An enhanced machine learning-based prognostic prediction model for patients with AECOPD on invasive mechanical ventilation.

Chronic obstructive pulmonary disease (COPD) causes irreversible airflow limitations, increasing global morbidity and mortality. Acute exacerbations (AECOPDs) worsen symptoms and may require mechanical ventilation, leading to complications. Understanding factors affecting AECOPD prognosis during mechanical ventilation is crucial. Inspired by rime ice physics, the RIME algorithm has been proposed but it had limitations in feature selection and solution space exploration. We improve RIME by adding a dispersed foraging mechanism and differential crossover operator, creating DDRIME. Our study analyzes patient data to identify factors related to invasive mechanical ventilation in AECOPD. DDRIME's performance is tested against RIME on 83 functions and 12 public datasets for feature selection. It outperformed most algorithms, with bDDRIME_KNN showing high accuracy in predicting AECOPD outcomes. Key indicators-chronic heart failure (CHF), D-dimer (D-D), fungal infection (FI), and pectoral muscle area (PMA)-predicted prognosis with >0.98 accuracy. bDDRIME is thus a valuable tool for predicting AECOPD patients' outcomes on mechanical ventilation.