“…Multivariable models were constructed including all variables independent of significance on univariable analysis. Multivariable hierarchical logistic regression was used to estimate the odds of morbidity, major morbidity, mortality, and failure to rescue by first constructing a model utilizing all the other variables and then entering either gender, race, or ethnicity as the main variable of interest to assess for additional predictive ability 21–23 . The following variables were used in our models: age, body mass index (BMI), emergency case, operative approach, diabetes, smoking, dyspnea, functional status, ventilator‐dependent, history of chronic obstructive pulmonary disease (COPD), ascites, history of congestive heart failure, hypertensive medications, renal failure, dialysis, disseminated cancer, wound infection, steroid use, weight loss, bleeding disorder, transfusion requirement, preoperative sepsis, preoperative sodium, preoperative blood urea nitrogen (BUN), preoperative creatinine, preoperative albumin, preoperative total bilirubin, preoperative serum glutamic oxaloacetic transaminase (SGOT), preoperative alkaline phosphatase, preoperative jaundice, preoperative biliary stent placement, chemotherapy within 90 days, radiation therapy within 90 days, pancreatic duct size, pancreatic gland texture, malignant disease type, T (tumor) stage, N (node) stage, and M (metastases) stage.…”