key: cord-0314565-0lmerywo
authors: Gerken, J.; Zapata, D.; Kuivinen, D.; Zapata, I.
title: Comorbidities and sociodemographic factors on COVID-19 fatalities
date: 2022-05-21
journal: nan
DOI: 10.1101/2022.05.20.22275397
sha: 18b1ef21a47893a403da47f686e27870d4fbda71
doc_id: 314565
cord_uid: 0lmerywo

Introduction: Previous studies have evaluated comorbidities and sociodemographic factors individually or by type but not comprehensively. This study aims to analyze the influence of a wide variety of factors in a single study to better understand the big picture of their effects on case-fatalities. Methods: County-level comorbidities, social determinants of health such as income and race, measures of preventive healthcare, age, education level, average household size, population density, and political voting patterns were all evaluated on a national and regional basis. Analysis was performed through Generalized Additive Models and adjusted by CCVI. Results: Factors associated with reducing COVID-19 case fatality rates were mostly sociodemographic factors such as age, education and income, and preventive health measures. Obesity, minimal leisurely activity, binge drinking, and higher rates of individuals taking high blood pressure medication were associated with increased case fatality rate in a county. Political leaning influences case case-fatality rates. Regional trends showed contrasting effects where larger household size was protective in the Midwest, yet harmful in Northeast. Notably, higher rates of respiratory comorbidities such as asthma and COPD diagnosis were associated with reduced case-fatality rates in the Northeast. Increased rates of CKD within counties were often the strongest predictor of increased case-fatality rates for several regions. Conclusion: Our findings highlight the importance of considering the full context when evaluating contributing factors to case-fatality rates. The spectrum of factors identified in this study must be analyzed in the context of one another and not in isolation.

case fatality rate in a county. Political leaning influences case case-fatality rates. Regional trends showed 23 contrasting effects where larger household size was protective in the Midwest, yet harmful in Northeast. 24

Notably, higher rates of respiratory comorbidities such as asthma and COPD diagnosis were associated with 25 reduced case-fatality rates in the Northeast. Increased rates of CKD within counties were often the strongest 26 predictor of increased case-fatality rates for several regions. 27

Conclusion: Our findings highlight the importance of considering the full context when evaluating contributing 28 factors to case-fatality rates. The spectrum of factors identified in this study must be analyzed in the context of 29 one another and not in isolation. 30 31 KEYWORDS: Sociodemographic, CCVI, politics, county-level 32 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.

The SARS-CoV-2 Virus, also known as the COVID-19 virus, has currently led to over 990,000 (April, 2022) 34 deaths in the United States 1 . The virus has ravaged not only the United States financially and put a stop to every 35 in-person activity for the last two years 2 but has highlighted how embedded community traits affect the 36 outcome of a whole community. In the end, it is characteristics like comorbidities and sociodemographic factors 37 that play in a defining role in determining a community's fatality outcome 3 . 38

While it has been shown that a wide range of comorbidities has an impact COVID-19 outcomes; there 39 has been a great effort to define the specific contributions of comorbidities in their impact on COVID-19 40 morbidity and mortality rates 4-7 . Comorbidities such as hypertension 8 , diabetes mellitus 9 , chronic kidney 41 disease 10 , chronic obstructive pulmonary disease 11 , and cardiovascular disease 12 among others, have important 42 repercussions on COVID-19 outcomes. However, there are also a variety of reasons for which to consider some 43 sociodemographic factors as deleterious, such as lacking healthcare insurance, being of a specific racial 44 background 7,13 , or even voting patterns. All these factors combined would point in a direction that suggests that 45 differences in COVID-19 fatality rates are a potential outcome of embedded community characteristics. 46

Even though previous studies have examined factors influencing COVID-19 fatalities [4] [5] [6] 14 , no study has 47 performed a comprehensive analysis of all the aforementioned factors together to the same extent as our study. 48

In our study, a multitude of key community indicators such as comorbidities, sociodemographic factors 49 (including voting patterns), and determinants of health have been examined to reflect trends and potential 50 associations that can be compared against each other. Therefore, the objective of this study was to perform a 51 comprehensive evaluation of comorbidities, sociodemographic factors, and determinants of health at a national 52 level using county aggregated to define their association to COVID-19 case-fatalities. These patterns may allow 53 us to alter the way communities handle public health crises, utilize public health interventions that could deflect 54 harmful outcomes, and provide resources to communities in a timely manner based on their community 55 characteristics. 56

Datasets 59

The focus of the study was to evaluate regional trends of COVID-19 case-fatality rate compared to 60 comorbidities and sociodemographic factors. This study was vetted and categorized as exempt by the 61 Institutional Research Board. Our study utilized countywide data for each county in the entire continental 62 United States and Hawaii; Alaska and Puerto Rico were excluded from the analysis due to differences in their 63 county data reporting. COVID-19 case-fatality rates were gathered from the COVID-19 Community Profile Report 64 15 for the January 2-8, 2021 week cutoff, this report included the CCVI 16 . This cutoff week was selected because 65 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Page 3 of 15 it allowed for the evaluation of the COVID-19 case fatality rate, without the influence of the vaccines or newer, 66 more infectious strains. Rates of various comorbidities such as chronic obstructive pulmonary disease (COPD), 67 hypertension, cancer, asthma, chronic heart disease (CHD), cholesterol, diabetes, chronic kidney disease (CKD), 68 smoking, stroke, and obesity were obtained on a per-county basis from the CDC 2020 Population Level Analysis 69 and Community Estimates (PLACES) project 17 . Rates of poor mental health, binge drinking, lack of health 70 insurance, time allocated to leisurely activity, and preventive care consisting of cervical cancer screening, 71 routine doctor visits, dental visits, cholesterol screening, and routine mammography were obtained from the 72 2020 CDC PLACES Project, as well. Other variables such as average household size and population density for 73 each county were acquired from the United States Census Bureau COVID-19 website 18 . Latitude of each county 74 was also included in the analysis and obtained from the United States Gazetteer Files 19 from the United States 75 Census Bureau. The 2020 Presidential voting records of each county were obtained from the Harvard Dataverse 76 20 . Racial makeup in each county was obtained from the 2020 decennial United States census 21 , while income, 77 age, and education level were retrieved from the 2019 American Community Survey 5-year estimates 22 . 78

79 Data was evaluated for associations using a Generalized Additive Models (GAMs) approach. COVID-19 80 case fatality rates per 100k people were set as the dependent variable while each comorbidity, 81 sociodemographic and health determinant factor was set as an independent variable. All models were adjusted 82 using COVID-19 Community Vulnerability Indexes (CCVI) 16 which normalized the data for inherent inequity 23 . 83

This study uses individual counties as the experimental unit. All analyses were evaluated on the national sample 84 and in a per region samples, as to determine regional patterns. Independent variables were introduced into the 85 model using smoothing splines starting with three degrees of freedom. Models assumed Gaussian residual 86 distributions. All analyses were performed using PROC GAM in SAS/STAT v.9.4 (SAS Inc., Cary, NC). Risk ratios 87 were estimated with 99% confidence intervals and the coefficients sign determined effect directions. Negative 88 coefficients indicated a reduction in COVID-19 case-fatality rates while positive coefficients indicated an increase 89 in case-fatalities. Coefficient standardization was done with a normally distributed Z-score transformation. All 90 associations presented were tested using two-tailed tests. Regional pattern models were performed 91 independently to identify the top contributors -negatively and positively associated. Even with 99% confidence 92 intervals, all tests were declared significant at a Bonferroni threshold. Regional pattern top contributors that did 93 not reach the Bonferroni threshold are indicated in the figures. 94

National level trends 96 Data from 3140 counties was included in our analysis. Our GAM approach examined the data for 97 associations to case-fatalities, all these values were adjusted to CCVI to normalize the inherent differences a raw 98 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10. 1101 /2022 Page 4 of 15 risk ratio would estimate, and the risk ratio standardized estimates are presented in Figure 1 . Raw estimates 99

showcase the extent of the association without considering their spread while standardized estimates adjust the 100 extent of the association to the spread. Raw estimate findings ( Figure 1A ) revealed that comorbidities above 101 sociodemographic factors have the largest effects associated with case-fatalities; however, these associations 102 can go in both positive and negative directions. A diagnosis of cancer provided the largest effect decreasing 103 COVID-19 case-fatalities while CKD and stroke had the largest effects increasing them. Similarly, asthma 104 (decreased risk), CHD (increased risk) and diabetes (increased risk) displayed moderate effects. Household size 105 was the largest significant sociodemographic factor in positive outcomes of COVID-19 with an effect in a range 106 comparable to relevant comorbidities. Other demographic effects such as age and education displayed 107 significant associations that reduced or increased case-fatality risk. Populations with higher educational 108 achievements displayed significantly reduced case-fatality rates. Increased income always displayed a protective 109 effect. Political preference was significantly associated with case-fatalities such as voting for Biden reduced case-110 fatalities while voting for Trump had the opposite effect. Racial and ethnic backgrounds were only associated to 111 COVID-19 case-fatalities for Pacific Islanders, Asian and Black groups. Determinants of health such as cervical 112 cancer screening and people using high blood pressure medication also showed mixed direction associations. 113

Cervical cancer screening had the largest case-fatality reducing effect from this category while the people using 114 high blood pressure medication had the largest opposite effect. Standardized estimates ( Figure 1B) show a 115 different perspective and allow for comparisons across factors as they are standardized. In this case, a routine 116 colonoscopy procedure was found to be the largest protecting effect against COVID-19 case-fatality followed by 117 a combination of sociodemographic factors such as age, education, and income. On the other side of the 118 spectrum, obesity had the largest negative impact deleterious outcome in COVID-19 patients followed by having 119 no leisure physical activity, binge drinking and higher proportions of people taking high blood pressure 120 medication in a specific county. 121

The main analysis was also replicated independently within each of the ten US Health and Human 123 Services defined regions. These models were also adjusted by CCVI. Risk ratio effect estimates for the ten 124 regions are displayed in Figure 2 . These analyses detected a wide array of effects that in some cases go in 125 opposite directions across all regions. No single factor was consistently associated for all regions suggesting that 126 regional associations are not generalizable. These regional assessments all have different sample sizes that are 127 based on the number of counties within each state. These can range from 67 (Region 1) to 736 (Region 4); 128 however, this discrepancy did not affect the capacity of each regional analysis to detect associations at a 129

Bonferroni level (adjusted for 470 tests across all sets). The top variables reducing and increasing COVID-19 case-130 fatalities for each region are presented in Figure 3 . The map in Figure 3A shows that the strongest protective 131 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10. 1101 /2022 Page 5 of 15 regional effects were observed towards the east of the country where Stroke and Cancer Diagnosis were highly 132 protective in the Northeast regions (Regions 1 and 2) and being of Pacific Islander descent was protective in the 133 Southeast United States (Region 4). The Midwest displayed some moderate protective effects where household 134 size was the top reducing factor in two regions (Region 5 and 7). Western regions displayed smaller COVID-19 135 case-fatality protective effects. Regions in the Western United States displayed smaller effect sizes in 136 comparisons to regions in the South or Midwest. Lastly, Figure 3B shows variables that most significantly 137 contributed to increased COVID-19 case-fatalities. Chronic kidney disease (CKD) was the most prevalent 138 comorbidity across several regions (Regions 4, 6, 8 and 9 ). Among other findings, household size had a negative 139 impact on COVID-19 outcomes specifically in the Northeast regions (Regions 1 and 2) . As previously mentioned, 140 some of the top variables displayed opposing effects which suggested that the interpretation must be done in 141 context with the specific characteristics of that region, and interpretations cannot be generalized to others. 142

The objective of this paper was to evaluate COVID-19 case-fatality rate in conjunction with 145 sociodemographic factors, comorbidities, and determinants of heath. Previous studies have evaluated the 146 influence of various socioeconomic factors 7,14,24 and comorbidities 4-6 on COVID-19 case-fatality rate; however, 147 these analyses do not pair together their findings to be comparable with each other. Our study evaluates COVID-148 19 fatality rates from a wide timeframe, without potential influence from vaccines reducing case-fatality rate 149 and the addition of major COVID-19 variants. Our study builds on the efforts of previous studies by presenting 150 together a wide array of variables that describe community characteristics. It is necessary to emphasize that the 151 associations between these factors and case-fatalities is not necessarily or entirely causative. All comorbidities, 152 sociodemographic and determinants of health variables presented describe characteristics of the population 153 that are not isolated or independent. Therefore, the causality that could be attributed to each factor evaluated 154 must always be provided with context as a community indicator as they are all dependent or interconnected on 155 each other, examples of this are binge drinking, mammography and visits to the dentist rates, which are likely 156 indirectly describing a characteristic of the community. In general, all these variables must be interpreted in a 157 continuum of causality that can vary across regions depending on the context. 158

Chronic kidney disease rates were the strongest predictor of increase COVID-19 case-fatality in several 159 US regions. This relationship is likely to be predominantly causal due to people with this condition being 160 medically vulnerable 25-28 . Other comorbidities followed a similar trend such as higher rates of hypertension, 161 diabetes, stroke, and smoking being associated with an increased case-fatality rate in a specific county. This 162 finding aligns with other studies 5,29,30 linking increased rates of comorbidities to poorer COVID-19 outcomes. 163

Even though comorbidities were most often associated with worse COVID-19 outcomes, stroke and cancer 164 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101/2022.05.20.22275397 doi: medRxiv preprint Page 6 of 15 diagnosis were linked to reduced case-fatality rates in the Northeast region. We speculate that a potential 165 explanation for this relationship is possible more frequent mask usage 31 and more precautions taken by this 166 group of people 32 . Northeastern states had mask adherence rates greater than 75% during the latter part of 167 2020 31 with usage potentially diminishing the influence of comorbidities such asthma and COPD on COVID-19 168 case-fatality rate. Those with asthma and COPD in these communities were maybe more likely to wear a mask, 169 further reducing their chance of acquiring and succumbing to Household size was identified early on to be a risk factor for COVID-19 transmission 33 . Our results have 171

shown a contrasting effect that supports this notion in the Northeast but has the opposite effect in the Midwest. 172

The difference between these regions is likely related to the specific context of living conditions. Although the 173 mean household size for regions 1 and 2 is not far from the mean household size for regions 5 and 7, with 2.48 174 and 2.41 respectively, this difference may capture differences in housing quality and composition 34 . This may 175 also be indirectly identifying behavioral factors that are not obvious but can be implied such as house proximity 176 (higher in cities) which can affect the capacity of self-isolation. Population density could be partially be 177 influenced by household size, which has shown to have an impact on transmission 35 . The Northeast states have 178 higher population densities when compared to midwestern states. In summary, living conditions, housing quality 179 and composition, and population density all be important component that define the impact of household size 180 on case fatality rate. Generally, higher income was also associated with decreased COVID-19 case fatality rates. 181

There could be a multitude of reasons why a higher income may be beneficial. This could include not being 182 classified as an essential worker leading to being more likely to take time off or work from home 36 , or even 183 being able to live outside high-density population and compact housing areas 7,37 . Income is often reflective and 184 associated with racial discrepancies in COVID-19 outcomes 24, 38, 39 . In a study examining neighborhood median 185 income and COVID-19, when examining the neighborhoods, Black populations more often lived in 186 neighborhoods with a significantly lower median income ($35,000) whereas White populations more often lived 187 in wealthier neighborhoods ($63,000). Communities with lower median incomes are more often Medicaid 188 patients and have COVID-19 complications that require invasive tactics such as mechanical ventilation 40 . Income 189 inequalities have been strong predictors of higher case numbers and fatalities throughout this pandemic 14 . 190 The context that defines the influence of social dynamics on COVID-19 is complicated. Political affiliation 191 has been repeatedly evaluated as a potential factor influencing the pandemic's mortality 41,42 . This pattern is not 192 new, as Republican party affiliation was associated with decreased influenza vaccinations and the Democrat 193 party was associated with increased compliance 43, 44 . The politicization of pandemic response has continued into 194 the COVID-19 pandemic, with behaviors such as masking, social distancing, and vaccination being often divided 195 along party lines. Some studies have shown a decrease in pandemic preventive health measures among 196

Republicans while there has been increased adherence to public health recommendations among Democrats 197 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Page 7 of 15 45, 46 . Similarly, Republicans have been shown to have lower COVID-19 risk perception 47,48 , when compared to 198 other political parties which may influence their likelihood of contracting COVID-19. The behaviors exhibited by 199 each political party may influence the results of this study. Our study showed that voting for Joseph Biden in the 200 2020 presidential election was mildly associated with decreased case-fatality rate while voting for Donald Trump 201 was associated with increased rate. With Democrats being shown to be more likely to adhere to public health 202 guidelines, they may be less likely to acquire and perish from COVID-19 while the inverse is true for Republicans. 203

In December 2020, states with Republican governors had higher rates of cases, deaths, and positive tests than 204 states with Democrat governors 49 . This trend is evident in a similar approach using national data presented by 205 NPR in collaboration with researchers at John's Hopkins University where it was shown that voting Republican 206 also had a deleterious effect 50 . Rural and urban differences have been shown to play a major role in case-fatality 207 rate as well with rural counties having a higher case-fatality ratio than urban counties 4 . Rural voters are more 208 likely to vote Republican 51 andtherefore, the influence of politics in our findings may also be capturing 209 geographical differences. Rural areas tend to have worse health outcomes in general and have significantly less 210 access to care compared to their urban counterparts 52 These disparities add to the likelihood of developing 211 comorbidities and ultimately, poorer COVID-19 outcomes. The link between political affiliation and COVID-19 212 case fatality rate is far more complex than the individual candidates that people of a county voted for. Political 213 affiliation in our study is an indicator of underlying sociodemographic, health, and psychological trends that are 214 more causative rather than associative. 215

Our study utilized aggregate data on a per-county basis instead of individual patient data; therefore, it is 217 not possible to evaluate factors that contribute to COVID-19 case-fatality on a per case fashion which could help 218 avoid any erroneous generalizations of specific regions. Another limitation of using county level data is that 219 there is significant variability in the size and number of counties across the United States. Some counties may 220

have only a few hundred people, while other counties may have a few million and this may lead, to some extent, 221 representation bias. Future directions of this study include using these results to guide the evaluation of 222 individuals' factors that contribute directly to illness outcomes. 223

Our study evaluated a multitude of factors that may affect COVID-19 case-fatality rate. Unlike previous 225 studies that evaluated these factors separately, we performed a comprehensive analysis of all these variables 226 together where they interact amongst each other. We identified several unique regionally dependent and 227 independent relationships that highlighted the various factors that might influence COVID-19. Like other studies, 228 we determined that comorbidities and demographic factors together are strong drivers of COVID-19 case 229 fatalities. However, our study presents an assessment that puts them side to side for direct comparison. Our 230 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 17.

US Center for Disease Control. 2020 PLACES. Local data for better health county data 2020 release. 293

Published 2021. Accessed March 12, 2021 . https://chronicdata.cdc.gov/500-Cities-Places/PLACES-Local-294

Data-for-Better-Health-County-Data-20/swc5-untb/data 295 18.

US Census Bureau. Average Household Size and Population Density -County. Published 2020. 296 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101/2022.05.20.22275397 doi: medRxiv preprint Figure 2. Regional risk ratio estimates for comorbidity, sociodemographic and determinants of health in 392 association to COVID-19 case-fatalities. All models were performed independently by region and are adjusted 393 to CCVI. Sample sizes correspond to the number of counties in the analysis for each region. Estimates signs 394 indicate effect direction. Red boxes are significant to a Bonferroni level (470 tests Padj≤1.06E-04). Orange boxes 395 are significant to a 95% confidence level. 396 397 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 Alaska was excluded from the analysis because of differences in their county level reporting 402 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022 

COVID-19 and Cardiovascular Disease

COVID-19, Social Determinants Past, Present, and Future, and African Americans' 284

Association Between Income Inequality and 286

County-Level COVID-19 Cases and Deaths in the US

White House COVID-19 Team

Vulnerable Communities and COVID-19: The Damage Done

CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 21, 2022. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All models are CCVI adjusted. A) Risk ratio raw estimates B) Risk ration standardized estimates (ordered). Blue 387 diamonds indicate case-fatality reduction, red diamonds indicate case-fatality increase, and Black diamonds 388 indicate no association. A total of 3140 counties were included in the study. Even when 99% CI are presented, 389 association are declared significant at a Bonferroni adjusted threshold (47 tests Padj≤1.06E-03). 390 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)The copyright holder for this preprint this version posted May 21, 2022. ; https://doi.org/10.1101 https://doi.org/10. /2022