A prediction model for severe postoperative hypoxemia after surgery for Standford type A aortic dissection

To study the risk factors of severe postoperative hypoxemia after surgery for Standford type A aortic dissection and establish a prediction model.Data of 411 consecutive patients from January 2014 to April 2015, who underwent surgery for Standford type A aortic dissection in the department of cardiovascular surgery of Beijing Anzhen Hospital, were retrospectively analyzed. All the cases were divided into two groups according to the appearance of severe postoperative hypoxemia. All the data about potential risk factors was put into the database and analyzed by logistic regression. The prediction model was then established upon acquired independent risk factors. Discrimination and calibration of the prediction model were assessed with ROC curve and Hosmer and Lemeshow goodness of fit test.The perioperative in-hospital mortality was 6.57%(27/411). Severe postoperative hypoxemia (PaO2/FiO2≤100 mmHg) happened in 69 cases within 48 hours after procedures, with an incidence rate of 17.1%. The logistic regression demonstrated that body mass index (BMI), age, preoperative serum myoglobin, preoperative serum creatinine, preoperative serumalanine aminotransferase, the time of cardiopulmonary bypass, re-exploration within 48 hours after procedures were the independent risk factors for severe postoperative hypoxemia. The prediction model was then established based on these independent risk factors. The area under ROC curve of the model was 0.785, and the P value in Hosmer and Lemeshow goodness of fit test was 0.625.The logsitic model built in this study succeeded to predict the incidence of severe postoperative hypoxemia after surgery for Standford type A aortic dissection, and it could meet the doctors' requirement with its excellent discrimination and calibration.