Objectives: The appetite test is used to risk stratify for children with severe acute malnutrition (SAM) in inpatient or outpatient care. The test is recommended in guidelines despite lack of evidence. We evaluated its ability to identify children at risk of a poor treatment outcome. Methods: We conducted an observational study of children diagnosed with SAM at three health facilities in Ethiopia. The appetite test was done independently, and the result did not affect decisions about hospitalisation and clinical care. Data were analysed using mixed linear and logistic regression models. Results: Appetite was tested in 298 (89%) of 334 children enrolled; 56 (19%) passed. Children failing the appetite test had a 6.6% higher weight gain per day (95% CI: 2.6, 10.8) adjusted for type of treatment, oedema, duration of follow-up and age than children passing the test. We found medical complications in 179 (54%) children. Medical complications were associated with blood markers of metabolic disturbance. Children with medical complications tended to have lower weight gain than those without complications (3.5%, 95% CI: −0.25, 7.0). Neither the appetite test nor medical complications were correlated with bacteraemia or treatment failure. Conclusions: Our findings question the use of the appetite test to identify children who need inpatient care. An assessment of medical complications alone could be a useful risk indicator but needs to be evaluated in other settings.