Additional predictor variables included age at time of hospital admission, sex, baseline Charlson Comorbidity Index score,21 Acute Physiology and Chronic Health Evaluation (APACHE) III score calculated 1 h after ICU admission, coronary artery disease (CAD), cerebrovascular disease (CVD), statin therapy at the time of hospitalization, and predicted risk of developing ALI. We elected to use the APACHE III score at 1 h after ICU admission rather than the traditional APACHE III score at 24 h following ICU admission due to the potential for cause-effect inversion with the latter. Predicted risk of ALI was determined using the Lung Injury Prediction Score (LIPS).22 This score weighs the following variables to calculate a predicted risk of developing ALI while in the hospital: high-risk trauma, high-risk surgery, aspiration, sepsis, shock, pneumonia, pancreatitis, alcohol abuse, smoking, hypoalbuminemia, tachypnea, Fio2 > 35%, chemotherapy, and diabetes mellitus. Standardized definitions were used to identify these variables (alcohol abuse,23,24 smoking,24,25 hypoalbuminemia,14,26 tachypnea,14,27 chemotherapy,27,28 diabetes mellitus14,29). Manual review of the medical record by a trained critical care investigator was used to determine the presence of these variables of interest.