COVID-19 Trends in a Northeastern Brazilian State from the Start of the Pandemic: Exploring an Alternative Time Series Method

Round 1

Reviewer 1 Report

This was an interesting paper, unfortunately Eq 2 is missing (and appears to have been replaced with a second instance of Eq 1).  That's too bad because Eq 2 seems like it would be the punch line of the paper.  Overall, the paper seems solid, but it is hard to say for sure because Eq 2 was not available for review.

One other thing I'll note, the authors make a few statements about government failure and the "right" way to mitigate, etc.  These statements really don't further the technical contribution and narrative of the paper.  While I don't take issue with them per se, they do feel out of place.

Line 114: equation 1 should be replaced with equation 2.

Line 161 - 162: the statement about "Rt with less lack of information" is confusing.  Maybe another sentence or two about what that means would help...?

I've highlighted a few things in the attached PDF file (a couple of language issues, the equation, etc.).

Comments for author File: Comments.pdf

Author Response

Dear Reviewer 1,

Thank you for your comments and careful review of our manuscript. We appreciate you taking the time to provide feedback.

We sincerely apologize for the error in Equation 2.  We corrected this mistake in the revised manuscript, ensuring Equation 2 is now properly presented.

We also acknowledge your point regarding the statements about government actions and the "right way" to mitigate the pandemic. We understand these may seem out of place and detract from the technical focus of the paper. In the revised version, we have removed the statement.

Furthermore, we have addressed the issues you highlighted in the PDF, including:

• Line 114: We have replaced the duplicate Equation 1 with the correct Equation 2.
• Line 161-162: We have rephrased the confusing statement about "Rt with less lack of information" to provide greater clarity.
• Language issues: We have carefully reviewed the entire manuscript and corrected any language issues you identified to ensure clarity and accuracy.

We believe these revisions have strengthened the manuscript and addressed your concerns. We are grateful for your feedback, which has been improving the quality of our work. We hope that the revised version now meets your expectations and reflects the contributions of our research.

Sincerely,

Matheus Paiva

Reviewer 2 Report

see the file

see the file

Comments for author File: Comments.pdf

Author Response

Dear Reviewer 2,

 

We are grateful that our manuscript, titled “From the Beginning, trends in COVID-19 Disease in a Northeastern Brazilian State: A New Time Series Analysis Method,” is being considered for publication.

 

Enclosed, you will find our detailed, point-by-point responses to the reviewer comments. We have carefully addressed each concern and made the necessary revisions to improve the manuscript. We appreciate the reviewers’ feedback, which has helped us enhance the quality of our work.

 

Sincerely,

Matheus Paiva

Corresponding Author

 

Overall Comment

The paper covers interesting data spanning two years of COVID-19 development in Pernambuco, Brazil. It promises “a new time series analysis method,” but from my standpoint, it does not deliver. I believe the paper could be worth publishing if major revisions are made.

 

Response:

We appreciate your feedback regarding the argument of a "novel time series analysis methodology". We recognize the need to elucidate its originality and execution more effectively, and we have amended the manuscript to thoroughly address these issues. With equal attention, we revised and enhanced the quality of the English language used. Therefore, we modified the title to better reflect the actual content of the manuscript: “From the Beginning, Trends in COVID-19 Disease in a Northeastern Brazilian State: Exploring an Alternative Time Series Method.”

 

Comment 1

What is the new time series analysis method proposed?

The authors need to explain the methodological novelty. Time-series analysis often uses Box-Jenkins ARIMA models, exponential smoothing, trend/seasonality analysis, or other techniques for modeling and forecasting. The methods in the manuscript are unclear and need to be explicitly stated.

 

Response:

We appreciate your question and understand the importance of clearly explaining the novelty of the method. Unlike traditional approaches, such as ARIMA, our method focuses on identifying stable periods when dealing with data gaps. This approach was designed to be more practical in challenging contexts, such as the pandemic, where data often present inconsistencies. It particularly manages missing data and provides a framework to determine stable temporal regions. Our goal is not to substitute traditional models but to complement them by offering a simpler way to segment and interpret critical periods, such as those observed in Pernambuco.

Comment 2

What is the focus of the study: the reproductive number (Rₜ), efficiency, or the trio of incidence, mortality, and case fatality? How are these related?

 

Response:

The study focuses on the indicators of incidence, mortality, and case fatality rates to understand disease transmissibility and segment the pandemic into temporal regions. These indicators are essential to understand the changes in the transmission and severity of the disease. Rt comes in as a complementary tool to reinforce the identification of patterns. Thus, the three components work together to provide a completer and more reliable picture.  This mixture permitted us to capture trends in illness behavior over time, identifying duration of stability and important changes.

 

Comment 3

To which of these time series is the Prais-Winsten regression applied? Where is the evidence of serial correlation (e.g., AR(1) type), which normally suggests the use of the Prais-Winsten method?

 

Response:

 

Thank you very much for your questions and for the opportunity to further clarify the novelty of our method for calculating the effective reproduction number (Rt). Traditionally, the calculation of Rt relies on probabilistic distributions, such as Weibull or gamma, to model the generation times of the disease. These approaches, although widely used, have some limitations, especially in contexts where the available data do not fit well with these distributions or where there are significant gaps.

Our method proposes a simpler and more practical alternative. Instead of simulating probability distributions, we directly use the observed daily incidence data, adjusting them through a robust regression approach. This allows us to deal with problems such as missing data or inconsistencies in a more direct way, without relying on complex statistical assumptions. With this technique, we are able to estimate Rt efficiently and reliably, especially in scenarios where resources or data quality are limited.

We believe that this approach not only simplifies the calculation of Rt, but also makes it more accessible and robust for different epidemiological realities, such as the one we faced in Pernambuco during the pandemic.

 

 

Comment 4

The time series plots provided depict the data only and do not suggest any modeling. What are the inference results regarding time-series models?

 

Response:

We understand your concern and thank you for highlighting this point. The study predominantly concentrates on a clear descriptive analysis, utilizing time series plots to visualize trends, and identify important patterns, such as peaks and periods of stability. These plots are not intended for predictive modeling but to provide insights that can guide public health decisions. While formal statistical inference methods (e.g., hypothesis testing or confidence intervals) were not applied, the observed trends describe patterns and highlight critical periods of change. Although we did not focus on predictive modeling in this work, our approach allows future studies to advance with more detailed inferential analyses.

 

Comment 5

The only inferential results involve t-tests comparing pairwise daily percentage changes across three segments (A, B, and C), treating them as independent samples. This approach ignores the finer structure of the time series and is not novel.

 

Response:

The t-tests were employed to facilitate direct comparisons of trends amongst the three segments (A, B, and C), emphasizing overarching temporal patterns. We recognize the constraint of not examining the more intricate structures within the time series and have elucidated this matter in the amended manuscript.

Comment 6

How are the boundaries between A, B, and C determined? There i  no mention of changepoint analysis (e.g., Zivot-Andrews or other methods), and the choice seems arbitrary.

 

Response:

The boundaries were defined by observing substantial changes in efficiency trends, as illustrated in Figure 2. Specifically, periods were defined when efficiency values approached 100%, ensuring minimal, impact of missing data on the calculations, indicating more stable periods. Although we did not use formal methods of change point analysis, we detailed in the manuscript how the choices were based on practical epidemiological criteria.

 

Comment 7

The observations of differences in behavior or periods A, B, and C are not fully justified.

 

Response:

Thank you for this observation. We explain how each period was characterized. Period A presented high mortality and lethality, while period B was marked by high incidence with stabilization in mortality. Period C, in turn, showed a consistent decrease in all indicators. We relate these differences to the interventions implemented and to the changes in the epidemiological scenario, which is detailed in the manuscript.

 

Comment 8

There are multiple typos and problems with definitions. For instance, there is no formula (2), and formula (1) is repeated twice. Also, the definition of Rₜ seems unclear.

 

Response:

Thank you for pointing out these issues. We have reviewed the entire manuscript, corrected typos and adjusted the formulas to ensure they are clear and accurate. Table 1 was corrected.  The definition of Rₜ has been revised for clarity, consistency and to make it more understandable and consistent with the methodology presented, addressing the concerns raised.

 

Comment 9

After addressing the abovementioned concerns, the paper may be ready for publication.

 

Response:

Thanks for the valuable feedback. We hope that our responses and revisions address all concerns and bring the manuscript to the required standard for publication.

 

Round 2

Reviewer 2 Report

The authors have addressed my comments. The paper seems to contain mostly qualitative analysis of time series which is adequately described by the changed title.

Comments from round 1 are addressed.