OBJECTIVE To evaluate severe community-acquired pneumonia (SCAP) patients in an intensive care unit (ICU) with regard to risk factors for mortality and to compare ICU patients with matched non-ICU patients to evaluate whether our judgement for ICU admission was appropriate or not. METHODS During a 7-year period, all patients with CAP who were admitted to the ICU were examined. They underwent clinical and radiographic evaluations, and two commonly used severity of illness scores were also calculated using the Simplified Acute Physiological Score (SAPS) and the Acute Physiology and Chronic Health Evaluation (APACHE) II methods. To detect risk factors for ICU admission using existing guidelines, each study patient was matched with two patients hospitalized in a general medical ward. RESULTS Seventy-two patients were identified during the study period. Their mean age was 72.9 years, and 35 patients (48.6%) subsequently died. For the univariate analysis, there were significant differences with the pulse rate > or = 130/min, blood urea nitrogen > or = 30 mg/dl, multilobar shadow, SAPS > or = 13, APACHE II > or = 23, and the occurrence of septic shock between the survivors and those who died. For the multivariate analysis, septic shock (p = 0.0005, odds ratio of 26.6) and blood urea nitrogen > or = 30 mg/dl (p = 0.037, odds ratio of 5.38) were associated with mortality. Regarding the characteristics of different clinical predictions for ICU admission, the revised American Thoracic Society criteria might have been the most accurate. CONCLUSIONS Septic shock was associated with high mortality, which is a more accurate and higher predictor of mortality than was physical examination, laboratory or radiographic findings.