We demonstrated that the standard clinical criteria of fever, leukocytosis, purulent sputum, and infiltrate on chest radiograph are nonspecific for the diagnosis of post-traumatic pneumonia, and only approximately 50 per cent of patients with these conditions have pneumonia. Quantitative cultures of bronchoalveolar lavage effluent will differentiate pneumonia (requiring antibiotic therapy) from systemic inflammatory response syndrome (not requiring antibiotics). Early identification of patients at risk for pneumonia can target populations for clinical research. Because risk factors for pneumonia when diagnosed by quantitative cultures have not been defined we reviewed our recent experience to identify variables predictive of pneumonia. Patients over a 22-month period who survived > 48 hours were identified from the trauma registry. Pneumonia was defined as growth of > or = 10(5) organisms per milliliter in the bronchoalveolar lavage effluent. Risk factors evaluated included injury severity and severity of shock. There were 7503 patients (75% with blunt and 25% with penetrating injuries). The incidence of pneumonia was 6 per cent (7% of patients with blunt and 2% of patients with penetrating injuries). Logistic regression analysis identified age; Glasgow Coma Scale score; Injury Severity Score; transfusion requirements during resuscitation; spinal cord injury; chest injury severity; and emergent femur fixation, craniotomy, and laparotomy as being independent predictors of pneumonia. We conclude that multiple risk factors, which are all able to be determined early after injury, are predictive of post-traumatic pneumonia. Prompt identification of this high-risk group of patients allows prognostic considerations relative to patient management schemes and targets populations for prophylactic measures or immunomodulation.