A number of different models have been used to assess prognosis in pneumonia. In this study the Pneumonia Severity Index was compared with the CURB and CURB-65 severity scores in predicting 30 day mortality.

3181 patients with community acquired pneumonia were assessed between January and December 2001 in 32 emergency departments in Connecticut and Pennsylvania. Patients were assessed according to the criteria of the Pneumonia Severity Index, CURB, and CURB-65 and then stratified into risk groups I–V, 0–4, and 0–5, respectively. Statistical analysis was performed to assess 30 day mortality and the accuracy of the risk classification based on sensitivity, specificity, likelihood ratios, and predictive values. The results indicated that the Pneumonia Severity Index stratified more patients as low risk (68%) than CURB (51%) or CURB-65 (61%), both p<0.001. Low risk patients identified by the Pneumonia Severity Index had slightly lower 30 day mortality (1.4%) than those stratified as low risk in CURB (1.7%) or CURB-65 (1.7%). It was found that the Pneumonia Severity Index had a higher sensitivity and lower specificity than the CURB scores. All three models had high negative predictive values (>95%).

The absolute difference in 30 day mortality based on the three models is small and of uncertain clinical significance. However, the more complex Pneumonia Severity Index does classify many more patients as low risk and this may have beneficial cost implications in terms of outpatient treatment.