Disparities in demand for COVID‐19 hospital care in the United States: Insights from a longitudinal hierarchical study

Abstract Background and aims This study examined disparities in hospitalization for COVID‐19 within the U.S. by racial and ethnic groups, health insurance status, and social support structure. Methods Using publicly available ecological case and contextual data from July 2020 to April 2021, a longitudinal hierarchical model for the 51 U.S. states was constructed. Results Racial/ethnic disparities were observed, such as that hospitalization rates were higher in states with a higher percentage of Black (β = .002, P = .009) and American Indian or Alaska Native persons (β = .003, P = .03). Conversely, lack of health insurance was related to a lower hospitalization rate (β = −.005, P = .002), and so was a stronger social support system (β = −.015, P = .05). Conclusion These differences suggest disparities in COVID‐19 incidence, symptom severity, and demand for hospital care. Understanding how they contribute to geographic differences in hospitalization can help guide public health decisions and resource allocation to address COVID‐19‐related health inequalities.


| INTRODUCTION
First reports of pneumonia of unknown etiology emerged in Wuhan, China, on December 31, 2019. The extremely contagious virus was identified as SARS-CoV-2, causing the disease COVID-19. While it spread quickly around the globe, differences in the outbreak existed between and within countries. [1][2][3] The pandemic strained hospital resources, and necessitated the need for models forecasting patient care demands. Because a community's susceptibility to any virus is determined by a variety of factors, including biological determinants, demographic profiles, and contextual factors, 4 in a country as diverse as the U.S., there was likely considerable intra-country variation. While first progress in understanding disparities in hospitalization rates for COVID-19 by racial and ethnic groups has been made, [5][6][7][8] differences in demand for care by health insurance status and social support structure are not well understood. Using ecological data from 51 U.S. states during July 2020 to April 2021, this study constructed a longitudinal hierarchical model to assess the COVID-19 hospitalization rate in relation to confirmed cases by race/ethnicity, health insurance status, strength of the social support system, and other contextual factors.

| MODELS AND METHODS
COVID-19 case, mortality, and hospitalization data were obtained from USAFacts and the U.S. Department of Health and Human Services (HHS). 9 F I G U R E 2 Funnel plot for COVID-19 hospitalization rates at state level. In this funnel plot, every dot represents a U.S. state. Two states with more than 2 million cases (Texas, California) and the District of Columbia were excluded from the figure. Almost all states (except Vermont, Wyoming and Delaware) are outside of the 99% control limits (with Bonferroni correction; dashed line), that is, they are subject to special cause variation that is, multiple observations t are nested within a single entity I. 13 When there is longitudinal data with temporal autocorrelation, a failure to accommodate a nested data structure would lead to improper statistical inferences. 14,15 At level 1, the model compared hospitalization with COVID-19 cases and the pandemic threat level, which is operationalized with the number of deaths: Then, at level 2, the model controlled the association of cases and hospitalization for state-level variables: The level-2 variables were not modeled to influence π 0i , because cases and deaths are baseline changes; hospitalization does not fluctuate without changes in those level-1 variables.
Deaths as a covariate represents the danger associated with the pandemic, in concordance with terror management theory 16 and the terror management health model. 17 Grounded in evolutionary theory, terror management theory offers predictions for how people behave in response to fear associated with mortality. It acknowledges that humans are unique in their use of conscious thought processes, and proposes that humans exhibit a biological predisposition toward self-preservation and reproduction. 18 Humans change behavior in novel circumstances, with hospitalization being such a behavioral health response to a threat. All predictors were centered around the group mean at level 1, and grand mean at level 2. P-values <.05 were considered statistically significant. SPSS (version 27; IBM) and HLM (version 7.3; Scientific Software International) were used to conduct all analysis.

| Statistical results
The intraclass correlation coefficient in a one-way random effects ANOVA (unconditional) model was ICC = .943, that is, along the timeline more than 94% of the variation in the hospitalization was between states, and only about 6% was within states; the variation between states was statistically significant (r 0 = 108 240.019, P < .001). Likewise, the average COVID-19 hospitalization rate, calculated as a percentage of total cases during July 2020 to April 2021 and shown in Figure 1, ranged from 3.9% (Rhode Island) to 31.5% (Hawaii). Most states were subject to special cause variation, as indicated by the funnel plot in Figure 2. Thus, it was deemed prudent to proceed with the longitudinal hierarchical model as specified above; Table 1

| DISCUSSION
States with a higher percentage of Black and AIAN persons had a disproportionally higher hospitalization for COVID-19; these findings are consistent with previous studies showing disproportionate COVID-19 incidence, hospitalization, and mortality among those racial/ethnic groups. [5][6][7][8] The health of these groups is affected by long-standing, systemic inequities, including limited access to quality health care and education, over-representation in essential jobs with less work-from-home flexibility, and possibilities for medical leave. 8 Contrary to prior studies, the effect was not significant for Hispanic or Asian persons as well as for the median age. 7,21 The percentage of the rural population was not significant either.
The productivity of social capital, which comprises of all aspects of human relationships that produce benefits, is unequally distributed across the U.S.; it is important for understanding national challenges and creating effective policies. 12 This study showed that states in which people enjoy more social support through family, friends, and neighbors are associated with lower demand for hospitalization; a withering associational life increases the need for external hospital care. People without health insurance coverage generally have worse access to health care than people who are insured. 10  Effectively protecting the health of a country's population relies on having data to assess reasons for health disparities. Continued use of this study's model can help to identify health disparities experienced by certain racial/ethnic groups. Importantly, access to hospital care is also related to its affordability.
The findings in this report are subject to at least five limitations. Because this approach is associated with 85% reductions in hospitalization, differences in access to this therapy may account for some of the differences observed. 22 sources were not involved in any decision at any part of the study.

CONFLICT OF INTEREST
The author is not aware of any competing interests.

TRANSPARENCY STATEMENT
The author has read and approved the final version of the manuscript.
Wolfgang Messner had full access to all of the data in this study and takes complete responsibility for the integrity of the data and the accuracy of the data analysis. He affirms that this manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.

ETHICS STATEMENT
No ethics approval was sought, because no humans or animals participated in this subject. Ethics approval or informed consent requirement were therefore not deemed to be applicable.