This study describes the epidemiology of community-acquired pneumonia (CAP) in elderly Australians for the first time. Using a case-cohort design, cases with CAP were in-patients aged > or = 65 years with ICD-10-AM codes J10-J18 admitted over 2 years to two tertiary hospitals. The cohort sample was randomly selected from all hospital discharges, frequency-matched to cases by month. Logistic regression was used to estimate risk ratios for factors predicting CAP or associated mortality. A total of 4772 in-patients were studied. There were 1952 cases with CAP that represented 4% of all elderly admissions: mean length of stay was 9.0 days and 30-day mortality was 18%. Excluding chest radiograph, 520/1864 (28%) cases had no investigations performed. The strongest predictors of CAP were previous pneumonia, history of other respiratory disease, and aspiration. Intensive-care-unit admission, renal disease and increasing age were the strongest predictors of mortality, while influenza vaccination conferred protection. Hospitalization with CAP in the elderly is common, frequently fatal and a considerable burden to the Australian community. Investigation is ad hoc and management empirical. Influenza vaccination is associated with reduced mortality. Patient characteristics can predict risk of CAP and subsequent mortality.