Résumé
Background: The severity of community acquired pneumonia (CAP) can be evaluated by the PSI and CURB-65 scales. However, it is unknown whether their predictive capacity varies according to the etiology of the disease. Aim: To compare the performance of these scales in adults with viral, bacterial, mixed, and no agent detected CAP. Material and Methods: We studied 725 patients hospitalized for CAP aged 18 to 95 years (47% females) Urinary S. pneumoniae and Legionella antigens were detected by immuno-chromatography (Binax®). Respiratory viruses and bacteria were detected by PCR in nasopharyngeal smears. The proportions of deaths, admission to the intensive care unit (ICU), and oxygen therapy were compared between mild and non-severe patients defined by PSI (I/II and I-III) and CURB-65 (1 and 1-2), according to the causative agent. Results: Ten percent of patients died. A causative agent was detected in 65%. The proportion of mild and non-severe patients according to PSI and CURB-65, and of deceased patients, admitted to the ICU and with oxygen therapy was similar in the four categories per agent. There were no deaths among non-severe patients with bacterial CAP. However, 6% of patients with CAP caused by virus or without causative agents, died. No deaths occurred among mild patients with bacterial CAP. In viral CAP, no deaths occurred among patients classified as mild only by PSI. The yields of PSI were greater than those of CURB-65 in non-severe patients who died and were admitted to the ICU with bacterial and viral CAP (5 and 14%; 7 and 12% respectively, p = 0.04). Conclusions: The prognostic performance of PSI in CAP varies according to the causative agent in adults. It is higher in non-severe bacterial cases, and superior to CURB-65.