Objective: To assess the Simplified Acute Physiology Score 3 (SAPS3) prognostic score performance across different body mass index categories. Methods: A retrospective cohort study in a general ICU in Brazil. A secondary analysis of medical records was performed with clinical and epidemiological data. Patients were stratified according to their body mass index (BMI) category, and a binary logistic regression was then performed to identify factors independently associated with mortality. SAPS3 accuracy was determined using the area under the receiver operating characteristics curve and the Hosmer-Lemeshow test. A modified Kaplan-Meyer plot was employed to evaluate death probability according to BMI. ICU mortality was evaluated as the primary outcome. Results: A total of 2,179 patients (mean age of 67.9 years and female predominance (53.1%)) were enrolled. SAPS3 was found accurate in all groups except in the underweight (AUC: 0.694 95% CI 0.616–0.773; HL = 0.042). The patients in the underweight group tended to be older, have longer hospital stay, have worse functional status, and have a higher value on prognostic scores. After the adjustments, no statistically significant difference between the BMI groups was noted in relation to mortality, except for the low weight that presented a likelihood of death of 3.50 (95% CI, 1.43–8.58, p = 0.006). Conclusion: This research showed that SAPS3 had poor accuracy in predicting ICU mortality in underweight patients. This group was shown to be an independent risk factor for worse clinical outcomes.