B24-17 Auxiliary Diagnosis of COPD With Cough Search and Portable Spirometry: A Multicenter, Prospective, Comparative Validation Study

Abstract Rationale Chronic Obstructive Pulmonary Disease (COPD) is widely underdiagnosed, leading to significant morbidity and healthcare costs. Novel, accessible screening tools are needed. This study aimed to evaluate the diagnostic accuracy of an multi-modal artificial intelligence (AI)-based cough analysis software Cough Search compared to a portable spirometry for the auxiliary diagnosis of COPD (Figure 1 a). Methods This was a prospective, multicenter, comparative study conducted at four hospitals covering North, South, East and West in China. Adults ( =30 years) with suspected COPD or at high risk were enrolled. Each participant underwent evaluation with the Cough Search cough analysis software (Figure 1 b) and a portable spirometry (index tests). The reference standard was a clinical diagnosis of COPD based on the GOLD 2024 guidelines, including post-bronchodilator spirometry (FEV1/FVC 0.70) with conventional spirometry in hospital. The primary endpoints were the sensitivity and specificity of each tool when used to assist physician diagnosis. Results A total of 378 participants (116 COPD, 262 non-COPD) were included in the final analysis. As a standalone tool, the Cough Search software demonstrated a sensitivity of 92% (95% CI, 86%-96%) and a specificity of 89% (95% CI, 84%-92%). The portable spirometry showed a sensitivity of 80% (95% CI, 72%-87%) and a specificity of 88% (95% CI, 83%-91%).When used to assist physicians, the Cough Search + physician diagnosis achieved a sensitivity of 93% (95% CI, 87%-97%) and a specificity of 94% (95% CI, 90%-96%). This was superior in sensitivity (difference = 8.6%; 95% CI, 0.6%-16.7%; p = 0.018) and non-inferior in specificity (difference = 0.3%; 95% CI, -3.8%-4.6%; p = 0.005) compared to the portable spirometry + physician diagnosis. Subgroup analyses showed the Cough Search maintained high sensitivity across age, sex, and smoking status, particularly outperforming the portable spirometry in older adults ( =65 years), smokers, and patients with early-stage (GOLD 1-2) COPD (Figure 1 c-f). Conclusions An AI-powered cough analysis algorithm demonstrated superior sensitivity and non-inferior specificity in assisting physician diagnosis of COPD compared to a portable spirometry. This non-invasive, rapid tool shows significant promise for improving early detection and screening for COPD, especially in high-risk populations. This abstract is funded by: None