OBJECTIVES: To estimate the hospitalizations and its costs that can be attributed to diabetes mellitus (DM) among hospitalizations occurred in Brazilian Public Healthcare System (SUS) between 2008-2010. METHODS: Number of hospitalization and associated costs for study period was obtained through a review of government administrative claims database (DATASUS). Hospitalizations with a first-listed diagnosis of diabetes were added to hospitalizations estimated to be due to diabetes by attributable risk methodology (ARM).  ARM is based on the formula: RAPi = [P x (RRi - 1)] / [P x (RRi - 1) +1], where RAPi is the fraction of population attributable risk for medical condition "i" due to diabetes, P represents prevalence rate of diabetes, and RRi is the relative risk of medical condition "i" for people with diabetes compared to those without it (ADA, 2003).  Diabetes prevalence was obtained from VIGITEL-2006, a nationwide random telephonic sample (54,369 individuals). Self-reported and expanded estimates according to diagnosis rate were used. Relative risks of hospitalization for chronic complications and general medical conditions for diabetic subjects were obtained from literature. Results were given for entire population and per 10,000 inhabitants according to national census bureau. RESULTS: According to self-reported data, a total of 896,727 hospitalizations were estimated to be related to DM per year in SUS. This corresponds to 47 hospitalizations per 10,000 inhabitants annually in the whole population and can reach up to 318,2 per 10,000 when considering the group aged 75+. Annual hospitalization costs were estimated to be Brz$1,167,386,000 or Brz$61,197,90 per 10,000 inhabitants (and Brz$398,058.61/10,000 for 75+ population). When considering the expansion of self-reported cases according to diagnosis rate, it was estimate 1,353,161 hospitalizations (70,8 per 10,000) and a total annual cost of Brz$1,778,992 (Brz$ 93,140.65/10,000). CONCLUSIONS: DM and its complications are associated with a relevant economic burden to SUS, especially when considering the elderly population.