The CURB65 pneumonia severity score outperforms generic sepsis and early warning scores in predicting mortality in community-acquired pneumonia

Background: The performance of CURB65 in predicting mortality in community-acquired pneumonia (CAP) has been tested in two large observational studies. However, it has not been tested against generic sepsis and early warning scores, which are increasingly being advocated for identification of high-risk patients in acute medical wards. Method: A retrospective analysis was performed of data prospectively collected for a CAP quality improvement study. The ability to stratify mortality and performance characteristics (sensitivity, specificity, positive predictive value, negative predictive value and area under the receiver operating curve) were calculated for stratifications of CURB65, CRB65, the systemic inflammatory response syndrome (SIRS) criteria and the standardised early warning score (SEWS). Results: 419 patients were included in the main analysis with a median age of 74 years (men = 47%). CURB65 and CRB65 stratified mortality in a more clinically useful way and had more favourable operating characteristics than SIRS or SEWS; for example, mortality in low-risk patients was 2% when defined by CURB65, but 9% when defined by SEWS and 11 - 17% when defined by variations of the SIRS criteria. The sensitivity, specificity, positive predictive value and negative predictive value of CURB65 was 71%, 69%, 35% and 91%, respectively, compared with 62%, 73%, 35% and 89% for the best performing version of SIRS and 52%, 67%, 27% and 86% for SEWS. CURB65 had the greatest area under the receiver operating curve (0.78 v 0.73 for CRB65, 0.68 for SIRS and 0.64 for SEWS). Conclusions: CURB65 should not be supplanted by SIRS or SEWS for initial prognostic assessment in CAP. Further research to identify better generic prognostic tools is required.