COVID-19 Mortality Among Hospitalized Medicaid Patients in Kentucky (2020–2021): A Geospatial Study of Social, Medical, and Environmental Risk Factors

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

1. Images 1 to 4 are blurry. It is recommended to enhance the clarity of the images.
2. It is suggested to supplement the details of sample size and data sources in the methods section of the abstract. Currently, only "cohort of hospitalized Medicaid patients" is mentioned, but there is a lack of specific quantity information.
3. There are two expressions in the abstract, "hours" and "96 hours plus". It is recommended to use a unified expression. Suggest changing 'Adjusted R2' to 'Adjusted R²'.
4. The current conclusion section is somewhat weak, and it is suggested to add an analysis of the geographical characteristics of Appalachian region specific risks.
5. There is a logical jump in the original text: the inference that "PM2.5 is a potential carrier" needs to be supplemented with key evidence chains, and it is suggested to revise the mechanism of virus particle binding.
6. Suggest supplementing the specific impacts of traditional model limitations to enhance the depth of methodological criticism. Explain the consequences of spatial non stationarity through examples and strengthen the necessity of using geographically weighted regression.
7. The current paragraph order jumps, causing a logical break between "However, several studies..." and the context. It is recommended to reconstruct the logical coherence of the paragraph.
8. The moderating effect of spatial heterogeneity on the effect of medical resources. The original text only mentions that there is no significant correlation between medical manpower, which can be further explored.
9. The spatial differences in COPD/mechanical ventilation were attributed to "differences in population health activities", but measurement data for relevant variables such as public health funding and community health cooperation project coverage were not provided in this study.
10. Suggest exploring potential confounding factors, such as whether better primary healthcare in urban areas (such as early intervention for COPD) has reduced the rate of severe disease conversion. Additional references are needed to support the use of "urbanized medical resources to buffer COPD risk".

Author Response

Thank you very much for the constructive feedback. They helped make the paper much better. We have addressed all your comments and provided the updates along with line numbers. 

1. Images 1 to 4 are blurry. It is recommended to enhance the clarity of the images.

This will help us ensure the best visual aspect of the maps. We have replaced Figures 1 to 4 in the manuscript with high-resolution images (300 DPI). This has improved clarity and visual sharpness. Additionally, we have uploaded a separate Word file containing all figures of their actual size to assist with the production process if needed.

1. It is suggested to supplement the details of sample size and data sources in the methods section of the abstract. Currently, only "cohort of hospitalized Medicaid patients" is mentioned, but there is a lack of specific quantity information.

We have revised the Background section of the abstract to include the sample size and data source as per your suggestion. The updated sentence now reads:

"Geospatial associations for COVID-19 mortality were estimated using a cohort of 28,128 hospitalized Medicaid patients identified from 2020–2021 Kentucky Health Facility and Services (KHFS) administrative claims data." Lines: 13-15.

1. There are two expressions in the abstract, "hours" and "96 hours plus". It is recommended to use a unified expression. Suggest changing 'Adjusted R2' to 'Adjusted R²'.

We have standardized the terminology by using “96 hrs. plus” consistently throughout the abstract when referring to prolonged mechanical ventilation use. We have also corrected the formatting of “adjusted R2” to the proper superscript form: “adjusted R²”.

1. The current conclusion section is somewhat weak, and it is suggested to add an analysis of the geographical characteristics of Appalachian region specific risks.

We have revised the Conclusion section of the abstract to highlight geographic and socioeconomic vulnerabilities of rural and Appalachian counties. “A risk of COVID-19 mortality was observed among patients with COPD and prolonged mechanical ventilation use, after controlling for social determinants, healthcare workforce, and PM_2.5 in rural and Appalachian counties of Kentucky. These counties are characterized by persistent poverty, healthcare workforce shortages, economic distress, and poor population health outcomes. Improving population health protection through multisector collaborations in rural and Appalachian counties may help reduce future health burdens.” Lines: 28-34.

1. There is a logical jump in the original text: the inference that "PM2.5 is a potential carrier" needs to be supplemented with key evidence chains, and it is suggested to revise the mechanism of virus particle binding.

We revised the relevant section of the Introduction in response to your great comment. The updated sentence now reads:

“Although SARS-CoV-2 particles are much smaller—approximately 70 to 90 nanometers (nm)—they may become embedded within or adsorbed onto larger particles such as PM2.5, potentially facilitating indirect airborne transmission by allowing viral particles to remain suspended in the atmosphere for extended periods or travel longer distances.” Lines: 52-56.

We have also added a citation to Belosi et al. (2021), a study published in Scientific Reports (Nature), which supports this point mentioned above.

1. Suggest supplementing the specific impacts of traditional model limitations to enhance the depth of methodological criticism. Explain the consequences of spatial non stationarity through examples and strengthen the necessity of using geographically weighted regression.

We updated the paragraph on traditional model in the Introduction to the following, “Traditional regression methods, i.e., ordinary least squares (OLS) and binary logistic regression, can identify a global association between an outcome and one or more independent variables. However, these regression models assume spatial stationarity—meaning the relationship between the independent variables and outcome is constant across all locations in the study area. This assumption can limit the model’s ability to account for local variations and lead to misleading or biased estimates.

For instance, the effect of chronic conditions like COPD on COVID-19 mortality may be stronger in rural counties with limited healthcare access than in urban counties with more resources. Failing to account for such heterogeneity can mask high-risk areas and under-inform interventions.” Lines: 62-71.

1. The current paragraph order jumps, causing a logical break between "However, several studies..." and the context. It is recommended to reconstruct the logical coherence of the paragraph.

Great suggestion. To improve the flow as per the suggestion, we have revised the following:

“Although results from this study did not reveal any independent effect of social determinants (e.g., sex, SDI score), healthcare workforce, or air quality (PM2.5) on county-level COVID-19-related mortality among Medicaid patients, numerous studies have demonstrated statistically significant associations between COVID-19 outcomes—particularly mortality—and social determinants of health among marginalized communities [3].

There are limited studies that have explored the relationship between the availability of the healthcare workforce and COVID-19 mortality in the U.S. Findings from these studies indicate a significant association between a lack of healthcare workforce and COVID-19 mortality [37, 38]. However, the current study found no such significant relationship between county-level physician and nurse rates and COVID-19-related mortality among Medicaid patients.” Lines: 302-337.

1. The moderating effect of spatial heterogeneity on the effect of medical resources. The original text only mentions that there is no significant correlation between medical manpower, which can be further explored.

This is rightly pointed out. We have updated the discussion section to include deeper exploration suggested by you:

“There are limited studies that have explored the relationship between healthcare workforce availability and COVID-19 mortality in the U.S. Findings from these studies indicate a significant association between provider shortages and increased mortality [37, 38]. However, the current study found no significant relationship between county-level physician and nurse rates and COVID-19-related mortality among Medicaid patients in either the global OLS regression or the local MGWR model.

This absence of significance may reflect broader structural mechanisms within hospitals. The relationship between healthcare workforce availability and patient outcomes may be influenced by external factors such as infection control policies and practices, ICU surge capacity, staffing ratios, availability of ventilators, and the operational readiness of hospitals to manage surges [39, 40]. These nuances are not captured by our model.” Lines: 332-42.

 

1. The spatial differences in COPD/mechanical ventilation were attributed to "differences in population health activities", but measurement data for relevant variables such as public health funding and community health cooperation project coverage were not provided in this study.
2. Suggest exploring potential confounding factors, such as whether better primary healthcare in urban areas (such as early intervention for COPD) has reduced the rate of severe disease conversion. Additional references are needed to support the use of "urbanized medical resources to buffer COPD risk".

Comments # 9 and 10 were helpful in polishing the ending remarks for the discussion. We make the following changes to incorporate your valuable comments:

“Population health interventions, particularly those focused on promoting physical activity and addressing risk factors such as smoking, can significantly impact the health of individuals with COPD [49]. Urban areas generally have better access to early population health intervention for COPD compared to rural areas [50]. These interventions may help achieve better lung functioning and avoid critical care, i.e., mechanical ventilation, thereby improving overall health outcomes.

Notably, differences in public health funding between rural and urban areas are key drivers of variations in population health activities and interventions. Urban centers often receive a larger share of resources, while rural communities may experience limitations in access to services and funding[51, 52]. Future research should directly measure and evaluate these contextual factors to better understand their contribution to geographic disparities in COVID-19-related outcomes.” Lines: 398-409.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

This study employs geospatial analysis methods to reveal the correlation between the COVID-19 mortality rate among hospitalized Medicaid patients in Kentucky, U.S., during 2020–2021 and factors such as air pollution, chronic lung diseases, and shortages of healthcare resources. The study used a Multiscale Geographically Weighted Regression (MGWR) model and found distinct differences in influencing factors between urban and rural areas. This study not only provides a new method for public health geography analysis but also plays an important role in optimizing the allocation of healthcare resources and improving the health of vulnerable groups.

1. It is suggested that the title be optimized to be more comprehensive and innovative.
2. It is suggested that the introduction section include more research content targeting the Medicaid patient population.
3. It is suggested that the novelty and contributions of this study relative to existing literature reviews be emphasized.
4. It is suggested that the rationale for selecting Kentucky as the study area be supplemented, such as how its geographical and socioeconomic characteristics affect COVID-19 mortality.
5. It is suggested that the data collection and processing procedures be described in detail, including data sources, data organization, and variable definitions.
6. It is suggested that the rationale for choosing the MGWR spatial regression model be supplemented.
7. It is suggested that the rationale for selecting specific independent variables be discussed and that the correlation of these variables with COVID-19 mortality be explained.
8. It is suggested that a discussion be held on in-depth analysis and interpretation of the results of the MGWR model, such as the reasons why some variables are not significant.
9. It is suggested that the policy and practical implications of the study findings be discussed, and specific recommendations or measures for improvement be proposed.

Author Response

Thank you very much for the constructive feedback. They helped make the paper much better. We have addressed all your comments and provided the updates along with line numbers.  

1. It is suggested that the title be optimized to be more comprehensive and innovative.

Updated title, “COVID-19 Mortality Among Hospitalized Medicaid Patients in Kentucky (2020–2021): A Geospatial Study of Social, Medical, and Environmental Risk Factors”

1. It is suggested that the introduction section include more research content targeting the Medicaid patient population.
2. It is suggested that the novelty and contributions of this study relative to existing literature reviews be emphasized.
3. It is suggested that the rationale for selecting Kentucky as the study area be supplemented, such as how its geographical and socioeconomic characteristics affect COVID-19 mortality.

We have added the following to the introduction section to address comments # 2-4:

  “Medicaid patients generally have a high prevalence of COPD, limited access to care, and greater exposure to adverse social determinants of health—making them more susceptible to COVID-19-related hospitalization and mortality [25]. Accordingly, Kentucky is an ideal location for examining COVID-19-related mortality, as it has a high prevalence of COPD and a large Medicaid-enrolled population [26, 27]. The state’s mix of rural, urban, and Appalachian counties introduces considerable geographic and socioeconomic diversity. Moreover, access to healthcare services varies widely across counties, with rural and Appalachian areas experiencing greater provider shortages [28].

 

Approximately half of the Kentucky physicians (49.1%) work in Fayette (19.6%, physician ratio [PR]: 1:148) or Jefferson (29.5%, PR: 1:244) counties. University of Louisville Health and University of Kentucky Healthcare, two of Kentucky’s largest healthcare organizations, are located in Jefferson and Fayette counties, respectively [28]. Therefore, patients from neighboring counties to Fayette or Jefferson have better access to care compared to far-distant counties.

 

The spatial heterogeneity observed in county-level mortality among hospitalized COVID-19 Medicaid patients provides a unique opportunity to measure non-stationary geographic variation using a MGWR model across counties in Kentucky (Figure 1). To the best of our knowledge, no prior study has conducted a spatial analysis focused specifically on COVID-19-related mortality among Medicaid patients.” Lines: 89-110.

 

1. It is suggested that the data collection and processing procedures be described in detail, including data sources, data organization, and variable definitions.

We have added or slightly modified the following in the methods section to provide more details as suggested:

“KFHS data provides the following patient demographic information: (1) adult age groups: 18-20, 21-24, 25-29, 30-34, 35-39, 40-44, 45-49, 50-54, 55-59, 60-64; and (2) sex: male, female. The CDC’s 2021 national COVID-19 death counts were used to identify the age-specific burden of mortality [31]. This identification served as guidance to recode the original age groups into the following categories: (1) 18–54 and (3) 55–64.” Lines: 125-129.

 

“Robert Graham Center’s SDI is a composite measure of area-level deprivation based on the following seven demographic parameters obtained from the American Community Survey: (1) percent living in poverty; (2) percent with less than 12 years of education; (3) percent single-parent households; (4) the percentage living in rented housing units; (5) the percentage living in an overcrowded housing unit; (6) the percentage of households without a car; and (7) the percentage of non-employed adults under 65 years of age.

 

SDI can help evaluate socioeconomic heterogeneity in health outcomes. SDI scores range from 0 to 100. Members of communities with higher SDI scores are at increased risk of disease and have limited access to care, which contributes to unmet healthcare needs and poor patient outcomes [35].

 

2021 county-level physician and nurse rates per 100,000 were obtained from the Area Health Resources Files (AHRF) [36]. The 2020 county-level highest annual average concentration of PM2.5 was obtained from the Centers for Disease Control and Prevention (CDC)—National Environmental Public Health Tracking Network [37]. County-level highest annual average concentration of PM2.5 for 2021 is not available in the database.

 

County-level SDI score, physician and nurse rates per 100,000, and annual average PM2.5 were merged with aggregated patient data at the county level using Federal Information Processing Standards (FIPS) codes.” Lines: 151-168.

 

1. It is suggested that the rationale for choosing the MGWR spatial regression model be supplemented.

We updated the paragraph on the traditional model in the introduction to the following, “Traditional regression methods, i.e., ordinary least squares (OLS) and binary logistic regression, can identify a global association between an outcome and one or more independent variables. However, these regression models assume spatial stationarity—meaning the relationship between the independent variables and outcome is constant across all locations in the study area. This assumption can limit the model’s ability to account for local variations and lead to misleading or biased estimates. For instance, the effect of chronic conditions like COPD on COVID-19 mortality may be stronger in rural counties with limited healthcare access than in urban counties with more resources. Failing to account for such heterogeneity can mask high-risk areas and under-inform interventions.” Lines: 62-71.

 

Proceeding paragraphs introduce the role of GWR and MGWR in accounting for non-spatial stationarity effects.

1. It is suggested that the rationale for selecting specific independent variables be discussed and that the correlation of these variables with COVID-19 mortality be explained.

The updates to the introduction for previous suggestions (2-5), in addition to current content in the introduction, address this comment.

1. It is suggested that a discussion be held on in-depth analysis and interpretation of the results of the MGWR model, such as the reasons why some variables are not significant.

We have provided the following updated contents in the discussion in response to this comment:

“Although results from this study did not reveal any independent effect of social determinants (e.g., sex, SDI score), healthcare workforce, or air quality (PM2.5) on county-level COVID-19-related mortality among Medicaid patients, numerous studies have demonstrated statistically significant associations between COVID-19 outcomes—particularly mortality—and social determinants of health among marginalized communities [3].

There are limited studies that have explored the relationship between the availability of the healthcare workforce and COVID-19 mortality in the U.S. Findings from these studies indicate a significant association between a lack of healthcare workforce and COVID-19 mortality [37, 38]. However, the current study found no such significant relationship between county-level physician and nurse rates and COVID-19-related mortality among Medicaid patients.” Lines: 322-312.

“There are limited studies that have explored the relationship between healthcare workforce availability and COVID-19 mortality in the U.S. Findings from these studies indicate a significant association between provider shortages and increased mortality [37, 38].

 

However, the current study found no significant relationship between county-level physician and nurse rates and COVID-19-related mortality among Medicaid patients in either the global OLS regression or the local MGWR model.

 

This absence of significance may reflect broader structural mechanisms within hospitals. The relationship between healthcare workforce availability and patient outcomes may be influenced by external factors such as infection control policies and practices, ICU surge capacity, staffing ratios, availability of ventilators, and the operational readiness of hospitals to manage surges [39, 40]. These nuances are not captured by our model.” Lines: 307-342.

 

1. It is suggested that the policy and practical implications of the study findings be discussed, and specific recommendations or measures for improvement be proposed.

We have provided the following updated contents in the end of the discussion in response to this comment:

 

“Population health interventions, particularly those focused on promoting physical activity and addressing risk factors such as smoking, can significantly impact the health of individuals with COPD [49]. Urban areas generally have better access to early population health intervention for COPD compared to rural areas [50]. These interventions may help achieve better lung functioning and avoid critical care, i.e., mechanical ventilation, thereby improving overall health outcomes.

Notably, differences in public health funding between rural and urban areas are key drivers of variations in population health activities and interventions. Urban centers often receive a larger share of resources, while rural communities may experience limitations in access to services and funding[51, 52]. Future research should directly measure and evaluate these contextual factors to better understand their contribution to geographic disparities in COVID-19-related outcomes.” Lines: 398-409.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

A well done research project that shows how useful spatial approach is within the context of epidemiology and health care provision. No flaws, nice work!

Author Response

Thank you very much for the kind words.