Objectives: The estimation of risk-adjusted in-hospital mortality is essential to allow each thoracic surgery team to be compared with national benchmarks. The objective of this study is to develop and validate a risk model of mortality after pulmonary resection.
Methods: A total of 18,049 lung resections for non-small cell lung cancer were entered into the French national database Epithor. The primary outcome was in-hospital mortality. Two independent analyses were performed with comorbidity variables. The first analysis included variables as independent predictive binary comorbidities (model 1). The second analysis included the number of comorbidities per patient (model 2).
Results: In model 1 predictors for mortality were age, sex, American Society of Anesthesiologists score, performance status, forced expiratory volume (as a percentage), body mass index (in kilograms per meter squared), side, type of lung resection,extended resection, stage, chronic bronchitis, cardiac arrhythmia, coronary artery disease, congestive heart failure, alcoholism, history of malignant disease, and prior thoracic surgery. In model 2 predictors were age, sex, American Society of Anesthesiologists score, performance status, forced expiratory volume, body mass index, side, type of lung resection, extended resection, stage, and number of comorbidities per patient. Models 1 and 2 were well calibrated, with a slope correction factor of 0.96 and of 0.972, respectively. The area under the receiver operating characteristic curve was 0.784 (95% confidence interval, 0.76-0.8) in model 1 and 0.78 (95% confidence interval, 0.76-0.797) in model 2.
Conclusions: Our preference is for the well-calibrated model 2 because it is easier to use in practice to estimate the adjusted postoperative mortality of lung resections for cancer.
Copyright Â© 2011 The American Association for Thoracic Surgery. Published by Mosby, Inc. All rights reserved.