Multi-Algorithm Machine Learning Model Enhances Exacerbation Risk Prediction in COPD

Key Details

• 1Retrospective analysis included 878 COPD patients with detailed clinical, biochemical, and pulmonary function data.
• 2Ninety-one machine learning models were tested; the best model combined stepwise Cox regression and random survival forest algorithms.
• 3Model training/testing split was 7:3; validation included ROC analysis, k-fold cross-validation, and subgroup analyses.
• 4Five biochemical, six demographic, and three pulmonary parameters were significant predictors for AECOPD risk.
• 5The integrated model significantly outperformed traditional predictive models (p < .05).
• 6A clinical risk score tool and online prediction platform were developed based on the model.