A cohort of 1,391 patients with community-acquired pneumonia of unknown etiology, atypical pneumonia, Legionella pneumophila pneumonia, viral pneumonia, or pneumococcal pneumonia was studied according to a standard protocol to analyse whether the addition of a macrolide to beta-lactam empirical treatment decreases mortality rates. Patients admitted to the intensive care unit were excluded. Severity was assessed using the PORT score. An etiological diagnosis was achieved in 498 (35.8%) patients (292 infections due to Streptococcus pneumoniae). Treatment was chosen by the attending physician according to his/her own criteria: beta-lactam agent in 270 and beta-lactam agent plus a macrolide in 918 cases. The mortality rate was 13.3% in the group treated with a beta-lactam agent alone and 6.9% in the group treated with a beta-lactam agent plus a macrolide (p=0.001). The percentage of PORT-group V patients was 32.6% in the group treated with a beta-lactam agent alone compared to 25.7% in the group who received a beta-lactam agent plus a macrolide (p=0.02). After controlling for PORT score, the odds of fatal outcome was two times higher in patients treated with a beta-lactam agent alone than in those treated with a beta-lactam agent plus a macrolide (adjusted OR = 2, 95%CI 1.24-3.23). The results suggest that the addition of a macrolide to an initial beta-lactam-based antibiotic regimen is associated with lower mortality in patients with community-acquired pneumonia, independent of severity of infection, thus supporting the recommendation of a beta-lactam-agent plus a macrolide as empirical therapy.