Review Reports - Forecasting Model Based on Lifestyle Risk and Health Factors to Predict COVID-19 Severity

Round 1

Reviewer 1 Report

Please, pay attention to split Section 4 into "Discussion" and "Conclusions" sections. I guess that the best split point is before row 207, when you say that "Figure 8 shows the geographical distribution in Italy".

Author Response

Question:

Answer:

Thanks for the interesting suggestion. We added section “Conclusions" in the main text. Furthermore, both the discussions and the conclusions have been enriched with new sentences that better highlight our work. For this reason, the division of the two paragraphs followed a complete meaning of both paragraphs.

Reviewer 2 Report

Dear Authors,

The article makes an interesting contribution to the field on social and health factors related to COVID-19. However, the article requires significant changes in the structure of the text. It needs to be put in order, and requires some additions. I have comments before the manuscript could be considered for further proceeding and publication:

Abstract

1. The Abstract requires changes in the structure, as of the adopted structure for the abstract of the article, for instance, (1) Introduction, (2) Aim of the study, (3) Material and Methods, (4) Results, and (5) Conclusions.

2. Keywords do not fully reflect the results of this study, so I suggest that you complete them, for instance with the new keywords, like lifestyle, flu, vaccination.

Introduction

3. Introduction does not contain comprehensive information on the Italian specificity related to COVID, especially the beginning of the pandemic compared to other European countries. In addition, it should provide information on the diversity of Italian regions, e.g. North-South, which is important in the context of the analysis being carried out. In general, there is a need for better explanation, why Italy is a good example to study COVID-related characteristics.

4. It would also be good to explain the concept of Machine learning strategy and the possibilities of using this tool for COVID analyses.

5. Lines 26-46 from the words „Figure 1 ….. January 2021” describe in fact information of Material and Methods, and should be removed to this section (Materials and Methods). In lines 26-46 is the sentence about limitations: „It is worth mentioning …… by the Italian local health system”. So it should be moved to the section Discussion. In lines 61- 66 is information on the strengths of the study, so this should be moved to the Discussion section as well.

The aim

6. The aim is not clearly stated in the manuscript and should be remade. The same aim should be included in the abstract.

Material and Methods

7. This section is recommended to split into subsections, like for instance, study design, settings, participants, data sources, study sizes, variables, statistical analysis. It is required to describe in detail these subsections.

8. In line 69, you presented the source of the data, but in the Abstract you gave two sources of the data. Please, explain.

9. Abbreviations used in the manuscript, such as GLM, RF, SVM, should appear in full formulation after the first time the term was used.

Results

10. Please provide the table containing descriptive statistics for all variables, with the appropriate division into groups of independent variables.

11. The results require editing and supplementing. In the section Discussion (in lines 180-208), there is a description of the results that should be moved to the Results section. Additionally, I suggest that you combine Figure 4 with Figure 5 and present it as one figure with two panels in a vertical orientation, e.g. (A): Crude Positivity Rate; (B): Crude Mortality Rate. Similarly, combine Figures 6 and 7.

12. Comments for Figures and Tables:

· F1:

No units on the Y axis - this can be included in the figure title. Generally, the title for Figure 1 should be in one sentence. In addition, the months should be changed to English.

· F2 should not have comments. The comments should be moved to the main text.

· F3 tytuł should be: “Correlation matrix for independent features, and Crude Mortality Rate (CMR), and Crude Positivity Rate (CPR).”

· F4, F5:

Change the layout to the panel (as above). Abbreviations CPR, CMR are not needed, it is enough to name the panel appropriately. The abbreviation SVM, GLM can be expanded in the Figure. The title of the new F4 should be corrected, and the comment contained in the title should be moved to the main text.

· F6, F7: Make similar as for F4, F5.

· T1: Table could be bigger, so the abbreviations CPR, CMR, SVM, GLM would be changed into the full text. In addition, in the table, please add MAPE (±SD), Adjusted R²(±SD), and delete comment from the title of the table. Under the table explain abbreviations MAPE and SD.

Discussion

13. In discussion there is lack of own comments for interpretation of the results achieved.

14. Please add additional paragraph with limitations and strengths of the study.

Conclusions

15. On page 15, there are no separate section of conclusions from the conducted study, which should be response to the purpose of the study – please, correct.

General comments:

16. Some sentences are repeating in the manuscript. Please avoid repeating.

Lines 39-40: The comparison between the predictions of a simple linear model and two machine learning algorithms makes our results more robust and reliable.

Lines 103-104: The use of three different kinds of models makes the results of our analysis more robust.

Lines 175-177: The use of three different models makes more robust our results, in particular we started with GLM and then used SVM and RF as more complex and alternative models to the linear hypothesis.

It is recommended to write about it in the one place. In opinion of the reviewer, you can write on it in the paragraph on the strengths of the study.

17.Please, highlight the changes to the revised version using a different colour and/or a different script.

Author Response

Dear Authors,

Abstract

Question:

The Abstract requires changes in the structure, as of the adopted structure for the abstract of the article, for instance, (1) Introduction, (2) Aim of the study, (3) Material and Methods, (4) Results, and (5) Conclusions.

Answer:

We thank the reviewer. We modified the abstract structure as suggested.

Question:

Keywords do not fully reflect the results of this study, so I suggest that you complete them, for instance with the new keywords, likelifestyle, flu, vaccination.

Answer:

We added the following keywords: lifestyle risk factors, flu, vaccination.

Introduction

Question:

Introduction does not contain comprehensive information on the Italian specificity related to COVID, especially the beginning of the pandemic compared to other European countries.

Answer:

As suggested by the reviewer, in the section “Introduction” we included specific information relating the beginning of the first wave of Covid-19 pandemic in Italy, comparing it to other European countries. We also added 3 new references.

Question:

In addition, it should provide information on the diversity of Italian regions, e.g. North-South, which is important in the context of the analysis being carried out. In general, there is a need for better explanation, why Italy is a good example to study COVID-related characteristics.

Answer:

In Italy there are substantial differences between north and south due to several socio-economic and geographical factors that we do not treat in this study. Our goal is to create a general model that can be adapted to the whole Italian territory starting from factors already connected to the spread of the covid from previous works, but which are not linked to the gap between north and south Italy. As the previous title could be misleading we changed it by highlighting the lifestyle risk factors instead of social factors.

Question:

It would also be good to explain the concept of Machine learning strategy and the possibilities of using this tool for COVID analyses.

Answer:

We added a sentence in the manuscript section “Material and Methods”: “The causes of the spread level of the COVID-19 pandemic can be a lot and range from social factors related (lifestyles or mobility) to environmental and health factors. The complex and multifactorial etiology of COVID-19 cannot be suitably explored by standard techniques and traditional approaches. A model that is able to manage data of a different nature could help to better understand the problem. In this context, techniques based on Machine Learning paradigms could be of great help, thanks to their intrinsic ability to manage multimodal data and represent complex problems.”

Question:

Lines 26-46 from the words „Figure 1 ….. January 2021” describe in fact information of Material and Methods, and should be removed to this section (Materials and Methods). In lines 26-46 is the sentence about limitations: „It is worth mentioning …… by the Italian local health system”. So it should be moved to the section In lines 61- 66 is information on the strengths of the study, so this should be moved to the Discussion section as well.

Answer:

We moved these sentences into section Discussion.

The aim

Question:

The aim is not clearly stated in the manuscript and should be remade. The same aim should be included in the abstract.

Answer:

We thank the reviewer for the suggestion, we added sentences in abstract, in introduction and in the discussion sections and modified the title. Our main goal was to study the progress of the early stages of the Covid-19 pandemic in Italy (long before the start of the vaccination campaign), through factors linked to lifestyle and respiratory diseases (factors highlighted by previous works). To do this we use a data driven approach by means of machine learning techniques that can find hidden patterns in the data.

Material and Methods

Question:

This section is recommended to split into subsections, like for instance, study design, settings, participants, data sources, study sizes, variables, statistical analysis. It is required to describe in detail these subsections.

Answer:

Question:

In line 69, you presented the source of the data, but in the Abstract you gave two sources of the data. Please, explain.

Answer:

We added a sentence in section “Data collection” to clarify this point.

Question:

Abbreviations used in the manuscript, such as GLM, RF, SVM, should appear in full formulation after the first time the term was used.

Answer:

Results

Question:

Please provide the table containing descriptive statistics for all variables, with the appropriate division into groups of independent variables.

Answer:

We added a new table in section Results.

Question:

The results require editing and supplementing. In the section Discussion (in lines 180-208), there is a description of the results that should be moved to the Results section. Additionally, I suggest that you combine Figure 4 with Figure 5 and present it as one figure with two panels in a vertical orientation, e.g. (A): Crude Positivity Rate; (B): Crude Mortality Rate. Similarly, combine Figures 6 and 7.

Answer:

Question:

Comments for Figures and Tables:

No units on the Y axis - this can be included in the figure title. Generally, the title for Figure 1 should be in one sentence. In addition, the months should be changed to English.

Answer:

We modified Figure 1 and the relative caption.

Question:

F2 should not have comments. The comments should be moved to the main text.

Answer:

We moved the comment in the main text.

Question:

F3 tytuł should be: “Correlation matrix for independent features, and Crude Mortality Rate (CMR), and Crude Positivity Rate (CPR).”

Answer:

Question:

F4, F5:

Answer:

Question:

F6, F7: Make similar as for F4, F5.

Answer:

Question:

T1: Table could be bigger, so the abbreviations CPR, CMR, SVM, GLM would be changed into the full text. In addition, in the table, please add MAPE (±SD), Adjusted R²(±SD), and delete comment from the title of the table. Under the table explain abbreviations MAPE and SD.

Answer:

Discussion

Question:

In discussion there is lack of own comments for interpretation of the results achieved.

Answer:

In section Discussion we have added our comments to better discuss the achieved results.

Question:

Please add additional paragraph with limitations and strengths of the study.

Answer:

We added this part in section discussion: “It is worth mentioning that our work is not an exhaustive study on the SARS-CoV-2 outbreak in Italy because we neglected many possible causes such as the different way of dealing with the pandemic by the Italian local health system. It is very difficult to choose a group of factors that can explain such a complex problem as the spread of Covid-19 in Italy in an exhaustive way. Also our model is also an approximation of the problem. Then even the regional scale is too broad to understand the precise dynamics of the territory, but the availability of the data did not allow us to choose a finer geographical scale.

The use of Machine Learning can weaken these limitations thanks to its ability to extract salient information from the data without making prior assumptions. Furthermore, the use of Covid-19 severity data available up to November 2020 avoided inserting an additional bias in our model due to the start of the vaccination campaign. Our model and the obtained results could be strategic to better understand way of diffusion and physiology of this disease.”

Conclusions

Question:

On page 15, there are no separate section of conclusions from the conducted study, which should be response to the purpose of the study – please, correct.

Answer:

DONE

General comments:

Question:

Some sentences are repeating in the manuscript. Please avoid repeating.

Lines 39-40: The comparison between the predictions of a simple linear model and two machine learning algorithms makes our results more robust and reliable.

Lines 103-104: The use of three different kinds of models makes the results of our analysis more robust.

It is recommended to write about it in the one place. In opinion of the reviewer, you can write on it in the paragraph on the strengths of the study.

Answer:

DONE

Question:

Please, highlight the changes to the revised version using a different colour and/or a different script.

Answer:

DONE

Reviewer 3 Report

The paper reports a comparison of three models for predictive investigation of COVID-19 positive and morality rates in Italy. Data of three groups of indicators are collected and their correlation are evaluated. Results are reported which are consistent with published works cited in the paper.

However, there are no explanations about why the source data collected are at different time ranges. Also, why are some other indicators missing, for example, medical conditions and economic status.

The analysis and comparison are conducted on more than twenty sub-regions in Italy. However geographic heterogeneity or geospatial regression are not investigated and presented with more details.

Author Response

Question:

However, there are no explanations about why the source data collected are at different time ranges.

Answer:

We thank the reviewer for this point. The different time intervals of the work were decided based on the available datasets, but we believe that the values of the dependent features we have chosen are quite constant over a wide time interval. We added a sentence in section Materials and methods to clarify this point.

Question:

Also, why are some other indicators missing, for example, medical conditions and economic status.

Answer:

We realize that the spread of covid is a complex problem to model. When we chose our variables we made a conscious approximation as we wrote in the introduction section.

Question:

The analysis and comparison are conducted on more than twenty sub-regions in Italy. However geographic heterogeneity or geospatial regression are not investigated and presented with more details.

Answer:

The data we used are only available at the regional level and we believe that a more detailed geographic or geospatial study needs a deeper level than the regional, at least provincial, level.

Reviewer 4 Report

In this article, the authors study the impact of several factors (e.g. respiratory /allergic diseases, flue vacinated persons, sedentary persons, alcohol consumers) on the mortality and positivity related to covid, focusing on Italian regions from feb to nov 2020.

The article includes a fair and quite focused state of the art (introduction and conclusion) that describes the works that have tried to establish the presence of factors impacting the propagation of the virus, which they themselves study in their works.

The authors then define precisely the terms used in the study, the techniques/models used to analyze the impact of the factors.

The authors propose to use the generalised linear model (GLM) to manage the possible dependence (not excluded and not established by the authors but difficult to establish from the dataset) between variables. The linear character having been proved during the evaluation. In addition to a simple linear model, traditional ML algorithms.

In addition, a convincing evaluation/analysis is provided and the results are explained/analysed in the conclusion.

Overall, the paper is quite easy to follow and provides an interesting analysis eventought the impact of these factors have been already independently evaluated.

Minor comment: sedentary behaviours and passengers in airports do not permit to establish the micro- and macro-mobility of persons in Italy or elsewhere. In the conclusion, I would avoid mentioning that mobility does not have an impact: many studies using GPS or publicly available datasets (e.g. the ones provided by social networks and goole) establish an impact..

Relevant references :

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8657163/

Author Response

The authors then define precisely the terms used in the study, the techniques/models used to analyze the impact of the factors.

In addition, a convincing evaluation/analysis is provided and the results are explained/analysed in the conclusion.

Overall, the paper is quite easy to follow and provides an interesting analysis eventought the impact of these factors have been already independently evaluated.

Question:

Relevant references :

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8657163/

Answer:

We deleted the following sentence from section Discussion: "An interesting aspect to highlight is that the mobility variable seems to have a very small influence for mortality and positivity rates of the SARS-CoV-2 outbreak in Italy, but this aspect could be explored in a future work".

Round 2

Reviewer 2 Report

Dear Authors,

I have read your article again and have further comments. Your new corrections will be important on the decision whether the manuscript could be considered for further proceeding and publication:

1.In the Introduction section, a better rationale for the aim of the work is needed, which should take into account the choice of 21 Italian regions for the analysis, and as well their North-South diversity.

2.The aim of the work is still not clear and appears in several places, for instance, in lines: 2-5; 36-37; 83-84; 89-91. After specifying one goal, insert it: in Abstract and above the Material and Methods section. Please, take in mind, that the aim in these two places should be identical. It should be remembered that the conclusions of the study should correspond to the aim of the study.

3.At the beginning of the Material and Methods section, you need a separate paragraph “Study design and settings” containing information about the research design and its type, and then paragraph “Study area” which should contain information as well about 21 Italian regions.

4.Figure 2 is mentioned in the paragraph the lines 104-109, therefore it should be inserted after the line 109.

5.Figures 4, 5 and 6 show the results and therefore should be included in the Results section (Figure 4 after the line 215; Figure 5 after the line 225; and Figure 6 with its description after the Figure 5).

6.In my previous comment No. 11, I recommended that the sentences in lines 180-209 would be moved to the Results section. These sentences are currently still in the Discussion section (now lines 235-259) and should be moved under the Table 2.

7.The conclusions presented in the abstract are not compatible to the conclusions in the main text. Moreover, according to the reviewer, there is a contradiction between the content in lines 12-13 and 318-320. Please correct it.

8.The present discussion is not comprehensive and requires an extension containing interpretations of the results concerning three groups of variables, such as incidence of respiratory diseases, life style and mobility.

9.The Discussion should contain the clearly separated paragraphs on the limitations and strengths of the study.

For the correcting their manuscript it may be helpful for authors to have a look at the similar articles published in the International Journal of Environmental Research and Public Health, for instance,

doi:10.3390/ijerph191710884; doi: 10.3390/ijerph19127051.

Please, highlight the changes to the revised version using a different colour and/or a different script.

Author Response

Question:

Answer:

We added some sentences in introduction section.

Question:

Answer

We clarified and uniformed the aim of our work in the different sections of the manuscript.

Question:

Answer

We modified the Material and Methods section as suggested by reviewer.

Question:

4.Figure 2 is mentioned in the paragraph the lines 104-109, therefore it should be inserted after the line 109.

Answer

We use latex as text editor and for this reason we can’t choose the precise point in which inserting the figure, because latex select automatically the best position of figure in the text. But we have moved Figure 2 to the closest point possible to that indicated by the reviewer.

Question:

Answer

See the previous answer.

Question:

Answer

We modified the Discussion section as suggested by reviewer.

Question:

Answer

We corrected the Conclusion section as suggested by reviewer.

Question:

Answer

We modified our Discussion as suggest by the reviewer.

Question:

9.The Discussion should contain the clearly separated paragraphs on the limitations and strengths of the study.

Answer

Done

Reviewer 3 Report

The authors have made great efforts to improve the paper. Nicely done.

As reported by medical researchers, SARS-CoV-2 has been circulating in some areas before the first case reported by the Chinese in December 2019. That is the reason that papers tend to make arguments like "The causes of Covid-19 spread in Italy in the first wave of the pandemic are difficult to explain". I suggest the authors may need review more available works, for example this one by Italy colleagues:

Giovanni Apolone etc, 2020, Unexpected detection of SARS-CoV-2 antibodies in the prepandemic period in Italy, Tumori, 107(5), 446-451， https://doi.org/10.1177/0300891620974755

Paragraphs between line 121-139 can converted to a table for easy reading.
Additional language editing may be needed.

Author Response

Question:

As reported by medical researchers, SARS-CoV-2 has been circulating in some areas before the first case reported by the Chinese in December 2019. That is the reason that papers tend to make arguments like "The causes of Covid-19 spread in Italy in the first wave of the pandemic are difficult to explain". I suggest the authors may need review more available works, for example this one by Italy colleagues:

Giovanni Apolone etc, 2020, Unexpected detection of SARS-CoV-2 antibodies in the prepandemic period in Italy, Tumori, 107(5), 446-451， https://doi.org/10.1177/0300891620974755

Answer:

We modified Introduction and Conclusion sections taking into account the suggestion of the reviewer.

Question:

Paragraphs between line 121-139 can converted to a table for easy reading.

Answer:

Done

Question:

Additional language editing may be needed.

Answer:

Done

Round 3

Reviewer 2 Report

I believe that the quality of manuscript improved after revision. Based on this observation, I recommend "Accept".