Objective: Establish the reliability of the scored Patient Generated-Subjective Global Assessment (PG-SGA) in determining nutritional status among Antiretroviral Therapy (ART) naive HIV-infected adults. Methods: A descriptive, cross sectional study among outpatient medical clinics, in The AIDS Support Organization (TASO), Mulago Centre, Kampala, Uganda. The study group (n=217) consisted of male (n=60) and female (n=157) HIV patients (18-67 years). Purposive sampling was used. Anthropometry (weight, height, BMI), nutritional history (body weight, dietary intake, gastrointestinal symptoms, functional capacity and infections), and clinical status were assessed. Sensitivity and specificity of PG-SGA were determined using Receiver Operating Characteristic (ROC) curve. Data collection was done from April-May 2008. Results: Only 12% of the subjects were underweight and over half (58.2%) had normal weight. The PG-SGA had low sensitivity (69.2%) and specificity (57.1%) at categorizing the risk for malnutrition indicated by BMI< 18.5. Conclusion: There was a high prevalence of malnutrition among the study group. Overall, this study indicated the PG-SGA could not adequately discriminate between underweight and normal patients. The tool was not reliable enough for determining nutritional status in this population.