agreement with minimum international standard set for management of severe acute malnutrition. Intervention to further reduce earlier deaths should focus on young children with shock, altered mental status, and vomiting. Key Words: Complicated SAM, Co morbid, Death Rate, Length of hospital status