RR\ID 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: Authors main claim is train models using early historical data can exhibit strong predictive capabilities.
Below are some line edits:
Data:
“However, for the sake of effectiveness and accuracy, only those characteristics demonstrating sufficient predictive power were included in our analysis”: How do you determine the predictive power of the variables before taking into the final consideration. Did you evaluate them by any statistical test?
“information. Category 1 includes diabetes”: Category and Stage is same?
Category and group is same? Please keep the same jargons throughout.
“Category 2 includes gender, age, state of birth, state of residence, and age.”: Age is mentioned twice here
“This implies that when carrying out a prospective study for patients in EW-5, there is uncertainty as to the specific day when a patient transitions from one stage to another.”: Not clear. Please paraphrase these lines or give more explanation.
Neural Network:
“Supervised machine learning provides computer algorithms with the ability to learn from a known dataset to identify features and generate predictions about the outcome given a specific set of features, not included in the learning stage.”: What Supervised machine learning model you run? Did you compare the model performance between the different models run?
“In this context, by making use of the publicly available database of Mexican COVID-19 patients, reference [17] reported on an artificial neural network capable…”: Why artificial neural network was used? Why not traditional supervised machine learning was used in this study?
“…classifying patients into two classes: a) patients who are more likely to recover than to die or b) patients who are more likely to die than to recover.”: Binary classes: mortality or survived, right?
“EW-2 through EW-5, irrespective of whether they lie within or outside the date span for the training data. To ensure a fair comparison in all cases, the neural network architecture is in all cases identical to that reported in Ref. [17].”: It is not clear that why you trained neural networks with the data from EW 2 through EW5? Why did you not include EW1 and EW6? It is also not clear that whether they lie within or outside the date span?
Result:
“As mentioned earlier, we have carried out our prospective study in two phases”: But this is based on historical data?
“In figure 3 we plot the number of new confirmed patients plotted vs date throughout the date span covered by the database, as was also done in Fig, 1a.”: Please describe in detail.
“…84.5% the clinical outcome for patients who are tracked from Stage 2, i.e.”: How did you get this value? Not clearly understandable.
Figure 1a and 1b:
1a: There are 2 peaks out of the 25th percentile. Please describe more regarding this results/figure.
Caption: “gray regions indicate inter-pandemic periods (IP)”: How do you define IP, but many died in this period.
Needs more description for both figures
Figure 2:
Patient Evolution Graph: Not clear too. Diagram needs sub categories like a,b,c
Prospective Study Diagram: Needs more descriptions. What these arrow mean?
Caption: “flow diagram in the bottom-left panel illustrates”: Instead denoted by alphabetical orders: a,b,c.
Figures need more description. Please describe all sub-categories of figures
Table 2:
Table 3:
Figure 3:
Figure 4:
“Prediction accuracy plotted, for each of stages 1-4 as labeled, vs the length of the training data date span, indicated in the horizontal axis as the most recent wave included”: Not clear.
“The accuracy values are obtained upon applying our neural networks to testing data resulting from our patient protocol tracking protocol…”: Not clear. May be the problem of language.
References:
Overall comments:
Check the language: A lot of confusing jargon is used. In addition, please keep using the same jargon throughout the paper.
More description of the figures. For example, Figure 1 could use more description. Also, how do you define inter-pandemic periods, but many died during this period?
What is the definition of inter-pandemic periods (IP)?
How did you determine the predictive power of the variables before taking it into final consideration? Did you evaluate them using any statistical test?
What supervised machine learning model was run? Did you compare the model performance between the different models run?
What artificial network was used?
Data is based on historical data, but you mention that this is a prospective
Whether there is a repository of coding.
Creative Commons Attribution 4.0 International License (CC-BY 4.0)
