External validation of the CURSI criteria (confusion, urea, respiratory rate and shock index) in adults hospitalised for community-acquired pneumonia

Abstract

Abstract Background For patients hospitalised due to community-acquired pneumonia (CAP), mortality risk is usually estimated with prognostic scores such as CRB-65 or CURB-65. For elderly patients, a new score referred to as CURSI has been proposed which uses shock index (SI) instead of the blood pressure (B) and age (65) criteria. The new score has not been externally validated to date. Methods We used data from a hospital-based CAP registry to compare the ability of CURSI, CURB-65 and CRB-65 to predict mortality at day 30 after hospital admission. Patients were stratified by score points as well as score-point-based risk categories, and mortality for each group was assessed. To compare test performance, receiver-operating characteristic (ROC) curves were constructed, and the areas under the curve (AUROC) were calculated with 95% confidence intervals (CI). Results We analysed 553 inpatients (45% females, median age 78 years) hospitalised between 2005 and 2009 for CAP. Overall, mortality at day 30 was 11% (59/553). The study sample was characterised by advanced comorbidity (chronic heart failure: 22%, chronic kidney failure: 27%) and functional impairment (nursing home residency: 26%, dementia: 31%). All risk scores were significantly associated with 30-day mortality. The AUROC values with 95% CI using score points for risk prediction were as follows: 0.63 [0.56-0.71] for CRB-65, 0.68 [0.61-0.75] for CURB-65 and 0.68 [0.61-0.75] for CURSI. The CURSI-defined low-risk group (0 or 1 score point) had a higher mortality (8%) than the low-risk groups defined by CURB-65 and CRB-65 (4% and 3%, respectively). Lowering the cut-off for the CURSI-defined low-risk group (0 point only) would lower the mortality to 4%, making it comparable to the CURB-65-defined low-risk group. Conclusions In our study, the CURSI-defined low-risk group had a higher 30-day mortality than the low-risk groups defined by CURB-65 and CRB-65. Lowering the cut-off value for the CURSI low-risk group would result in a mortality comparable to the CURB-65-defined low risk group. Even then, however, CURSI does not perform better than the established risk scores.