Document is current

Current Referee Status

Referee status: Not Approved, Approved with reservations, Approved with reservations

Article Reports

Referee Report: https://doi.org/10.5256/f1000research.58239.r124239, Sam Moore, Zeeman Institute for Systems Biology and Infectious, University of Warwick, Coventry, 11 Mar 2022, version 1, 1 approved with reservations, 1 not approved

Referee Comment: <b>Glenn Young</b>; <i>Posted: 04 Oct 2022</i>; We thank the reviewer for his useful suggestions for improving our paper. We have addressed all comments&nbsp; below. 1. The model firstly assumes a homogeneous population. While all modelling necessitates a certain degree of approximation, a detailed spatial structure may be intractable and of limited benefit, some stratification by age and/or vulnerability is arguably essential in the study of COVID-19 dynamics given the highly uneven level of risk, especially given results in the paper are principally presented in terms of severe disease outcomes. Due to the highly contagious nature of the disease and damaging socio-economic consequences of prolonged NPIs, most countries around the world have aimed to use vaccination principally to reduce the worst disease effects by targeting the vulnerable rather than aiming for epidemic turnover as concentrated on in this study. Thank you for your feedback. We understand that considering a homogeneous population limits our model’s predictive scope, as we mention in our manuscript’s discussion. However, the stated goal of the work in this manuscript is to qualitatively understand how pharmaceutical and non-pharmaceutical interventions influence the spread of COVID-19, particularly by focusing on the incidence of severe cases requiring intensive care. These theoretical results are most effectively found through the rigorous analytical methods we employ here, methods that would not be available to us with an age-structured model. Our assumption of a homogeneous population is therefore appropriate because homogeneity simplifies the required analysis and parameterization, and our results are robust to parameter perturbations. We have added text to the discussion elaborating on the necessity of our model’s simplicity. 2. The vaccine dynamics included in the model are also overly simplistic. While in the introduction the authors do make some discussion about the multiple benefits of vaccination, in terms of reducing severe disease outcomes and symptom prevalence as well as infection and transmission, their model makes no attempt to incorporate this. Many of the vaccines used have proved highly effective in reducing severe disease outcomes (as much as 98% effective in preventing mortality) which would have significant consequences for the projected hospital/ICU numbers presented. We agree that many of the vaccines have proven highly effective in reducing severe disease outcomes. In fact, when we submitted this manuscript, initial data on the widely administered MRNA vaccines (Pfizer, Moderna) reduced all COVID-19 cases by ∼ 95%, not just severe [1]. This is reflected in our assumption that vaccinated individuals’ risk of infection is reduced by a factor of ξ = 0.94. We further emphasize that this assumption, that vaccination reduces only the risk of infection and does not subsequently reduce the risk of severe outcomes, was based on data available at the time, and that limitations on available data would have made any assumption on the vaccine’s downstream effects impossible to parameterize in any meaningful way. &nbsp; Further: in our model, hospitalization is being driven primarily by the unvaccinated, and consequently differentiating the rates at which vaccinated versus unvaccinated individuals develop severe symptoms has no qualitative effect on our results. We include here a visualization of ICU cases with total ICU cases (solid curves) and ICU cases among unvaccinated individuals (dotted curves) visualized separately. For both zero (green curves) and high (blue curves) NPI adherence, the unvaccinated population comprises over 97% of ICU cases at their peak. These results suggest that even if we assume that vaccinated individuals cannot develop severe symptoms, our results remain qualitatively unchanged. We have added text and Figure 5 in the manuscript clarifying this important point. 3. Results throughout the study are framed with a view to disease eradication, assumingly achieved by reducing the effective reproduction below 1. There are several issues with this measure in the context however; firstly, waning immunity and rapidly emerging viral variants both mean that the herd immunity threshold is by no means fixed, and secondly the study frames presents the threshold as a function of a fixed NPI level, while in reality no country is likely to leave control measures in place in perpetuity and these should rather be explored as a temporary dynamic and not to effect a final threshold. We again ask the reviewer to understand that this work was submitted in August 2021, shortly before evidence of waning immunity began to emerge [2] (note that the cited article was first published on medRxiv on Au- gust 31, 2021). Before this evidence was released, many considered herd immunity through a combination of vaccination and infection to be an achievable goal [3]. Moreover, while we do present conditions such that the effective reproductive number is reduced below unity, we also consider the impact of subsequent outbreaks in on communities with low vaccination rates and relaxed social measures in Figure 6. This consideration was aimed to address the reality that the virus might never be eradicated. We have added text regarding waning immunity to the discussion. &nbsp; However, we do agree that terms like “eradication” overstate what was ever considered feasible. We have modified the language throughout the manuscript to focus on control, rather than eradication. 4. Finally, the model is fitted to provide a single a static set of parameters. However, a broad parameter set can clearly match the same data. Due to the sensitivity of the results to the fitting, it would be greatly beneficial to see this explored to some extent, to provide some sort of prediction intervals. Due to significant changes in both viral characteristics and available treatment over time, many of these parameters are also not constant as presented for instance one assumes the parameters are chosen to match the characteristics of variants in circulation at one particular time, but this is a rapidly changing situation. The control parameter presented is also entirely abstract, and one feels it would be greatly more informative if it had some grounding in reality what range has actually been observed in practice? &nbsp;Thank you for the comment; we agree that many of the parameters in our model can and will change over time. As stated above, the goal of our work is to assess the influence of intervention strategies on the incidence of severe covid cases, and while changing parameters affects the quantitative results of our model, the qualitative results remain unchanged. We illustrate this by the addition of two new figures, both in Fig 3 in the manuscript. Since we do not know for certain how parameters will change over time, we allow every parameter in our model except φ (vaccine coverage), ξ (vaccine efficacy), μ (vaccination rate), and a (proportion of population adhering to NPIs) to randomly vary by up to 20% above or below its baseline value. We then simulate our model and collect data to make the figures, as described below. &nbsp; The left figure shows the region in which 80% of simulations fall as we vary parameters as described above over 5000 iterations. The solid curve shows the ICU caseload over time with φ = 0.6 and a = 0 (the same curve shown in black in Figure 4 in our manuscript) and 80% of simulations fall within the two dashed curves. &nbsp; The right figure shows peak ICU caseloads over φ. The solid middle curve shows the peak ICU caseload for our baseline parameters with φ = 0.6 and a = 0 (same curve shown in red in Figure 7 in our manuscript). We then vary φ from 0 and 1 in 0.2-unit steps and simulate our model 5000 times for each value of φ with parameters randomly selected as described above. 80% of peak ICU caseloads fall within the dash-dotted curves above and below the central curve. &nbsp; These results suggest that our model’s results are qualitatively robust; that is, we expect disease eradication or control to require a combination of vaccination and NPI strategies even as these characteristic parameters change over time. [1] &nbsp;Polack, Fernando P., et al. “Safety and efficacy of the BNT162b2 mRNA Covid-19 vaccine.” New England Journal of Medicine (2020). [2] &nbsp;Goldberg, Yair, et al. “Waning immunity after the BNT162b2 vaccine in Israel.” New England Journal of Medicine 385.24 (2021): e85. [3] &nbsp;Randolph, Haley E., and Luis B. Barreiro. “Herd immunity: understanding COVID-19.” Immunity 52.5 (2020): 737- 741.

Referee Report: https://doi.org/10.5256/f1000research.58239.r143358, Matthew Adewole, Department of Computer Science and Mathematics, Mountain Top University, Prayer City, 30 Aug 2022, version 1, 1 approved with reservations, 1 not approved

Referee Comment: <b>Glenn Young</b>; <i>Posted: 04 Oct 2022</i>; We thank the reviewer for his useful suggestions for improving our paper. We have addressed all comments below. 1. Authors assumed a closed population where there is no demographic data. This kind of scenario can only be true when the time-duration is small enough. However, Figure 2 shows a time series plot over a period of about 400 days. Saying there is no demographic process within this period is far from reality. I suggest that the authors include demographic data. While Figure 3 displays data out through 365 days, the most important dynamics (namely time to peak ICU caseload) occur within 100 days of outbreak. Similarly, the caseload reduces to 20% of its maximum within about half a year. These are a short enough timespans relative to the dynamics of the model that the demographics within the population should not change enough to influence the course of the outbreak. For this reason we do not suspect that the addition of demographic processes will influence our results in any qualitative way. 2. It was said that S represents “susceptible without access to vaccine”. How come individuals in S compartments are vaccinated at a rate μ? I think something is wrong with the nomenclature. The S population is the proportion of individuals who have not yet had the opportunity to receive a vaccination. We agree that our description within the manuscript is unclear and confusing. We have changed the description of S from “susceptible without access to vaccine” to “susceptible population prior to vaccine access. 3. The first term on the right side of Equation 1c should be μφS. •\ You are correct. Thanks for catching this mistake. 4. A significant part of this work is the parameter estimation. Authors should make this more explanatory. Authors should display graphs showing data and fitted curves. Authors should differentiate between estimated parameter values and the parameter values taken from literature. The primary goal of this work is to investigate the tradeoffs between pharmaceutical and non-pharmaceutical interventions for controlling the spread of COVID, not on parameter estimation. The parameter estimation work was done in another paper by co-authors Newcomb and Michael [1], and we consequently leave the details of the method out of the present manuscript. However, we agree that additional detail regarding the method could benefit readers, and so we have added a new figure (now Figure 2) showing a fitted curve of COVID case data and include a slightly longer, though still brief, description of the method used to fit parameters. 5. The data used for parameter estimation are quite outdated. Therefore the estimated parameter values and, consequently, the results of this work may not represent the present-day reality. This is true and is of course a limitation of any covid modeling work that is not continuously updated with new data. However, the goal of this work is not to inform real-time decisions, but to understand the qualitative effects of coupled pharmaceutical and non-pharmaceutical interventions on the spread of the virus. In an effort to understand how robust our results are to changes in parameters, we have included a sensitivity analysis in our revised manuscript. Please see our response to Reviewer #1 comment 4. [1] &nbsp;Newcomb, Ken, et al. “Iterative data-driven forecasting of the transmission and management of SARS-CoV-2/COVID- 19 using social interventions at the county-level.” Scientific Reports 12, no. 1 (2022): 1-19.

Referee Report: https://doi.org/10.5256/f1000research.139468.r244831, Sangeeta Saha, MS2 Discovery Interdisciplinary Research Institute, Wilfrid Laurier University (Ringgold ID: 8431), Waterloo, Ontario, 07 Mar 2024, version 2, 2 approved with reservations, 1 not approved

Grant Information: The author(s) declared that no grants were involved in supporting this work

Copyright: This is an open access article distributed under the terms of the Creative Commons Attribution Licence, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.