Converting a Cough Counter into a Cough Monitor: A Way Forward?

Background/Objective: To identify respiratory pathology, automated cough counting is frequently proposed. A trial validating an early warning system for exacerbations in chronic obstructive pulmonary disease (COPD) patients was recently concluded successfully. This paper aims to review the critical design choices for converting a cough counter into a patient-friendly continual cough monitor. Furthermore, it provides a basis for a practical reliability metric for continual cough monitoring. Methods: Design choices made in the development of a cough-based alert mechanism called XACT are discussed. A practical approach for reliability assessment is outlined based on cough counts, day-to-day variation and specificity data. Results: In post hoc analysis, it is shown that the described approach enables differentiation between high-quality cough estimates and less reliable data. The approach is used to underpin an earlier cohort subdivision into patients with and without increased cough during exacerbation. Conclusions: The validated alert mechanism has various patient-oriented design choices (unobtrusiveness, privacy-preserving). The examples illustrate how to screen for potential issues in automated cough count data without resorting to laborious annotation. It creates a practical basis for confidence metrics of medical inferences made from cough data, e.g., exacerbation forecasts. The proposed concepts need further validation.