ABSTRACT
Background: Cough represents a cardinal symptom of acute respiratory tract infections. Generally associated with disease activity, cough holds biomarker potential and might be harnessed for prognosis and personalised treatment decisions. Here, we tested the suitability of cough as a digital biomarker for disease activity in coronavirus disease 2019 (COVID-19) and other lower respiratory tract infections. Methods: We conducted a single-centre, exploratory, observational cohort study on automated cough detection in patients hospitalised for COVID-19 (n=32) and non-COVID-19 pneumonia (n=14) between April and November 2020 at the Cantonal Hospital St Gallen, Switzerland. Cough detection was achieved using smartphone-based audio recordings coupled to an ensemble of convolutional neural networks. Cough levels were correlated to established markers of inflammation and oxygenation. Measurements and main results: Cough frequency was highest upon hospital admission and declined steadily with recovery. There was a characteristic pattern of daily cough fluctuations, with little activity during the night and two coughing peaks during the day. Hourly cough counts were strongly correlated with clinical markers of disease activity and laboratory markers of inflammation, suggesting cough as a surrogate of disease in acute respiratory tract infections. No apparent differences in cough evolution were observed between COVID-19 and non-COVID-19 pneumonia. Conclusions: Automated, quantitative, smartphone-based detection of cough is feasible in hospitalised patients and correlates with disease activity in lower respiratory tract infections. Our approach allows for near real-time telemonitoring of individuals in aerosol isolation. Larger trials are warranted to decipher the use of cough as a digital biomarker for prognosis and tailored treatment in lower respiratory tract infections.
ABSTRACT
BACKGROUND: Clinical deterioration can go unnoticed in hospital wards for hours. Mobile technologies such as wearables and smartphones enable automated, continuous, noninvasive ward monitoring and allow the detection of subtle changes in vital signs. Cough can be effectively monitored through mobile technologies in the ward, as it is not only a symptom of prevalent respiratory diseases such as asthma, lung cancer, and COVID-19 but also a predictor of acute health deterioration. In past decades, many efforts have been made to develop an automatic cough counting tool. To date, however, there is neither a standardized, sufficiently validated method nor a scalable cough monitor that can be deployed on a consumer-centric device that reports cough counts continuously. These shortcomings limit the tracking of coughing and, consequently, hinder the monitoring of disease progression in prevalent respiratory diseases such as asthma, chronic obstructive pulmonary disease, and COVID-19 in the ward. OBJECTIVE: This exploratory study involved the validation of an automated smartphone-based monitoring system for continuous cough counting in 2 different modes in the ward. Unlike previous studies that focused on evaluating cough detection models on unseen data, the focus of this work is to validate a holistic smartphone-based cough detection system operating in near real time. METHODS: Automated cough counts were measured consistently on devices and on computers and compared with cough and noncough sounds counted manually over 8-hour long nocturnal recordings in 9 patients with pneumonia in the ward. The proposed cough detection system consists primarily of an Android app running on a smartphone that detects coughs and records sounds and secondarily of a backend that continuously receives the cough detection information and displays the hourly cough counts. Cough detection is based on an ensemble convolutional neural network developed and trained on asthmatic cough data. RESULTS: In this validation study, a total of 72 hours of recordings from 9 participants with pneumonia, 4 of whom were infected with SARS-CoV-2, were analyzed. All the recordings were subjected to manual analysis by 2 blinded raters. The proposed system yielded a sensitivity and specificity of 72% and 99% on the device and 82% and 99% on the computer, respectively, for detecting coughs. The mean differences between the automated and human rater cough counts were -1.0 (95% CI -12.3 to 10.2) and -0.9 (95% CI -6.5 to 4.8) coughs per hour within subject for the on-device and on-computer modes, respectively. CONCLUSIONS: The proposed system thus represents a smartphone cough counter that can be used for continuous hourly assessment of cough frequency in the ward.