Machine Learning-Based Model for Predicting Severe Exacerbations in Adult-Onset Type 2 Inflammatory Asthma

Area of Science:

• Pulmonary Medicine
• Computational Biology
• Data Science

Background:

• Machine learning is increasingly used for predicting acute asthma exacerbations.
• Asthma's inflammatory phenotype heterogeneity necessitates tailored predictive models.
• Developing phenotype-specific models is crucial for accurate clinical prediction.

Purpose of the Study:

• To develop predictive models for severe exacerbations in adult-onset type 2 inflammatory asthma.
• To identify independent risk factors associated with severe asthma exacerbations.
• To facilitate early diagnosis and intervention, potentially reducing healthcare costs.

Main Methods:

• Retrospective analysis of 509 patients with type 2 inflammatory asthma exacerbations.
• Categorization into mild-to-moderate (n=300) and severe (n=209) exacerbation groups.
• Evaluation of four machine learning models: decision trees, logistic regression, random forests, and LightGBM.

Main Results:

• Low ACT scores, low FEV1/FVC ratio, diabetes, high neutrophil count, and family history of asthma were independent risk factors.
• LightGBM achieved the highest AUC (0.9344), outperforming other models.
• Key predictors identified by LightGBM included ACT score, FEV1/FVC ratio, and age.

Conclusions:

• A LightGBM-based prediction model for severe adult-onset type 2 inflammatory asthma exacerbations was developed.
• The model demonstrates strong predictive performance, aiding early detection.
• This tool can support timely intervention and prevention strategies for severe asthma exacerbations.