Location Scheme of Routine Nucleic Acid Testing Sites Based on Location-Allocation Models: A Case Study of Shenzhen City
Shanshan Han
Round 1
Reviewer 1 Report
The manuscript combines two Location-Allocation models to solve NAT siting problems. It can not only realize the determination of numbers and distribution of NAT services but also choose the optimal candidate plans. Although China’s present COVID-19 prevention and control policy has been relaxed, it is still worth researching for future public and health service facilities siting. However, there are still some problems that need to be improved or explained clearly:
1. In Section 4.1.5, the comparison between the actual NAT siting deployment and the manuscript’s solution is made in two subdistricts. It mentions that the population coverage of your solution is larger than the actual deployment. However, I notice in both examples that your solution has more NATs than it actually does. To give an extreme example, as long as we set enough NATs without considering the cost, we can achieve high coverage without setting any location-allocation function. Therefore, I’m not sure if such an explanation is still persuasive when the NAT number of your solution is larger than that of the actual one.
2. I think the siting of NAT may not only consider the population but also other attributes of the population, like age, mobility, etc. For instance, a certain residential area is not densely populated but mostly populated by older people with low mobility. Such an area may need a NAT that has a comparatively low travel cost, while the siting solution may not fully consider it when only considering population.
3. Such small mistake like reference source not found error in Word (e.g. in Line 325). Please check the whole manuscript carefully.
Author Response
Thanks for the valuable comments, which are very helpful to improve our work. We have written a point-by-point response and hope that the revision will address your concerns.
Comment 1:
In Section 4.1.5, the comparison between the actual NAT siting deployment and the manuscript’s solution is made in two subdistricts. It mentions that the population coverage of your solution is larger than the actual deployment. However, I notice in both examples that your solution has more NATs than it actually does. To give an extreme example, as long as we set enough NATs without considering the cost, we can achieve high coverage without setting any location-allocation function. Therefore, I’m not sure if such an explanation is still persuasive when the NAT number of your solution is larger than that of the actual one.
Response 1：
Thanks for the suggestions. To eliminate potential ambiguity, we added “people’s average travel cost to the nearest site” as an indicator to compare the performance of different siting solutions. In the Guangming subdistrict, our scheme has the same number of NAT sites as the actual scheme, so we can directly compare the population coverage and average time. However, in the Pingshan subdistrict, the number of NAT sites differs between our scheme and the actual scheme. To control the effect of NAT number, we added a simulation experiment in which we randomly sampled the same number (i.e. NAT number in the actual one) of NAT sites from our solution multiple times and calculated the average coverage and travel cost for the comparison. Table 4 compares the performance of the actual scheme and our scheme in the two selected subdistricts. With the same number of NAT sites, our scheme achieved a significantly higher population coverage and lower average time in both Pingshan and Guangming subdistricts, which demonstrates the superiority of our methods. Please find the details from line 414 to 426.
Comment 2:
I think the siting of NAT may not only consider the population but also other attributes of the population, like age, mobility, etc. For instance, a certain residential area is not densely populated but mostly populated by older people with low mobility. Such an area may need a NAT that has a comparatively low travel cost, while the siting solution may not fully consider it when only considering population.
Response 2：
Thanks for the suggestion. It is true that different populations would have different impedances to accessing the NAT site, this is also a significant issue in recent studies on social equity. Unfortunately, in this instant study, we cannot distinguish different populations as our data do not contain personal information. However, once personal information is available, it can be easily implemented into our framework by increasing the weight of the demand point where people with lower mobility live. We have added some discussion about this issue in the “Conclusions” section.
Please find the details from line 467 to 474.
Comment 3:
Such small mistake like reference source not found error in Word (e.g. in Line 325). Please check the whole manuscript carefully.
Response 3：
Sorry for the inconvenience caused by our mistakes. We have made thorough proofreading to correct all such mistakes.
Reviewer 2 Report
This paper presents a research study to adapt classical location models and a bi-objective optimization model to build a data-driven framework for designing the 15-NAT sites layer. It optimizes the number of locations of NAT sites and dynamically adapts to different window times of the day, balancing various cost constraints.
The paper is interesting; however, some aspects must be improved.
It is necessary to justify this research work and the importance of the analysis. It could be described in the “Introduction Section” in more detail.
What is the main difference between other models of the state-of-the-art?
Please, in the “Methods Section,” it is crucial to describe in more detail the the applied and implemented mecorrect geoprocessing.
Correct the reference error on page 12.
Please, incorporate the section “Literature review or Related work,” it is necessary to report research studies concerning this particular topic and their influence and effectiveness in this field. Moreover, at the end of the section, authors must compare their proposal and the reported works in the literature review, emphasizing the differences and advantages.
Please, separate the “Discussion” and “Conclusion” into different sections and extend the interpretation of the analysis of the results in the “Discussion section”. Conclusions should emphasize the contributions and future work.
Author Response
Thanks for the kind comments. Based on your suggestions, we have made a major revision, especially on the comparison with state-of-the-art methods. Our responses are listed as follows:
Comment 1:
It is necessary to justify this research work and the importance of the analysis. It could be described in the “Introduction Section” in more detail.
Response 1：
Although the impact of COVID-19 is currently declining, its rapid mutations and repeated outbreaks on a global scale make it still an important threat to the global public health system. Studying the deployment scheme of public NAT services is not only an effective preventive measure for future risks of COVID-19, but also provides a reference for preventing other infectious diseases and establishing rapid public health responses.
Please find the details from line 144 to 148.
Comment 2:
What is the main difference between other models of the state-of-the-art?
Response 2：
We added some reviews about the state-of-the-art works in “Introduction”. Existing methods could be classified into two categories. The first category could be termed as ‘scoring-based’, in which a list of factors that are related to the alternative site needs to be evaluated, and the scores will be further analyzed by methods like AHP to get a synthesis score of each alternative. By ranking these scores, one can easily select the highest one as the best location. However, the scoring-based method will suffer from subjectivity issues during the selection of factors. Another batch of studies usually adopted location-allocation algorithms to objectively find the optimal locations for new facilities under different kinds of constraints. However, although existing method can solve the optimizaiton problem when either the number of new facilities or alternative locations are given, they cannot deal with the situation when these two variables are unknown at the same time.
In contrast, our method could solve the optimization problem when neither the number of new facilities nor their locations are not determined.
Please find the related contents from line 50 to 93, and line 450 to 460.
Comment 3:
Please, in the “Methods Section,” it is crucial to describe in more detail the applied and implemented mecorrect geoprocessing.
Response 3：
The geoprocessing involved in this study is mainly implemented in ArcGIS (both the LSCP and P-median models were calculated using the location assignment tool in the Network Analysis module of ArcGIS 10.7) and the finding of knee points is implemented in Python. Please find the details from line 278 to 279, and line 290 to 292.
Comment 4:
Correct the reference error on page 12.
Response 4：
Thank you for your help. We have corrected the error.
Comment 5:
Please, incorporate the section “Literature review or Related work,” it is necessary to report research studies concerning this particular topic and their influence and effectiveness in this field. Moreover, at the end of the section, authors must compare their proposal and the reported works in the literature review, emphasizing the differences and advantages.
Response 5：
Thanks for the suggestions. First we added some reviews about the state-of-the-art methods in “Introduction” as we mentioned in the response to Comment 2. Further, we added some justification for this study at the end of “Introduction” from line 144 to 148. We also emphasized the pros and cons of this study in “Conclusion” from line 451 to 474.
In brief, the current siting method cannot directly solve the optimization problem when neither the number nor the distribution of new facilities is known at the same time. Even though some scoring-based methods might be applicable to deal with that problem, they require prior knowledge to design the evaluation factors which are easily affected by the subjective issue. Our method is superior in producing the siting solution without knowing the number and distribution of new facilities, which would be suitable in the case of planning NAT sites.
The disadvantages of our method include: 1) redundant sites might be assigned near the district boundary; 2) only two cost factors are considered, while more factors could be included; 3) the mobility difference in population is ignored. These disadvantages are discussed at the end of this article.
Comment 6:
Please, separate the “Discussion” and “Conclusion” into different sections and extend the interpretation of the analysis of the results in the “Discussion section”. Conclusions should emphasize the contributions and future work.
Response 6：
We removed the discussion section because we think enough discussions about the experimental results have been made in the “Result” section. Also, we added some content in the “Conclusion” to emphasize the contributions and limitations of this work. Please find the details from line 443 to 474.
Reviewer 3 Report
page 1. keyword, MOOP => Multi-Objective Optimisation Problem (MOOP)
page 5. Figure 1, It is necessary to present an index map so that Shenzhen city can know where it is in China. And The content of Figure 1(b) overlaps with the legend of 1(a). Figure 1(b) is can be deleted.
- In Figure 3, the maps of working hours and non-working hours are so similar that it is difficult to see a difference in distribution between the two. If you want to show the difference between the two in your case study, you need to show the point of difference on the map as well.
- It is necessary to clearly present the means of transportation (walking or car) used in the model presented in the case study.
- In Figure 13, at the city scale, there seems to be no difference in the distribution of NAT between working and non-working hours.
-Traditional Location-Allocation models is generally case studies using cars as a means of transportation. Therefore, the research presented to the class seems to suggest walking or public transportation as a means of transportation, and additional explanations for these differences are needed.
Author Response
Thanks for the valuable comments, which help us a lot in improving the quality of this work. Our responses are listed as follows:
Comment 1:
page 1. keyword, MOOP => Multi-Objective Optimisation Problem (MOOP)
Response 1：
Thanks, we corrected that keyword following the suggestion.
Comment 2:
page 5. Figure 1, It is necessary to present an index map so that Shenzhen city can know where it is in China. And The content of Figure 1(b) overlaps with the legend of 1(a). Figure 1(b) is can be deleted.
Response 2：
Thanks for the suggestion. First, we removed Figure 1(b). Then, we redrew Figure 1 and added the index map and administrative subdistrict information to it. We also removed the original Figure 2 to avoid duplication of information.
Please see the revised Figure 1 in line 154.
Comment 3:
In Figure 3, the maps of working hours and non-working hours are so similar that it is difficult to see a difference in distribution between the two. If you want to show the difference between the two in your case study, you need to show the point of difference on the map as well.
Response 3：
To highlight the differences between the two time periods, we performed a kernel density analysis of the population data and replaced the original Figure 3 with a population density map. We hope that this modification meets the requirements. Please find the new figure in line 199.
Comment 4:
It is necessary to clearly present the means of transportation (walking or car) used in the model presented in the case study.
Response 4：
Thanks for the suggestion.
In this study, the distance range of [s_min,s_max] was determined by the average walking speed V and predefined time threshold.
Please see the detail from line 235 to 237.
Comment 5:
In Figure 13, at the city scale, there seems to be no difference in the distribution of NAT between working and non-working hours.
Response 5：
Thanks for the suggestion. To highlight the difference in the results between the two time periods, we replotted Figure 13 by integrating the siting results of three selected subdistricts into one figure (i.e. Figure 12) to enrich the visualization. We also removed the original Figure 14 to avoid duplication.
Please see the revised Figure 12 in line 376.
Comment 6:
Traditional Location-Allocation models is generally case studies using cars as a means of transportation. Therefore, the research presented to the class seems to suggest walking or public transportation as a means of transportation, and additional explanations for these differences are needed.
Response 6：
People who seek NAT testing are likely to have been infected, and they are usually encouraged to access nearby NAT sites by walking to reduce the risk of viral transmission. Therefore, we reasonably chose the walking mode and took 5-15 minutes as the time interval to evaluate the travel cost, which is also consistent with the "15-minute life circle" policy in China.
We add this justification in the revision, please find details from line 237 to 244.
Round 2
Reviewer 1 Report
The author has revised the manuscript point by point. I think that after revision the manuscript is qualified to be considered for publication in IJGI as a research article.
