Truace
Subscribe

American Association for Thoracic Surgery Quality Gateway Procedural Outcomes Models for General Thoracic Surgery Quality Assurance

To develop quality assurance models for adverse procedural events after lung resection and esophagectomy for cancer. Using procedural outcomes after 16,430 lung resections and 3,068 esophagectomies for cancer at 9 hospitals in 7 hospital systems, we developed 5 individual and 2 composite outcome models from preoperative information. Random forest quantile classification for imbalanced data with model-free variable priority screening identified predictors. Out-of-bag prediction provided regularized estimates and…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-2118· cite & verify
American Association for Thoracic Surgery Quality Gateway Procedural Outcomes Models for General Thoracic Surgery Quality Assurance
Adenocarcinoma - CT scan (5499628365) by Yale Rosen from USA. CC BY-SA 2.0 · https://creativecommons.org/licenses/by-sa/2.0

In brief

Researchers analyzed procedural outcomes from 16,430 lung resections and 3,068 esophagectomies for cancer across 9 hospitals in 7 hospital systems. They trained random forest models on preoperative information to produce five individual and two composite models of adverse procedural events.

The models could help surgeons and hospital systems judge patient risk and compare institutions fairly for quality assurance. They are described as initial models, and their predictive performance was moderate, so their reliability in routine clinical use is not yet established.

Main points

  1. Of 71 lung resection and 63 esophagectomy preoperative variables, 35 and 29 were identified as important predictors across the seven outcome models.
  2. Important predictors included demographics, advanced cancer, organ dysfunction, higher acuity, and planned surgical approach and resection extent.
  3. Continuous risk factors, particularly pulmonary function, carried high-importance information.
  4. For lung resection operative mortality, hospital-specific contrasts stayed stable when each focal hospital was excluded from the benchmark.

The gain

Random forest models using preoperative variables can estimate adverse procedural event risk after lung and esophageal cancer surgery, giving surgeons and hospitals tools for risk assessment and quality improvement.

The rundown

The method used random forest quantile classification for imbalanced data with model-free variable priority screening to select predictors. Out-of-bag prediction supplied regularized estimates and internal validation, and a hospital-holdout analysis tested whether institutional benchmarks were stable.

Out-of-bag geometric mean ranged from 0.588 to 0.713 for lung resection and 0.517 to 0.649 for esophagectomy. Risk-factor magnitude and risk-adjusted outcomes differed among hospitals, assessed through counterfactual inference within a framework of virtual twins.

Sources

  1. Peer-reviewedThe Journal of Thoracic and Cardiovascular Surgery2026-10-20

The debate