Mediating Effect of the Stay-at-Home Order on the Association between Mobility, Weather, and COVID-19 Infection and Mortality in Indiana and Kentucky: March to May 2020
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsHere are some points to be addressed for the improvement of this article:
The study mentions high demographic, socioeconomic, weather, and air pollution heterogeneity across the geographical locations analyzed. Such variability can obscure the true effect of state-level health policies on COVID-19 incidence and make it difficult to control for all confounding factors.
The study notes that missing data for human mobility and weather indicators were not completely at random, as indicated by Little's MCAR test. This can introduce bias into the results and affect the validity of the conclusions.
Differences in the compliance and enforcement of stay-at-home orders and other interventions between Indiana and Kentucky might lead to differences in reported COVID-19 incidence and mortality that are not solely attributable to the interventions themselves.
The study period, from March 1, 2020, to May 15, 2020, is relatively short. A longer duration might be necessary to fully understand the long-term effects of social distancing and mask mandates on COVID-19 transmission and mortality.
Although the study focuses on the association between mobility, weather, and COVID-19 incidence, it does not analyze the impact of specific non-pharmaceutical interventions (NPIs) like social distancing and mask-wearing on transmission patterns.
Differences in reporting practices between states and counties could influence the accuracy of COVID-19 infection and mortality data. Such discrepancies can skew the results and conclusions drawn from the data.
The findings are specific to Indiana and Kentucky, which may limit the generalizability of the results to other states or regions with different demographic, socioeconomic, and environmental conditions.
While multiple imputation is a robust method for handling missing data, the imputation process itself can introduce biases if the assumptions underlying the imputation model are not fully met.
External factors such as state-level healthcare capacity, public health infrastructure, and public adherence to guidelines were not controlled for, which could significantly influence COVID-19 outcomes.
The use of aggregated mobility data from Google Community Mobility Reports might not capture the nuanced variations in individual mobility patterns and behaviors, potentially leading to an incomplete understanding of mobility's impact on COVID-19 spread.
Comments on the Quality of English LanguageModerate editing of English language required.
Author Response
Dear Reviewer # 1,
Thank you for your constructive feedback. We are providing the following responses (marked in red) for each of the following comments:
- The study mentions high demographic, socioeconomic, weather, and air pollution heterogeneity across the geographical locations analyzed. Such variability can obscure the true effect of state-level health policies on COVID-19 incidence and make it difficult to control for all confounding factors.
This comment is referring to lines 66-70 in the introduction section. Prior lines, 63-66, are highlighting studies that have investigated the association between human mobility and COVID-19 incidence rate. The statement relating to heterogeneity is being referred to geographical locations in those studies. We agree with your comment about the true effect of state-level health policies. This is the very same point we made in the statement, “The investigation included states in different phases of the COVID-19 pandemic surges. Statistical significance was not achieved, likely because of the high heterogeneity that was not adequately controlled.” (lines: 69-70). For better clarification, we have rephrased the sentences.
Original sentences (lines: 63-70):
“The association between mobility and COVID-19 incidence was analyzed in small populations as well as at highly aggregated levels such as regions and states in the U.S. [15-17] and found that state-level health policies (i.e., stay-at-home orders, mask mandates, curfews) were not associated with COVID-19 incidence [18]. The geographical locations analyzed had high demographic, socioeconomic, weather, and air pollution heterogeneity. The investigation included states in different phases of the COVID-19 pandemic surges. Statistical significance was not achieved, likely because of the high heterogeneity that was not adequately controlled.”
Revised sentences (lines: 63-66)
“Multiple studies have investigated the association between mobility and COVID-19 incidence [15-17]. The geographical locations analyzed in these studies had high demographic, socioeconomic, weather, and air pollution heterogeneity [15-17]. Such heterogeneity can mask the effect of COVID-19-related policies to curb the infection rate.”
- The study notes that missing data for human mobility and weather indicators were not completely at random, as indicated by Little's MCAR test. This can introduce bias into the results and affect the validity of the conclusions.
We believe that this is a result of joining different datasets. However, the missing value patterns chart shows that we do not see consistent patterns in the missing data for the variables except grocery and pharmacy. Since we conducted a multiple imputation using a linear regression technique, inconsistent patterns in missing data add validity to the imputation. But imputation of missing data is not flawless and may add bias to results. We have added the following in the limitation section:
“A key limitation is the use of multiple imputations to replace missing values. This particular method can cause biased results. However, the missing value patterns chart shows no consistent patterns in the missing data for the variables except for grocery/pharmacy. Since a multiple imputation using a linear regression technique was implemented, inconsistent patterns in missing data strengthened the validity of imputed data.” (Lines: 473-478).
- Differences in the compliance and enforcement of stay-at-home orders and other interventions between Indiana and Kentucky might lead to differences in reported COVID-19 incidence and mortality that are not solely attributable to the interventions themselves.
This is an excellent point, which we tried to address in the concluding paragraph in the discussion section. We added a bit more context on differences between the stay-at-home orders for Indiana and Kentucky, which clarifies this point better.
“Although findings on the state-associated risk of COVID-19 infection and mortality in relation to human mobility are contrary, it is important to consider that Indiana had a less stringent stay-at-home order compared to Kentucky [20]. Indiana residents were allowed to leave home for certain employment, outdoor activities, and to take care of others during March 24 to May 1, 2020. However, travel was restricted, and only life-sustaining businesses (retail, gas station, etc.) could remain open in Kentucky from March 25 to April 29, 2020. Furthermore, social distancing and hygiene were required, and the use of facial masks was promoted in public settings. The use of face masks in community settings is known to reduce the transmission risk of COVID-19 infection [54]. The difference in compliance and enforcement of stay-at-home orders between Indiana and Kentucky may help explain some of the variability in state-level differences in COVID-19 infection and mortality observed in the current study.” (Lines: 458 – 470).
- The study period, from March 1, 2020, to May 15, 2020, is relatively short. A longer duration might be necessary to fully understand the long-term effects of social distancing and mask mandates on COVID-19 transmission and mortality.
We considered March 1, 2020, to May 15, 2020, to capture the effect of social mobility and weather during the stay-at-home order. In both the states, initiatives were taken to phase out stay-at-home orders starting in May. This point is emphasized and explained in the introduction section (Lines: 68–94).
We added the following lines to limitations section, “Future studies may consider longer duration, segmenting analyses by different seasons for comparison, and the inclusion of socio-economic information to fully capture the long-term effects of social distancing and mask mandates on COVID-19 transmission and mortality.” (Lines: 484-487).
- Although the study focuses on the association between mobility, weather, and COVID-19 incidence, it does not analyze the impact of specific non-pharmaceutical interventions (NPIs) like social distancing and mask-wearing on transmission patterns.
- Differences in reporting practices between states and counties could influence the accuracy of COVID-19 infection and mortality data. Such discrepancies can skew the results and conclusions drawn from the data.
- The findings are specific to Indiana and Kentucky, which may limit the generalizability of the results to other states or regions with different demographic, socioeconomic, and environmental conditions.
- While multiple imputation is a robust method for handling missing data, the imputation process itself can introduce biases if the assumptions underlying the imputation model are not fully met.
- External factors such as state-level healthcare capacity, public health infrastructure, and public adherence to guidelines were not controlled for, which could significantly influence COVID-19 outcomes.
Comments 5-9 are all great points and we have added them to the limitations section.
We did not have data to capture a person’s behavior that could provide insight to non-pharmaceutical interventions (NPIs). We included this point in the limitations section based on your comment.
“This study has several limitations, notably that such an observational study cannot establish causality. Findings cannot be generalized to other U.S. states or regions. A key limitation is the use of multiple imputations to replace missing values. This particular method can cause biased results. However, the missing value patterns chart shows no consistent patterns in the missing data for the variables except for grocery/pharmacy. A multiple imputation using a linear regression technique was implemented; therefore, inconsistent patterns in missing data strengthened the validity of imputed data. The lack of information on other seasons (summer, fall, winter), non-pharmaceutical interventions (NPIs) (use of facial masks, social distancing compliance), and socio-economic factors (i.e., income and occupation) are underlying limitations to this study’s analytical models. Other limitations include possible discrepancies in reported COVID-19 infection and mortality and a lack of information on state-level healthcare capacity, public health infrastructure, and public adherence to guidelines. Future studies should consider longer duration, segmenting analyses by different seasons for comparison, and the inclusion of socio-economic information to fully capture the long-term effects of social distancing and mask mandates on COVID-19 transmission and mortality.” (Lines: 472-487).
- The use of aggregated mobility data from Google Community Mobility Reports might not capture the nuanced variations in individual mobility patterns and behaviors, potentially leading to an incomplete understanding of mobility's impact on COVID-19 spread.
This is true and we stated this in our discussion.
“The differences in findings between prior studies and the current study may be due to the limitations of human mobility data. Cellphone human mobility data was not able to capture an individual’s protective behaviors, such as mask usage and social distancing. Behavioral compliance may vary significantly across sub-populations and fluctuate dramatically over the course of the pandemic.” (Lines 354-359).
Reviewer 2 Report
Comments and Suggestions for Authors
Mediating Effect of Stay-At-Home Order on the Association between Mobility, 2 Weather, and COVID-19 Infection and Mortality in Indiana and Kentucky: March 3 to May, 2020 4
Shaminul H. Shakib (M.P.H.) 1, Bert B. Little (Ph.D.) 1,2, Seyed Karimi (Ph.D.) 1,2, W. Paul McKinney (M.D.) 3, 5 Michael Goldsby (Ph.D.) 1, Maiying Kong (Ph.D.)
Reviewer comments: (copied from paper with associated line)
This is an interesting topic with an appropriate statistical model applied as Cox regression model with covariates is good for analysis of time to event data (differently timed outcomes), censoring, and multiple covariates.
The writing is clear, cogent and effective. The premise, though not novel at this moment in time, does provide yet another important source of information concerning the restrictions that aid/or undermine efforts to protect the population, prevent overwhelm of tertiary care providers, and balance economic activity with agility in getting through what could (depending on the epidemic and the time) be a civilization crippling or even eliminating event. This is not hyperbole on my part, but a matter of historical record.
The methodology is interesting in regard to the number of data points (16,000+) and their sources pulled into the analysis.
Things to address grammatically or from the perspective of content:
120 For each day of the week and for each nation, baseline day median values were calculated [22]. For each nation? Do you mean State? Clarify and refine this.
130 . NOAA’s GHCNd database includes information from over 80,000 weather stations There is a period here that starts this sentence, a typo.
139-148 You used imputation to replace missing data. It is good that you report this and give the MCAR for it, great! Is there any indication as to why the mobility data reflected non-random missing data? Please mention this if there is. Also, I did not see the MVP for mobility in Supplemental Table 1 figure…I see it is included in table 2, supplemental 2 but that is not related to imputation, was it to difficult to display visually?
Not sure I understand the significance of Supplemental Table 3? Is it just to provide information on methodological approach?
Where do you refer to the supplemental tables in the methodology?
350-361 Discussion is interesting, but what about not capturing the enforcement of social distancing or mask wearing in the retail and/or work area. This is another set of variables to consider that where not measured and are worth noting and discussing.
376 – Findings are consistent with the benefits of UV sterilization but we still do not require indoor air HEPA/UV filtration in public buildings and businesses….unbelievable! Perhaps you could discuss this a little more and cite relevant studies. Here is a list of some:
Beggs, C. B., Avital, E. J., & Hewlett, P. (2021). The use of ultraviolet light in the decontamination of SARS-CoV-2 and other viruses in the built environment: A review. Science of The Total Environment, 777, 145855. https://doi.org/10.1016/j.scitotenv.2021.145855
Heilingloh, C. S., Aufderhorst, U. W., Schipper, L., Dittmer, U., Witzke, O., Yang, D., Zheng, X., Trilling, M., & Alt, M. (2020). Susceptibility of SARS-CoV-2 to UV irradiation. American Journal of Infection Control, 48(10), 1273-1275. https://doi.org/10.1016/j.ajic.2020.07.031
Buonanno, M., Welch, D., Shuryak, I., & Brenner, D. J. (2020). Far-UVC light (222 nm) efficiently and safely inactivates airborne human coronaviruses. Scientific Reports, 10, 10285. https://doi.org/10.1038/s41598-020-67211-2
Raeiszadeh, M., & Adeli, B. (2020). A critical review on ultraviolet disinfection systems against COVID-19 outbreak: Applicability, validation, and safety considerations. ACS Photonics, 7(11), 2941-2951. https://doi.org/10.1021/acsphotonics.0c01245
418 I am glad to see that you begin addressing the availability of medical care facilities at this point. However, you should clearly note that you have not considered medically served and medically underserved (aligns with poverty and rural areas) areas when doing you comparisons, nor did you analyze political affiliation as personal protective practices in this recent epidemic was very aligned with political ideology. You should at least acknowledge that these factors were not measured. Consider adding this to section 463-469.
related to COVID-19 mortality across 3141 U.S. counties showed that highly urbanized 432…you need a comma on this figure.
Author Response
Dear Reviewer # 2,
Thank you for your constructive feedback. We are providing the following responses (marked in red) for each of the following comments:
120 For each day of the week and for each nation, baseline day median values were calculated [22]. For each nation? Do you mean State? Clarify and refine this.
Thank you for this comment. Google’s community mobility reports (CMR) have worldwide mobility data. For U.S., there is county-level data by state. In this sentence, we were referring to countries as nations. In consideration to your comment, we have replaced the word “nation” to counties. We used Google’s county-level mobility data to conduct analysis. (Lines: 119-120).
130 . NOAA’s GHCNd database includes information from over 80,000 weather stations There is a period here that starts this sentence, a typo.
We have taken out the period. Thank you. (Line: 131).
139-148 You used imputation to replace missing data. It is good that you report this and give the MCAR for it, great! Is there any indication as to why the mobility data reflected non-random missing data? Please mention this if there is.
“You might see data gaps for some categories in your region. These gaps are intentional and happen because the data doesn’t meet the quality and privacy threshold—when there isn’t enough data to ensure anonymity.” (Source: Google Mobility Data). Our understanding is that some counties are low in population size which lead to Google’s decision not to disclose their mobility. We included the following line in the methods section, “Google's CMR had missing data for categories in counties where anonymity could not be ensured due to insufficient mobility among the residents.” (Lines: 120-122).
Also, I did not see the MVP for mobility in Supplemental Table 1 figure…I see it is included in table 2, supplemental 2 but that is not related to imputation, was it to difficult to display visually?
Supplemental Figure 1: Missing Value Patterns (MVP) includes all of the independent variables. Variables, “GroceryandPharmacy”, “RetailandRecreation”, and “Workplace” are displayed in the MVP chart.
Not sure I understand the significance of Supplemental Table 3? Is it just to provide information on methodological approach?
Yes.
Where do you refer to the supplemental tables in the methodology?
“Missing value patterns were not identical for mobility data on workplace, retail/recreation, and grocery/pharmacy (Supplemental Figure 1)”. (Lines: 153-154).
“Unweighted Cox regression results for COVID-19-related mortality were reported in Supplemental Table 2.” (Lines: 168-170).
“In the sensitivity analysis (Supplemental Table 4-5), bordering counties (two counties deep) in each state were analyzed (Supplemental Table 3).” (Lines: 172-174).
350-361 Discussion is interesting, but what about not capturing the enforcement of social distancing or mask wearing in the retail and/or work area. This is another set of variables to consider that where not measured and are worth noting and discussing.
Thank you for this comment. This is true and we have updated our limitation section to put a focus on this.
“The lack of information on other seasons (summer, fall, winter), non-pharmaceutical interventions (NPIs) (use of facial masks, social distancing compliance), and socio-economic factors (i.e., income and occupation) are underlying limitations to this study’s analytical models.” (Lines: 478-481).
We also have updated our discussion section with the following:
“Although findings on the state-associated risk of COVID-19 infection and mortality in relation to human mobility are contrary, it is important to consider that Indiana had a less stringent stay-at-home order compared to Kentucky [20]. Indiana residents were allowed to leave home for certain employment, outdoor activities, and to take care of others during March 24 to May 1, 2020. However, travel was restricted, and only life-sustaining businesses (retail, gas station, etc.) could remain open in Kentucky from March 25 to April 29, 2020. Furthermore, social distancing and hygiene were required, and the use of facial masks was promoted in public settings. The use of face masks in community settings is known to reduce the transmission risk of COVID-19 infection [54]. The difference in compliance and enforcement of stay-at-home orders between Indiana and Kentucky may help explain some of the variability in state-level differences in COVID-19 infection and mortality observed in the current study.” (Lines: 458-470).
376 – Findings are consistent with the benefits of UV sterilization but we still do not require indoor air HEPA/UV filtration in public buildings and businesses….unbelievable! Perhaps you could discuss this a little more and cite relevant studies. Here is a list of some:
Beggs, C. B., Avital, E. J., & Hewlett, P. (2021). The use of ultraviolet light in the decontamination of SARS-CoV-2 and other viruses in the built environment: A review. Science of The Total Environment, 777, 145855. https://doi.org/10.1016/j.scitotenv.2021.145855
Heilingloh, C. S., Aufderhorst, U. W., Schipper, L., Dittmer, U., Witzke, O., Yang, D., Zheng, X., Trilling, M., & Alt, M. (2020). Susceptibility of SARS-CoV-2 to UV irradiation. American Journal of Infection Control, 48(10), 1273-1275. https://doi.org/10.1016/j.ajic.2020.07.031
Buonanno, M., Welch, D., Shuryak, I., & Brenner, D. J. (2020). Far-UVC light (222 nm) efficiently and safely inactivates airborne human coronaviruses. Scientific Reports, 10, 10285. https://doi.org/10.1038/s41598-020-67211-2
Raeiszadeh, M., & Adeli, B. (2020). A critical review on ultraviolet disinfection systems against COVID-19 outbreak: Applicability, validation, and safety considerations. ACS Photonics, 7(11), 2941-2951. https://doi.org/10.1021/acsphotonics.0c01245
This is a great suggestion. We have included the following lines in the discussion section, “Far-UVC radiation (207-222 nautical miles [nm]) effectively destroys the SARS-CoV-2 virus without causing damage to human tissues [46]. UV disinfection technologies can be used to disinfect biocontaminate air and surfaces to curb the transmission of the COVID-19 infection [47]. In consideration of the protective nature of UVC radiation against COVID-19 infection, UV disinfection technologies could be implemented in public settings, i.e., health care facilities, shopping malls, and airports, for disinfection of frequently touched surfaces and circulating air streams.” (Lines: 409-415).
418 I am glad to see that you begin addressing the availability of medical care facilities at this point. However, you should clearly note that you have not considered medically served and medically underserved (aligns with poverty and rural areas) areas when doing you comparisons, nor did you analyze political affiliation as personal protective practices in this recent epidemic was very aligned with political ideology. You should at least acknowledge that these factors were not measured. Consider adding this to section 463-469.
Yes, this is true, and we have addressed this as limitations.
“The lack of information on other seasons (summer, fall, winter), non-pharmaceutical interventions (NPIs) (use of facial masks, social distancing compliance), and socio-economic factors (i.e., income and occupation) are underlying limitations to this study’s analytical models. Other limitations include possible discrepancies in reported COVID-19 infection and mortality and a lack of information on state-level healthcare capacity, public health infrastructure, and public adherence to guidelines.” (Lines: 478-483).
related to COVID-19 mortality across 3141 U.S. counties showed that highly urbanized 432…you need a comma on this figure.
Thank you. We have added the comma. (Line: 473).
Reviewer 3 Report
Comments and Suggestions for AuthorsThis manuscript describes a study in Indiana and Kentucky in which the association of Covid-19 infection and mortality with mobility and weather was compared. Increased risk of infection and mortality was associated with higher maximum and minimum temperature and metropolitan status. The authors report that infection and mortality was higher in Indiana than Kentucky and suggest this may be due to “enforcement of and adherence to stay-at-home orders between Indiana and Kentucky”.
Major issues
This type of analysis depends upon the quality of the information provided by the databases but there is very little discussion on the quality of the data held by the New York Times database and the Google database. Furthermore, no citations are provided for either database. The authors use the following variables (“retail/recreation, grocery/pharmacy and workplace”) to assess mobility but no details are provided as to what the numerical values mean and how Google has calculated them. Are these measures really indicative of motility? Further information, references and clarity is required.
The analysis for this period covers the time period from March 1 2020 to May 15 2020. Why was this time period chosen? The stay-at-home orders were declared from March 26 to May 11 in Kentucky and March 24 to May 1 in Indiana and if the aim was to look at the effect of stay-at-home orders surely the same time period (March 24 to May 1) only should have been examined in both States.
The authors suggest that their results “may be due to enforcement of and adherence to stay-at-home orders between Indiana and Kentucky”. The results do not appear to be that consistent with this conclusion. Cox regression analysis for infection and mortality (Tables 4, 5) indicates that the main risk factors were metropolitan status and living in Indiana after adjustment for motility and other factors. What is the evidence that supports their conclusion?
Minor issues
Much of the material identified as supplementary material is correctly identified but they are placed in the main text. This should be removed and placed in a supplementary file.
Lines 118-119: The following statement “Monday data would be contrasted with Monday baseline data for the same date prior to the COVID-19 pandemic” indicates that the data for the same date was compared pre and during Covid. Is this correct? If the same date, then presumably it is for the previous year? Yet the authors indicate that the baseline data was for “a five-week period beginning on January 3, 2020, and ending on February 6, 2020”. Please clarify.
Table 2: Various variables are provided but without units. Units need to be provided for each of them.
Table 3 in essence provides one set of data which can be written as text. This Table should be deleted.
Discussion: The key findings of the study (lines 436-448) really should be introduced much earlier in the discussion.
Sections 3.1.2 and 3.1.3 have the same title. Please change the title of one.
Author Response
Dear Reviewer # 3,
Thank you for your constructive feedback. We are providing the following responses (marked in red) for each of the following comments:
- This type of analysis depends upon the quality of the information provided by the databases but there is very little discussion on the quality of the data held by the New York Times database and the Google database. Furthermore, no citations are provided for either database. The authors use the following variables (“retail/recreation, grocery/pharmacy and workplace”) to assess mobility but no details are provided as to what the numerical values mean and how Google has calculated them. Are these measures really indicative of motility? Further information, references and clarity is required.
Thank you for the comment. We have added the following to provide information on the New York Times database.
“The New York Times released data files with cumulative counts of COVID-19 cases and deaths at the state and county level in the U.S. over time. Data was collected from state and local governments and health departments. COVID-19 case and death counts included both laboratory-confirmed and probable cases using criteria that were developed by states and the federal government [21]. Cumulative COVID-19 infection and mortality county-day rates were computed from the New York Times for Indiana and Kentucky.” (Lines: 105-110).
Following is the details we provided for Google database. We slightly updated it to reflect your comment.
“Google’s community mobility reports (CMR) provide movement trends over time by geography across different categories of places such as retail, recreation, groceries, pharmacies, parks, public transit stations, workplaces, and residential. For each geographical category, CMR contains the percent change in activity from baseline days prior to the introduction of COVID-19. Daily activity variations are contrasted with the matching baseline figure day; for instance, Monday data would be contrasted with Monday baseline data for the same date prior to the COVID-19 pandemic. For each day of the week and for each county, Google calculated baseline day median values. Google's CMR had missing data for categories in counties where anonymity could not be ensured due to insufficient mobility among the residents [22]. Daily human mobility data in the analytical models with respect to essential commuting needs during the stay-at-home order were included: (1) retail/recreation; (2) grocery/pharmacy; and (3) workplace.” (Lines: 113-124).
- The analysis for this period covers the time period from March 1 2020 to May 15 2020. Why was this time period chosen? The stay-at-home orders were declared from March 26 to May 11 in Kentucky and March 24 to May 1 in Indiana and if the aim was to look at the effect of stay-at-home orders surely the same time period (March 24 to May 1) only should have been examined in both States.
We chose this time period as restrictions were slowly being applied before the issuance of executive orders for both states during the mentioned timeframe. Also, the executive orders were being phased out slowly after the dates stated. We included a few days before (mostly) and after to count for those restrictions.
- The authors suggest that their results “may be due to enforcement of and adherence to stay-at-home orders between Indiana and Kentucky”. The results do not appear to be that consistent with this conclusion. Cox regression analysis for infection and mortality (Tables 4, 5) indicates that the main risk factors were metropolitan status and living in Indiana after adjustment for motility and other factors. What is the evidence that supports their conclusion?
Thank you for this comment. Yes, you have rightly pointed out the main risk factors, metropolitan status and living in Indiana. We have emphasized on these risk factors in discussion section as well as the conclusions section.
“A key finding from the current study is that Indiana residents were at higher risk of COVID-19 infection (aHR = 1.18, 95% CI = 1.10–1.26) and mortality (aHR = 1.59, 95% CI = 1.57–1.60) compared to Kentucky residents. In addition to state of residence, other in-fluences included human mobility, weather, and metropolitan status during the period of stay-at-home order. Precipitation and UV index had a protective effect against COVID-19 infection and mortality, similar to others [33,34,44,45]. Limited evidence suggests metropolitan areas are at increased risk of COVID-19 infection and mortality, probably attributable to population density [55,56]. Average precipitation and UV index were higher in Kentucky compared to Indiana during the study period by 59% and 11%, respectively. Indiana has 12% more metropolitan counties compared to Kentucky, and the number of non-metropolitan counties was lower in Indiana by 51%. Precipitation, UV index, and metropolisation may contribute to the state-associated increased risk of COVID-19 infection and mortality among Indiana residents compared to Kentucky residents.” (Lines: 439-451).
We have also updated the discussion section that should help address the comment.
“Although findings on the state-associated risk of COVID-19 infection and mortality in relation to human mobility are contrary, it is important to consider that Indiana had a less stringent stay-at-home order compared to Kentucky [20]. Indiana residents were allowed to leave home for certain employment, outdoor activities, and to take care of others during March 24 to May 1, 2020. However, travel was restricted, and only life-sustaining businesses (retail, gas station, etc.) could remain open in Kentucky from March 25 to April 29, 2020. Furthermore, social distancing and hygiene were required, and the use of facial masks was promoted in public settings. The use of face masks in community settings is known to reduce the transmission risk of COVID-19 infection [57]. The difference in compliance and enforcement of stay-at-home orders between Indiana and Kentucky may help explain some of the variability in state-level differences in COVID-19 infection and mortality observed in the current study.” (Lines: 458-470).
- Much of the material identified as supplementary material is correctly identified but they are placed in the main text. This should be removed and placed in a supplementary file.
Thank you. Correctly pointed out. Journal asked for a single file. We thought that the journal would create a separate file for supplementary material before publication. We have deleted the supplementary material from the manuscript and have created a separate file.
Lines 118-119: The following statement “Monday data would be contrasted with Monday baseline data for the same date prior to the COVID-19 pandemic” indicates that the data for the same date was compared pre and during Covid. Is this correct? Yes. If the same date, then presumably it is for the previous year? Yes. Yet the authors indicate that the baseline data was for “a five-week period beginning on January 3, 2020, and ending on February 6, 2020”. Please clarify.
Thank you very much for pointing this out. The quote line was inserted by mistake. According to Google official website for the database, “For each region-category, the baseline isn’t a single value—it’s 7 individual values.”. We hope this clarifies the concern. We have updated the statement to the following:
“For each geographical category, CMR contains the percent change in activity from baseline days prior to the introduction of COVID-19.” (Lines: 115-117).
Table 2: Various variables are provided but without units. Units need to be provided for each of them.
Thank you. We have updated Table 2 with units.
Table 3 in essence provides one set of data which can be written as text. This Table should be deleted.
We have deleted this Table 3 and updated the other following table numbers according both to table titles and text.
Discussion: The key findings of the study (lines 436-448) really should be introduced much earlier in the discussion.
Thank you for the comment. We mentioned the following in the initial discussion, which is the key finding of this study, “The present study found that Indiana residents were at a higher risk of contracting COVID-19 infection by 18% and, if infected, 59% more likely to have a fatal outcome compared to Kentucky residents after controlling for human mobility, weather, and metropolitan status.” (Lines: 337-340). Following this, we discussed the independent variables that we controlled for leading to a concluding discussion tying to differences observed in the states.
Sections 3.1.2 and 3.1.3 have the same title. Please change the title of one.
Thank you. It has been addressed.
Round 2
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors have addressed the majority of the issues I raised about the manuscript. I would still suggest though that the final sentence of the abstract "This may be associated with enforcement of and adherence to stay-at-home orders between Indiana and Kentucky." still puts too much emphasis on the role of these factors given the limited evidence in this study for their role. I would recommend putting this into the overall context of the study results and suggesting that it is rewritten. Indeed the text (lines 453-455) "The difference in compliance and enforcement of stay-at-home orders between Indiana and Kentucky may help explain some of the variability in state-level differences in COVID-19 infection and mortality observed in the current study" indicates that this may explain only some of the variability.
Author Response
The authors have addressed the majority of the issues I raised about the manuscript. I would still suggest though that the final sentence of the abstract "This may be associated with enforcement of and adherence to stay-at-home orders between Indiana and Kentucky." still puts too much emphasis on the role of these factors given the limited evidence in this study for their role. I would recommend putting this into the overall context of the study results and suggesting that it is rewritten. Indeed the text (lines 453-455) "The difference in compliance and enforcement of stay-at-home orders between Indiana and Kentucky may help explain some of the variability in state-level differences in COVID-19 infection and mortality observed in the current study" indicates that this may explain only some of the variability.
Thank you for the feedback. We have updated the final line from the abstract with the following:
"This may be attributed to variations in stay-at-home orders compliance and enforcement between Indiana and Kentucky."