This secondary data analysis project resulted in a preoperative calculator (tool) identifying thoracic surgery patients at risk for developing postoperative pneumonia (POP). The National Surgical Quality Improvement Program (NSQIP) database was used to collect data on 23 predictors identified in the literature to be associated with pneumonia in thoracic surgery patients. Variables were analyzed to determine significant predictors for POP following thoracic surgery. Three methods for identifying significant predictors were used: logistic regression (Model 1), machine learning with eXtreme Gradient Boosting (XGBoost, Model 2), and an analysis of a thoracic surgery expert panel survey (Method 3). Model 1 identified nine significant risk factors (p < .05; AUC ROC = .74, 10-fold cross-validated AUC = .72) used to create the preoperative risk-based POP calculator. Model 2 (10-fold cross-validated AUC = .75) validated Model 1. The expert panel method (Method 3) did not validate Models 1 and 2 or fit the data well (AUC ROC = .6), but it independently identified three of the nine variables used in the risk-based tool. The nine risk factors (p < .05) used in the risk-based calculator were sepsis (effect size [ES] 1.43), SIRS (ES 1.04), male gender (ES 0.77), bleeding disorder (ES 0.57), current smoker within one year (ES 0.39), disseminated cancer (ES 0.39), hypoalbuminemia (ES 0.33), history of severe COPD (ES 0.31), and anemia (ES 0.05). The findings highlight that logistic regression and machine learning models are able to classify POP patients more accurately than expert opinions, and, more importantly, that the developed risk-based tool is valid. Certified Registered Nurse Anesthetists can use the tool to preoperatively identify at-risk thoracic surgery patients, implement perioperative measures to optimize patients prior to surgery, and prevent the occurrence of POP following thoracic surgery.