The aim of this study was to evaluate the accuracy of three score systems: the pneumonia severity index (PSI); CURB-65 (confusion; urea .7 mM; respiratory rate o30 breaths?min -1 ; blood pressure ,90 mmHg systolic or f60 mmHg diastolic; aged o65 yrs old); and modified American Thoracic Society rule for predicting intensive care unit (ICU) need and mortality due to bacteraemic pneumococcal pneumonia. All adult patients (n5114) with invasive pneumococcal pneumonia at the Karolinska University Hospital, Sweden, 1999-2000, were included in the study. Severity scores were calculated and the independent prognostic importance of different variables was analysed by multiple regression analyses.PSI oIV, CURB-65 o2, and the presence of one major or more than one minor risk factor in mATS all had a high sensitivity, but somewhat lower specificity for predicting death and ICU need. The death rate was 12% (13 out of 114). Severity score and treatment in departments other than the Dept of Infectious Diseases were the only factors independently correlated to death. Patients treated in other departments more often had severe underlying illnesses and were more severely ill on admission. However, a significant difference in death rates remained after adjustment for severity between the two groups.In conclusion, all score systems were useful for predicting the need for intensive care unit treatment and death due to bacteremic pneumococcal pneumonia. The pneumonia severity index was the most sensitive, but CURB-65 was easier to use.