Severity assessment tools in patients with community-acquired pneumonia can serve multiple purposes. These include the identification of low risk patients who may be suitable for management at home, higher risk patients who should be managed in hospital and patients with more severe illness who may require admission to the intensive care unit (ICU). The best known are the PSI and CURB-65, which are based on patients risk of dying within 30 days. While such tools are reasonably accurate at stratifying mortality risks, they are much less accurate in predicting which patients will be admitted to the ICU. This reflects the fact that there is often little overlap between the patients who die and those admitted to ICU. This begs the question of how useful it is to identify the higher mortality patients if they usually die because active medical treatment is withdrawn. Thus there is increasing interest in tools that predict which patients are more likely to need management in ICU. Examples include SMART-COP, REA-ICU, SCAP and the minor criteria of the IDSA/ATS rule. While all are at least moderately accurate for this purpose, limitations of previous studies, such as not recording ‘do not resuscitate’ status makes valid comparisons difficult.