Objective:

To develop and validate machine learning-based prediction models for postoperative atrial fibrillation (POAF) and to construct a clinically applicable nomogram for individualised risk estimation.

Key Findings:

• Six predictors of POAF were identified: age, education level, hypertension, marital status, postoperative pain score, and surgical approach.
• Logistic regression (LR) showed the highest AUC (0.855) and accuracy (0.857) in the test cohort, with AUC values in the training cohort ranging from 0.827 to 0.995.
• Calibration curves indicated good agreement between predicted and observed POAF risks for the LR model.

Interpretation:

The study successfully developed a machine learning-assisted risk prediction framework for POAF after lung cancer surgery, demonstrating that LR provided superior stability and clinical utility compared to more complex models.

Limitations:

• The study was conducted at a single center, which may limit generalizability.
• The retrospective design may introduce biases and limit the ability to establish causation.
• The potential for overfitting in machine learning models may affect the robustness of the findings.

Conclusion:

An LR-based nomogram was developed for early postoperative POAF risk assessment, providing a practical tool for perioperative monitoring and personalized management.