Background Severe acute malnutrition is defined by <70% weight for length/height, by visible severe wasting, by the presence of pitting edema, and in children 6 to 59 months of age, mid upper arm circumference <110 mm. Severe acute malnutrition remains to be a worldwide problem, claiming lives of millions of children, especially in sub-Saharan Africa and south Asia. Though the Ethiopian national guideline states the total length of stay in therapeutic feeding units should not be more than four weeks, there is huge difference, varying from 8 to 47 days of stay. Therefore, the objective of this study was to assess length of stay to recover from severe acute malnutrition and associated factors among under five children hospitalized to the public hospitals in Aksum Town. Methods Sample size was calculated using STATA version 12.0. A retrospective cohort study was conducted using pretested questionnaire in the public hospitals in Aksum on children aged 0-59 months. Cleaned data was entered to Epi info version 7.1.4 and then exported into SPSS version 21 for analysis. Bivariable and multivariable analyses were performed using Kaplan Meier and Cox regression models. During bivariable analysis, variables with p-value < 0.05 were selected for multivariable analysis to identify independent factors associated with length of stay. Results A total of 564 participants enrolled to the study. The rate of recovery was 56% with median length of stay of 15 days (95% CI: 14.1, 15.9). The independent predictors of length of stay to recovery were presence of diarrhea at admission (AHR = 0.573, 95% CI: 0.415-0.793), being HIV positive (AHR = 0.391, 95% CI: 0.194-0.788), palmar pallor (AHR = 0.575, 95% CI: 0.416-0.794), presence of other co-morbidities at admission (AHR = 0.415, 95% CI: 0.302-0.570) and not being treated with plumpy nut (AHR = 0.368, 95% CI: 0.262-0.518).