The aim of the present study was to investigate whether the pneumonia severity index (PSI) could adequately predict the severity of community-acquired pneumonia (CAP) and could be used as a severity of illness classification system. Furthermore, reasons that may influence the decision to admit low risk patients were analysed. In a prospective study 260 patients with CAP were included. Stratification in five risk classes according to the PSI was compared with parameters that are closely related to severity of CAR A significant difference in severity parameters, such as length of stay (P < 0.001) and simplified acute physiologic score and acute physiologic and chronic health evaluation II score (P < 0.001) was found between the five risk classes. Furthermore, a positive British Thoracic Society (BTS) rule and modified BTS rule score was significantly more prevalent in the higher risk classes (P < 0.001). The patient population had an average 30-day mortality of 10% and a mean Intensive Care Unit (ICU) admission rate of 8%. The mortality rate and ICU admission rate significantly differed between the five risk classes (P < 0.001), in which the highest ICU admission rate (40.9%) and the highest mortality percentage (40.9%) were both found in risk class V. Several clinical factors (n = 64), such as an exacerbation of chronic obstructive pulmonary disease in 17 patients and clinical appearance of being ill in 16 patients, lack of improvement on outpatient antibiotic therapy (n = 15) and social circumstances (n = 3) were reasons that influenced the decision to hospitalise low risk patients (n = 82). The results show that the PSI adequately predicted the severity of CAP and can be used as a severity of illness classification in CAP. Clinical and social factors other than those mentioned in the PSI have to be considered when making the decision to hospitalise patients with CAP.