Cough biomarkers for diagnosis and monitoring of respiratory disease: a systematic review

Clinical Snapshot

Key Findings

• P Value: Not applicable (systematic review without meta-analysis)

• Effect Size: Sensitivities and specificities frequently reported ≥90% for tuberculosis, asthma, and COVID-19 detection; deep learning models (convolutional and recurrent networks) predominated (n=56/89) and tended to achieve highest accuracies; ML ensemble methods and logistic regression also demonstrated strong performance with well-engineered features

• Primary Outcome: Diagnostic and prognostic performance of cough acoustic biomarker models across three domains: cough detection (n=31 studies), disease classification (n=55 studies), and disease severity prediction (n=3 studies)

• Nnt Or Sensitivity: Sensitivity ≥90% and specificity ≥90% frequently reported across tuberculosis, asthma, and COVID-19 studies; however, these estimates are based predominantly on internal validation only (88.8% of studies lacked external validation) and are likely substantially inflated due to high risk of bias in 71.1–87.6% of included studies

• Confidence Interval: Not reported (no meta-analytic pooling performed; narrative synthesis only)

Clinical Application

Cough acoustic analysis is non-invasive, potentially low-cost, and deployable via smartphone or wearable devices, making it theoretically feasible for point-of-care and telehealth applications. However, current evidence does not support clinical deployment due to lack of external validation, absence of standardised recording protocols, and unresolved data privacy considerations. Implementation would require regulatory clearance, prospective multicentre validation, and integration into existing clinical workflows. In Australia, cough acoustic biomarker tools are not currently TGA-approved as diagnostic devices, and no PBS-listed applications exist. The RACGP does not currently include cough acoustic analysis in guidelines for asthma, COPD, or respiratory infection management. The technology has potential relevance to Australian primary care, particularly for remote and Indigenous community health settings where access to spirometry, chest imaging, and specialist review is limited. The University of Melbourne authorship team (Department of Audiology and Speech Pathology and Department of Anatomy and Physiology) positions this as an area of active Australian research interest. Future TGA regulatory pathways for Software as a Medical Device (SaMD) would be relevant if clinical-grade tools are developed. Tuberculosis screening applications may have relevance in high-risk communities. COVID-19 applications, while prominent in the literature, are of reduced immediate clinical priority in the post-pandemic context. Adults and children with suspected or confirmed respiratory conditions including tuberculosis, asthma, COVID-19, COPD, and other respiratory diseases; potential utility is greatest in low-resource, remote, or community screening settings where conventional diagnostics are inaccessible

Abstract

Cough is a common and physiologically informative component of respiratory morbidity, but its potential for diagnosing and monitoring disease is not thoroughly investigated. This systematic review synthesised the literature on algorithmic and statistical models analysing cough acoustics for diagnosing or monitoring respiratory conditions. Following PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines, five databases (PubMed, Embase, Scopus, Web of Science and CENTRAL) were systematically searched for studies published from January 2010 to June 2025. Eligible studies performed quantitative acoustic feature analysis of human coughs using statistical, machine learning or deep learning models and reported diagnostic or prognostic performance. 89 studies from 34 countries were assessed, covering cough detection (n=31), disease classification (n=55) and disease severity prediction (n=3), reflecting potential applications in disease monitoring. Deep learning approaches, especially convolutional and recurrent networks, were predominant (n=56) and tended to achieve the higher accuracies, although machine learning ensemble methods and logistic regression also demonstrated strong performance, particularly with well-engineered features. Across different diseases, sensitivities and specificities were often reported to be ≥90%, notably for tuberculosis, asthma and COVID-19. However, methodological weaknesses were common, with only 11.2% of studies introducing external validation, 71.1-87.6% demonstrating high risk of bias (according to PROBAST-AI) and most based on small, homogeneous or crowdsourced cohorts with limited generalisability. These limitations contribute to inflated internal performance and uncertainty about real-world applicability. Cough acoustic biomarkers hold promise as an adjunctive tool for screening and longitudinal monitoring in low-resource environments. Nevertheless, widespread implementation will require large, multicentre validation, standardised calibration, bias control and incorporation into privacy-preserving workflows.

References

1. 1.Xu, J., Kathiresan, T., Siddiqui, A., Mazzone, S. B., & Vogel, A. (2026). Cough biomarkers for diagnosis and monitoring of respiratory disease: a systematic review. European Respiratory Review, 35. https://doi.org/10.1183/16000617.0288-2025