OBJECTIVES: Contemporary community-acquired pneumonia (CAP) guidelines recommend that intensive care unit (ICU) patients empirically receive a beta-lactam plus a macrolide (BLM) or a beta-lactam plus a fluoroquinolone (BLF). This study compares the impact of the initial antibiotic choice on time to clinical stability (TTCS), time to switch therapy (TTST), length of hospital stay (LOS), and in-hospital mortality. METHODS: Patient demographics, laboratory and physical exam findings, empiric antibiotic therapy, and hospital course (over 200 variables in all) were extracted from the medical records of all adult CAP patients admitted to the ICUs of 5 community hospitals between 1 November 1999 and 30 April 2000. Patients were divided into two groups (BLM and BLF) based on antibiotics received within the first 24h of hospitalization. TTCS, TTST, LOS and in-hospital mortality were compared using regression models that included the outcome as the dependent variable, antibiotic therapy as the independent variable, and Pneumonia Severity of Index (PSI score) as a covariate. RESULTS: Overall, there were 129 ICU patients of which 34% received BLF and 13% received BLM. Groups were similar with respect to age, sex, comorbidities, PSI score, pre-admission antibiotics, shock, acute renal failure, and the need for mechanical ventilation. Compared to patients who received BLM, those who received BLF had a significantly longer (median; OR, 95%CI) TTCS (2 vs. 4 days; 1.54, 1.04-2.20), TTST (6 vs. 10 days; 1.47, 1.06-2.00), and LOS (6.5 vs. 12 days; 1.58, 1.11-2.11). In-hospital mortality was similar between patients who received BLM and BLF (13% vs. 15%; 1.17, 0.23-8.95). CONCLUSION: Severe CAP patients who initially received a beta-lactam plus a macrolide had a faster time to clinical stability, shorter time to switch therapy, and reduced time to hospital discharge compared to patients treated with a beta-lactam plus a fluoroquinolone.