Despite the existence of standard protocol, many stabilization centers (SCs) continue to experience high mortality of children receiving treatment for severe acute malnutrition. Assessing treatment outcomes and identifying predictors may help to overcome this problem. Therefore, a 30-month retrospective cohort study was conducted among 545 randomly selected medical records of children <5 years of age admitted to SCs in Gedeo Zone. Data was entered by Epi Info version 7 and analyzed by STATA version 11. Cox proportional hazards model was built by forward stepwise procedure and compared by the likelihood ratio test and Harrell’s concordance, and fitness was checked by Cox–Snell residual plot. During follow-up, 51 (9.3%) children had died, and 414 (76%) and 26 (4.8%) children had recovered and defaulted (missed follow-up for 2 consecutive days), respectively. The survival rates at the end of the first, second and third weeks were 95.3%, 90% and 85%, respectively, and the overall mean survival time was 79.6 days. Age <24 months (adjusted hazard ratio [AHR] =2.841, 95% confidence interval [CI] =1.101–7.329), altered pulse rate (AHR =3.926, 95% CI =1.579–9.763), altered temperature (AHR =7.173, 95% CI =3.05–16.867), shock (AHR =3.805, 95% CI =1.829–7.919), anemia (AHR =2.618, 95% CI =1.148–5.97), nasogastric tube feeding (AHR =3.181, 95% CI =1.18–8.575), hypoglycemia (AHR =2.74, 95% CI =1.279–5.87) and treatment at hospital stabilization center (AHR =4.772, 95% CI =1.638–13.9) were independent predictors of mortality. The treatment outcomes and incidence of death were in the acceptable ranges of national and international standards. Intervention to further reduce deaths has to focus on young children with comorbidities and altered general conditions.