Mobile Phone Auscultation to Diagnose Influenza A in Patients Presenting to Three Different Emergency Departments

Influenza is a major global health burden, causing widespread illness, complications, and high economic costs each year. Current diagnostic methods and traditional auscultation (MPA) are limited by delays, subjectivity, and accessibility challenges. Mobile phone auscultation with computational modeling offers a promising rapid, scalable, and noninvasive alternative for detection.

To evaluate the feasibility and preliminary diagnostic performance of MPA in distinguishing influenza A from a heterogeneous group of respiratory conditions and controls. Successful classification would support remote influenza A diagnosis using telehealth audio recordings, expanding access for underserved populations.

In this prospective cohort study, participants were enrolled into five predefined groups: influenza A, pneumonia, acute bronchitis, other respiratory illness, and controls. Subjects were recruited from three emergency departments. Lung sound recordings were collected using unmodified mobile phones. Computational nonlinear biofluid dynamics techniques were applied to extract features, including Maximal Lyapunov Exponent (MLE) and Correlation Dimension (Dcorr). Train-test sets were created by 80/20 clustered random sampling. Time Series Dynamics models were fitted, and logistic regression classifiers were developed to distinguish influenza A from comparison groups.

A total of 294 subjects were enrolled, including 59 with influenza A. Baseline differences between influenza A and comparison cohorts were limited to race. All complete recordings were analyzed. Modeling performed well, generating only one false negative. Sensitivity was 92%, specificity 85%, and area under the curve (AUC) 89%.

MPA demonstrates feasibility and moderate discriminatory ability for identifying influenza A from a heterogeneous set of respiratory conditions and controls. These findings support further investigation of MPA as a potential adjunctive or screening tool, particularly in resource-limited settings. Additional studies are needed to validate performance, define clinical use cases, and determine its role relative to existing diagnostic methods.