European risk models for morbidity (EuroLung1) and mortality (EuroLung2) to predict outcome following anatomic lung resections: an analysis from the European Society of Thoracic Surgeons database

Abstract

Objectives: To develop models of 30-day mortality and cardiopulmonary morbidity from data on anatomic lung resections deposited in the European Society of Thoracic Surgeons (ESTS) database.

Methods: Retrospective analysis of 47 960 anatomic lung resections from the ESTS database (July 2007-August 2015) (36 376 lobectomies, 2296 bilobectomies, 5040 pneumonectomies and 4248 segmentectomies). Logistic regression analyses were used to test the association between baseline and surgical variables and morbidity or mortality. Bootstrap resampling was used for internal validation and to check predictors of stability. Variables that occurred in more than 50% of the bootstrap samples were deemed reliable. User-friendly aggregate scores were then created by assigning points to each variable in the model by proportionally weighting the regression coefficients. Patients were grouped in classes of incremental risk according to their scores.

Results: Cardiopulmonary morbidity and 30-day mortality rates were 18.4% (8805 patients) and 2.7% (1295 patients). The following variables were reliably associated with morbidity after logistic regression analysis (C-index 0.68): male sex ( P < 0.0001); age ( P < 0.0001); predicted postoperative forced expiratory volume in 1 s (ppoFEV1) ( P < 0.0001); coronary artery disease (CAD) ( P < 0.0001); cerebrovascular disease (CVD) ( P < 0.0001); chronic kidney disease ( P < 0.0001); thoracotomy approach ( P < 0.0001); and extended resections ( P < 0.0001). All variables occurred in more than 95% of the bootstrap samples. An aggregate score was created that stratified the patients into six classes of incremental morbidity risk ( P < 0.0001). The following variables were reliably associated with mortality after logistic regression analysis (C-index 0.74): male sex ( P < 0.0001); age ( P < 0.0001); ppoFEV1 ( P < 0.0001); CAD ( P = 0.003); CVD ( P < 0.0001); body mass index ( P < 0.0001); thoracotomy approach ( P < 0.0001); pneumonectomy ( P < 0.0001); and extended resections ( P = 0.002). All variables occurred in more than 80% of bootstrap samples. An aggregate score was created that stratified the patients into six classes of incremental mortality risk ( P < 0.0001).

Conclusions: The updated ESTS morbidity and mortality models can be used to define risk-adjust outcome indicators for auditing quality of care and to counsel patients about their surgical risk.

Keywords: Lobectomy; Lung cancer; Morbidity; Mortality; Pneumonectomy; Risk modelling.