Prevalence of undernutrition at admission and over the weeks and evaluate the association between nutritional status and length at a Hospital in the south of Brazil. A cohort study was carried out from March to October in 2004. Patients between 1 month and 12 years were included. Anthropometric data were collected up to 48 hours after admission and, weekly, up to hospital discharge. Children below 5 years of age, standard WHO/2006 for the classification of the z-score for the stature/age (S/A),weight/age (W/A) and weight/stature (W/S) scores was used. Children from 5 to 10 years of age, standards NCHS, 1977 were used to classify the same rates as reference values. Children above 10 years of age, classification of BMI was used. In order to compare the z-scores, ANOVA was used for repeated measurements, with Bonferroni’s Post-Hoc test, and, for the evaluation of the in-hospital stay length, according to the nutritional status, survival curve was used, in SPSS program, v. 12.0, was used. 426 patients were included. 57% were male and 50.7% were below 1 year of age. At admission, the prevalence of malnutrition was of 10%, 18%, 21% and 14.7%, according to the W/S, W/A, S/A, and BMI criteria,r espectively. Improvement of the nutritional status was observed in children below 5 years (p = 0.001) and S/A (p = 0.007) and also in 5 to 10 years of age (S/A, p = 0.024). Undernourished patients, compared to nourished (according to the W/S rate for children), showed a higher probability of remaining hospitalized (HR = 1.41; IC 95%: 1.02 to 1.92). The prevalence of malnutrition by the W/S score was about half the prevalence found by the W/A and W/A indices, reflecting a chronic impairment of the nutritional status.