Intensive Care Unit (ICU) is an important department within a hospital since it deals mostly with complex cases and it generates the highest amount of costs, thus requiring adequate control on its care treatments. Nonconformities such as poor communication and treatment errors are commonly responsible for a bad performance in ICUs. However, evaluating the performance of an ICU is not an easy task and there are no gold-standard indicators. The most common metrics are the Standardized Mortality Ratio (SMR) and the Standardized Resource Use (SRU), which measure mortality and resource utilization, respectively. Hence, this study aims to analyze different ICUs in terms of mortality, resource use, and institutional factors, combining the methods Efficiency Chart, Rankability and Risk Profile. The analysis was performed considering a total of 12,100 patients in 116 ICUs provided by a clinical trial study. As results, it was verified that most ICUs were from hospitals with public administration (47.41 per cent), which had significantly high lethality rate compared to private hospitals. Four different clustering approaches were tested, which identified similar case-mixes between the best and lower performance groups of ICUs, and a high variability in expected risks for low severity patients. Using a resampling approach, it was evidenced that the mortality indicator varies strongly on low-risk groups of patients, while high-risk patients had a smaller range of SMR values, which may lead to biased conclusions when comparing ICUs with similar mortality and different case-mixes.