Objective:

To investigate the use of machine learning-based LASSO to predict the risk of short- and long-term clinically driven target-lesion revascularization (TLR) compared with conventional Cox regression methods.

Approach:

• Study Design: A register-based study using data from the Western Denmark Heart Registry, including 24,360 patients treated with PCI and stent implantation from 2002 to 2022.
• Predictive Models: Developed models for TLR at 0–1 and 1–5 years post-PCI using full Cox regression, stepwise variable selection, and ML-based Cox-LASSO.
• Statistical Analysis: Compared models using Harrell's C-index and assessed model discrimination with the log-rank test.

Key Findings:

• Full Cox and stepwise Cox models performed equally at 0–1 years (Harrell's C 0.6743).
• Cox-LASSO showed a minor improvement in predictive performance for short-term TLR (0.6774).
• Stepwise Cox had the best predictive performance at 1–5 years (0.6831), not outperformed by Cox-LASSO (0.6818).
• Most identified risk factors for TLR were consistent across both Cox models and Cox-LASSO.

Interpretation:

The ML-based Cox-LASSO model did not improve predictive performance over conventional Cox regression models for TLR.

Limitations:

• The models demonstrated intermediate predictive performance and require further validation.
• Current models may not be precise enough for definitive bedside decision-making for individual patients.

Conclusion:

The study indicates that while machine learning methods like Cox-LASSO are explored for risk stratification, they do not yet surpass traditional Cox regression in predictive accuracy for TLR.