Detecting changes in cough rates is complicated by significant stochastic variability. This study monitored cough continuously in a large cohort, allowing for comparison of cough estimates from different observation periods. Intermittent sampling as 24-hour monitoring can often be misleading.
Authors: Juan C. Gabaldón-Figueira, Eric Keen, Matthew Rudd, Virginia Orrilo, Isabel Blavia, Juliane Chaccour, Mindaugas Galvosas, Peter Small, Simon Grandjean Lapierre, Carlos Chaccour
This paper explores the complexities involved in monitoring cough frequency to detect changes in clinical status. The study involved 616 participants, who cumulated a mmonitoring period longer than nine person-years, registering a total of 62,325 coughs. This comprehensive monitoring was achieved using smartphone-based acoustic artificial intelligence software developed by Hyfe AI.
One of the key findings is the stochastic nature of an individual's cough patterns. The data revealed that these patterns follow a binomial distribution, indicating significant variability both within and between days. A critical insight from the study is the challenge of accurately detecting changes in cough frequency. The researchers found that relying on intermittent sampling, such as 24-hour monitoring, can lead to inaccurate estimates of changes in cough frequency. This inaccuracy is particularly pronounced in individuals with low cough rates or those exhibiting high variance in their cough patterns.
The study's findings are significant in the context of using cough frequency as a tool for monitoring clinical status changes, especially in respiratory diseases. It highlights the importance of prolonged and continuous monitoring to obtain reliable data, which is more representative of an individual's cough pattern over time. The study's results emphasize the limitations of short-term cough monitoring and the potential risks of misinterpretation in clinical assessments based on such limited data.
The paper also discusses the implications of these findings for the development and use of digital health tools, particularly those employing artificial intelligence for passive cough monitoring. It underscores the need for careful consideration of the duration and method of cough monitoring to ensure accurate and meaningful clinical interpretations.