OBJECTIVE: To identify variables contributing to interfacility differences in mortality among residents of long-term care facilities who have lower respiratory tract infection. DESIGN: Multicenter, prospective, 1-year observational study. SETTING: Twenty-one long-term care facilities in 4 geographic areas of Canada. PARTICIPANTS: Residents of long-term care facilities prescribed antimicrobials for treatment of lower respiratory tract infection. METHODS: Mortality rates were calculated for 3 definitions of lower respiratory tract infection: episodes with a clinical or radiographic diagnosis and treated with antimicrobials (definition 1); episodes with a physician diagnosis of pneumonia (definition 2); and episodes with chest radiography findings consistent with pneumonia (definition 3). Multilevel modeling was used to evaluate variables describing premorbid resident status, clinical presentation, management, and facility characteristics. Multivariable models were developed to identify independent predictors of mortality and determine whether facility-level variables remained independently associated with mortality rate after incorporation of individual-level variables. RESULTS: Facility mortality rates varied from 0% to 17.8% for definition 1, from 0% to 47.1% for definition 2, and from 0% to 37.5% for definition 3. There were significant differences in mortality rate depending on which definition was used; for definitions 1 and 2, there were significant differences in mortality rate across facilities. Poorer premorbid resident status and a more severe presentation remained independent predictors of mortality in the multivariable analysis. There were also significantly increased mortality rates for episodes in which a fluoroquinolone was prescribed for initial treatment. For definitions 1 and 3, facility-level variables remained independently associated with mortality rate in the final multivariable model. CONCLUSIONS: Rates of mortality due to lower respiratory tract infection varied among long-term care facilities and differed within a facility, depending on the definition applied. Variables describing premorbid resident status, severity of presentation, and management did not fully explain the variation in mortality rate. Some facility-level variables remained independent predictors of mortality.