Artificial intelligence (AI) offers new opportunities to improve asthma management across the care process. This review synthesizes evidence on AI applications in diagnosis, classification, monitoring, and treatment. We conducted a narrative review based on a PubMed search (1995-2025). Eligible studies included peer-reviewed reports on AI applied to diagnosis, monitoring, prediction, and treatment of asthma. Outcomes of interest included diagnostic accuracy, risk prediction, adherence, and clinical decision support. Of 943 records screened, 32 studies met the inclusion criteria. Diagnostic tools integrating clinical data and data from objective tests performed best: deep neural networks combining spirometry and bronchial challenge tests achieved up to 98% accuracy, and automated pulmonary function test interpretation outperformed specialists in consistency and accuracy. Acoustic analyses of cough and respiratory sounds demonstrated sensitivity and specificity above 90%, supporting remote monitoring. In prediction of exacerbations, the electronic health record combining peak expiratory flow and symptom data achieved areas under the curve of up to 0.85. Unsupervised clustering approaches provided clinically meaningful asthma phenotypes. Evidence on treatment optimization remains scarce. AI has the potential to enhance diagnostic accuracy, phenotyping, and monitoring in asthma. However, most studies remain proof-of-concept, with limited external validation and little evidence of clinical impact. Future research should prioritize pragmatic trials, responder stratification, and real-world implementation to confirm clinical and cost-effectiveness.
Almonacid et al. (Fri,) studied this question.