Review 1: "The Acoustic Dissection of Cough: Diving into Machine Listening-based COVID-19 Analysis and Detection"

RR:C19 Evidence Scale rating by reviewer:

• Potentially informative. The main claims made are not strongly justified by the methods and data, but may yield some insight. The results and conclusions of the study may resemble those from the hypothetical ideal study, but there is substantial room for doubt. Decision-makers should consider this evidence only with a thorough understanding of its weaknesses, alongside other evidence and theory. Decision-makers should not consider this actionable, unless the weaknesses are clearly understood and there is other theory and evidence to further support it.

***************************************

Review:

The attempt of the authors in this study is well-appreciated. Not only have the authors tried to distinguish between COVID and non-COVID subjects, but they also focused on the explainability of the models by comparing feature relevance. The limitations of the work are clearly described, particularly those related to the COUGHVID dataset.

1. The dataset in terms of COVID (n = 210) vs. non-COVID (n = 1201) classes is highly imbalanced, which could affect the performances of the classifiers. Although it is mentioned that balanced class weights were applied to each SVM model in order to mitigate the data imbalance problem, the corrective measures for other models were not applied or described.

2. The authors used the unweighted average recall to evaluate the classification performance purposefully without considering the data imbalance characteristics. The rationale for this approach should be further elaborated.

3. The UARs of best models exceed 0.5 by a small degree, albeit with statistical significance. Further improvement is needed before they can be used for automatic detection of COVID-19.

4. It would be helpful if the authors can discuss the pros and cons of all seven classifier models and which model is most suitable for the task given the nature of the dataset.

5. The feature analysis in the result section is hard to follow, particularly for readers not well familiar with the feature sets used in the study. More elaboration would be helpful.