Prediction modeling of postoperative pulmonary complications following lung resection based on random forest...

Area of Science:

• Thoracic Surgery
• Medical Informatics
• Pulmonary Medicine

Background:

• Postoperative pulmonary complications (PPCs) are a major cause of morbidity and mortality following lung resection.
• Effective early detection and intervention strategies for PPCs are crucial for improving patient outcomes.
• Existing risk prediction models may lack accuracy or clinical applicability.

Purpose of the Study:

• To develop and validate a risk prediction model for PPCs after lung resection.
• To utilize the random forest (RF) algorithm for enhanced predictive accuracy.
• To identify key predictive factors for PPCs in lung resection patients.

Main Methods:

• Retrospective analysis of 180 lung resection patients.
• Development of a random forest (RF) model using Python.
• Evaluation of model performance using ROC, calibration, and decision curve analyses.

Main Results:

• The RF model demonstrated favorable discrimination and calibration capabilities.
• Key predictors of PPCs included blood loss, maximal resection length, lymph node count, FEV1, and age.
• The model showed potential for clinical applicability in predicting PPCs.

Conclusions:

• The random forest algorithm is effective for predicting PPCs post-lung resection.
• The developed model shows promise for clinical application in early detection and intervention.
• Accurate PPC prediction can lead to improved patient management and reduced complications.