Prospective comparison of severity scores for predicting mortality in community-acquired pneumonia.

INTRODUCTION: Specific prognostic models for community acquired pneumonia (CAP) to guide treatment decisions have been developed, such us the Pneumonia Severity Index (PSI) and the Confusion, Urea nitrogen, Respiratory rate, Blood pressure and age ≥ 65 years index (CURB-65). Additionally, general models are available such as the Mortality Probability Model (MPM-II). So far, which score performs better in CAP remains controversial. The objective was to compare PSI and CURB-65 and the general model, MPM-II, for predicting 30-day mortality in patients admitted with CAP. METHODS: Prospective observational study including all consecutive patients hospitalised with a confirmed diagnosis of CAP and treated according to the hospital guidelines. Comparison of the overall discriminatory power of the models was performed by calculating the area under a receiver operator characteristic curve (AUC ROC curve) and calibration through the Goodness-of-fit test. RESULTS: One hundred and fifty two patients were included (mean age 73.0 years; 69.1% male; 75.0% with more than one comorbid condition). Seventy-five percent of the patients were classified as high-risk subjects according to the PSI, versus 61.2% according to the CURB-65. The 30-day mortality rate was 11.8%. All three scores obtained acceptable and similar values of the AUCs of the ROC curve for predicting mortality. Despite all rules showed good calibration, this seemed to be better for CURB-65. CURB-65 also revealed the highest positive likelihood ratio. CONCLUSIONS: CURB-65 performs similar to PSI or MPMII for predicting 30-day mortality in patients with CAP. Consequently, this simple model can be regarded as a valid alternative to the more complex rules.