key: cord-0840264-7c92u3im
authors: Tsai, Thomas C.; Jacobson, Benjamin H.; Orav, E. John; Jha, Ashish K.
title: Association of community-level social vulnerability with US acute care hospital intensive care unit capacity during COVID-19
date: 2021-12-22
journal: Healthc (Amst)
DOI: 10.1016/j.hjdsi.2021.100611
sha: 1b4b61e023f24d4878a459fa4818b1e7d4932411
doc_id: 840264
cord_uid: 7c92u3im

The COVID-19 pandemic has placed unprecedented stress on US acute care hospitals, leading to overburdened ICUs. It remains unknown if increased COVID-19 ICU occupancy is crowding out non-COVID-related care and whether hospitals in vulnerable communities may be more susceptible to ICUs reaching capacity. Using facility-level hospitalization data, we conducted a retrospective observational cohort study of 1753 US acute care hospitals reporting to the US Department of Health and Human Services Protect database from September 4, 2020 to February 25, 2021.63% of hospitals reached critical ICU capacity for at least two weeks during the study period, and the surge of COVID-19 cases appeared to be crowding out non-COVID-19-related intensive care needs. Hospitals in the South (OR = 3.31, 95% CI OR 2.31–4.78) and West (OR = 2.28, 95% CI OR 1.51–3.46) were more likely to reach critical capacity than those in the Northeast, and hospitals in areas with the highest social vulnerability were more than twice as likely to reach capacity as those in the least vulnerable areas (OR = 2.15, 95% CI OR 1.41–3.29). The association between social vulnerability and critical ICU capacity highlights underlying structural inequities in health care access and provides an opportunity for policymakers to take action to prevent strained ICU capacity from compounding COVID-19 inequities.

With rising novel 2019 coronavirus disease cases over the course of Fall and Winter 23 2020, hospitals increasingly became overburdened, leading to a critical shortage of intensive 24 care unit (ICU) capacity. The ability of acute care hospitals to deliver timely, effective care is 25 fundamental not just for patients with COVID-19 but for everyone else who relies on hospital 26 care, and there is growing concern that escalation in COVID-19-related ICU admissions may be 27 crowding out non-COVID-19 ICU care. 28 29 Tracking real-time data on hospital ICU care is critically important but has largely been 30 unavailable at the individual hospital level, hampering our ability to fully understand how 31 effectively our hospitals are able to manage the influx of patients. Prior studies have focused 32 primarily on modeled projections of anticipated hospital capacity. 1,2 For the first time since the 33 start of the pandemic, detailed hospital-level data on hospital capacity became available with 34 the public release of the HHS Protect data on December 7, 2020. This opened up the possibility 35 of assessing trends in ICU occupancy and the factors associated with hospitals exceeding a 36 margin of safety and reaching critical occupancy. There is reason to believe that capacity and 37 strain will vary across hospitals. It may be that some types of hospitals-such as for-profit 38 hospitals-are less likely to postpone elective surgeries, thus risking higher levels of ICU 39 utilization. Additionally, there has been a disproportionate burden of COVID-19 infections on 40 racial minority and economically vulnerable populations. [3] [4] [5] This disparity in infections has 41 resulted in age-adjusted rates of COVID-19 hospitalization that are approximately four-fold 42 higher for racial minority groups. 6 Already stressed hospitals located in communities serving a 43 socially vulnerable population therefore may be even more likely to exceed ICU capacity, 44 compounding underlying structural disparities in health care outcomes. Empirical data on the 45 types of hospitals most likely to exceed capacity would be immensely helpful for policymakers 46 to direct resources towards communities and hospitals at risk while mitigating disparities in 47 COVID-19 care. 48 49 Using hospital-level ICU capacity data, we sought to answer three pressing policy questions. 50

First, we examined trends and variation in ICU capacity during the third phase of the pandemic 51 (Winter 2020-2021) across US hospital markets. Second, we examined the structural features of 52 hospitals associated with reaching critical ICU capacity. Finally, we examined the relationship 53 between community-level vulnerability and likelihood of hospitals reaching critical ICU capacity. 54

We hypothesized for-profit hospitals, due to financial incentives, and hospitals in communities 55 with high social vulnerability would be more susceptible to reaching or exceeding critical ICU 56 capacity during the third phase of the COVID-19 pandemic. 57 58

Data 60

Data for this study covered the period from September 4, 2020 to February 25, 2021, and data 61 were accessed on March 9, 2021. The HHS Protect Public Data Hub consolidates multiple data 62 sources from HHS and CDC to create a dashboard of hospital capacity in the US in response to 63 -19. 7 Additionally, HHS requires all hospitals licensed to provide 24-hour care to report 64 data to the HHS Protect Effort. Initial dashboards were only at the state level, and due to 65 delays, a hospital-level public dataset was not released until December 7, 2020. The HHS 66 J o u r n a l P r e -p r o o f Protect hospitalization data file was merged with the 2018 American Hospital Association (AHA) 67

Annual Survey using hospital CMS certification numbers (CCN) to obtain hospital-level 68 structural features. The Area Health Resource File landscape file was then merged to the 69 analytic dataset using county FIPS codes to obtain county-level socioeconomic variables. 8 70 Hospitals in the top decile of the disproportionate share index were defined as safety net 71 hospitals. County-level Social Vulnerability Index (SVI) categorizations from the CDC were also 72 linked to hospital data. 9 Data points smaller than 4 were suppressed by HHS and these hospital-73

week observations were excluded from analysis. 105 of the 4,450 hospitals in the dataset did 74 not have observations for all 25 weeks since September 1, 2020 and were excluded. Children's 75 hospitals and long-term care hospitals were also excluded. 76 77

Our main dependent variable is a hospital reaching critical ICU capacity, which was defined as a 79 hospital's ICU occupancy exceeding 90% capacity for at least two weeks since September 4. We 80 chose overall ICU occupancy for both COVID and non-COVID-related admissions as our main 81 dependent variable as it reflects the effect of the burden of COVID-19 on the overall ability of 82 acute hospitals to deliver timely and effective care for the sickest patients. Overall ICU 83 occupancy has also emerged as an important gating criterion for re-institution of 84 nonpharmacologic interventions such as restaurant closures or stay-at-home measures, and 85 states including California have mandated cancellation of elective procedures when ICU 86 occupancy exceeds the 90% threshold. 10 Hospital occupancy for each week was determined by 87 dividing the number of occupied ICU beds by the number of staffed ICU beds. 88 J o u r n a l P r e -p r o o f Hospital-level variables explored for association with reaching critical ICU capacity included 90 hospital size (less than 100 beds is small, 100-399 beds is medium, 400 or more beds is large), 91 teaching status (non-teaching, major teaching, or minor teaching), profit status, safety net 92 status, nurse-bed ratio, number of intensivists, number of operating rooms, and region. Safety-93 net was defined as the top decile of the disproportionate share (DSH) index which was obtained 94 from the 2017 CMS Healthcare Cost Report Information System (HCRIS). Regions were divided 95 using US Census Bureau categorizations as follows: Northeast, Midwest, South, and West. 96

Our main measure of community-level social vulnerability was the Centers for Disease Control 98

and Prevention (CDC) Social Vulnerability Index (SVI). The SVI assesses the ability of 99 communities to respond to wide-scale hazardous events like natural disasters or disease 100 outbreaks. The SVI represents a percentile ranking across US census tracts and is based on 15 101 social factors across the domains of socioeconomic status, household composition, 102 race/ethnicity/language, and housing/transportation (Supplemental Methods 1). SVI was 103 aggregated at the county level, and these percentile rankings were categorized into quartiles. 104 105

The unit of analysis was acute-care hospitals with unsuppressed ICU capacity data reported to 107 HHS Protect over our study time frame. We first compared the characteristics of the HHS 108

Protect sample to the overall sample of acute care hospitals in the AHA Annual Survey. 109

Statistical testing of hospital characteristics was performed with chi-squared tests. ICU 110 occupancy and capacity were aggregated across all hospitals and stratified by week and by 111 region to determine national and regional trends of critical ICU capacity. Total and available ICU 112 beds were aggregated within each Hospital Referral Region (HRR), as this geography represents 113 a hospital market within which patients may be rebalanced between hospitals to prevent 114 exceeding capacity. Given this rebalancing, we set the threshold for HRR critical capacity at 70% 115 occupancy, based on observations that while hospitals often saw occupancy around 70% prior 116 to the pandemic, few HRRs exceeded this 70% threshold (Supplemental Figure 1) . We 117 calculated the number of weeks during the third wave of the pandemic that each HRR exceeded 118 70% occupancy and displayed this data as a heatmap. We also aggregated capacity and 119 occupancy data at the state level and ranked both HRRs and states by occupancy in the most 120 recent week of data (February 19 to February 25). 121

We next assessed the relationship between hospital-and community-level factors and a 123 hospital reaching critical ICU capacity. Mean values of hospital-and community-level factors 124 were calculated independently for ICUs that did and did not exceed 90% occupancy for at least 125 two weeks since September 1. In order to assess the relationship between hospital-level factors 126 and a hospital reaching critical ICU capacity, we first compared bivariate relationships. We then 127 created two multivariate logistic regression models to assess the relationship between hospital-128 level factors and a hospital reaching critical ICU capacity. The first model assessed the 129 relationship between hospital structural features and the likelihood of a hospital reaching 130 critical ICU capacity. Hospital features included hospital-level variables as described above. 131 Given the regional waves of the pandemic, in the multivariate hospital characteristics model we 132 J o u r n a l P r e -p r o o f included regional fixed effects to assess the relationship between hospital structural features 133 and critical ICU capacity within a region. We then created an additional multivariate logistic 134 regression model to assess the relationship of community-level vulnerability with critical ICU 135 capacity, controlling for hospital structural features and regional fixed effects. 136 137

As a sensitivity analysis, we modeled the association between social vulnerability and critical 139 capacity at the county and HRR level, given concerns that a hospital-level model does not 140 capture the redistribution of patients within larger markets to avoid a specific hospital 141 exceeding capacity (Supplemental Table 4 ). The number of staffed ICU beds and occupied ICU 142 beds was summed for all hospitals within each HRR in each week to determine critical capacity, 143

and HRRs were considered to have reached critical capacity if they exceeded 90% ICU 144 occupancy for at least two weeks. In the county-level model, each county was assigned the 145 critical-capacity status of the HRR in which it was located. In the HRR-level model, each HRR 146 was assigned a social vulnerability score by calculating a population-weighted average of the 147 SVI scores of each county in the HRR. County and HRR multivariable logistic regressions 148 contained both social vulnerability and region, but not other hospital structural features as 149 hospital-specific metrics like for-profit or safety-net status cannot necessarily be aggregated at 150 larger geographies. We also assessed the relationship of each individual domain of the SVI by 151 creating multivariate logistic regression models using each of the four sub-measures of SVI 152 rather than the composite measure (Supplemental Table 5 ) and controlling for hospital 153 structural features. We also performed a mediation analysis by sequentially adding in 154 J o u r n a l P r e -p r o o f covariates for hospital characteristics in addition to SVI (Supplemental Table 6 From September 2020 to February of 2021, ICU occupancy increased across the US, though the 166 upward trend was most substantial in the South, Midwest, and West (Figure 1) the South and increasing at more than twice that rate in other regions. Total ICU occupancy 179 increased by considerably less in each region, and Figure 1 Jersey dipping below 50% occupancy(Supplemental Table 2 ). 192

Of the 1,753 hospitals with ICU occupancy data, 63% (1100) reached critical ICU capacity for at 194 least two weeks since September, and these hospitals differed from those that did not reach 195 critical capacity on a number of county-and hospital-level factors (Supplemental Table 3 and controlling for the region, only the regional effect remained significant, suggesting the 202 widespread regional effect of community-level transmission of COVID-19. In adjusted analyses, 203

there was a statistically significant regional effect, with the South having over three times the 204 odds (adjusted OR=3.31, 95% CI OR 2.31-4.78, p<0.001) of reaching critical capacity as those in 205 the Northeast. Hospitals in the West also had increased odds (adjusted OR=2.28, 95% CI OR 206 1.51-3.46, p<0.001) of reaching critical capacity relative to those in the Northeast. 207 208 Social vulnerability was strongly associated with the odds of hospitals reaching critical ICU 209 capacity in both unadjusted and adjusted models. Adjusting for hospital structural features and 210 region, hospitals located in areas of the highest vulnerability had more than twice the odds of 211 reaching critical ICU capacity as those in less vulnerable areas (adjusted OR 2.15, 95% CI OR 212 1.41-3.29, p<0.001). When assessing critical ICU capacity at larger geographies, both counties 213 and HRRs had higher odds of reaching critical ICU capacity when in more socially vulnerable 214 areas (Supplemental Table 4 ). 215 216 When examining individual components of social vulnerability, minority status and 217 socioeconomic vulnerability had the largest influence on odds of reaching critical capacity, 218 while household composition and housing type were not significantly associated with critical 219 capacity (Supplemental Table 5 ). 220

When sequentially adding additional control variables for hospital characteristics, odds of 222 critical ICU capacity decreased for hospitals located in the highest vulnerability counties, but 223 patterns were generally similar to our main results (Supplemental Table 6 ). 224 225

In this analysis of official national acute-care hospital ICU occupancy data, approximately 3 in 5 227 hospitals were overburdened and reached critical ICU capacity, exceeding 90% occupancy for at 228 least two weeks during the third phase of the COVID-19 pandemic. While hospital occupancy 229 grew in all regions over the course of the study period, hospitals in the South and West were 230 most likely to reach critical levels of ICU capacity. Most importantly, hospitals located in 231 communities of the highest social vulnerability had more than twice the odds of being 232 overburdened. Taken together these findings highlight the disproportionate impact of COVID-233 19 cases on vulnerable communities, and in turn pose a concern that overburdened acute care 234 hospitals may potentially be widening disparities of vulnerable populations by compounding 235 structural inequities in access to medical care. 236 237 Other than regional variation, the strongest predictor of overburdened hospital ICUs was the 238 county-level Social Vulnerability Index. Adjusting for age, racial and ethnic minority groups 239

including Indigenous, Black, and Hispanic populations have approximately four-fold higher rates 240 of hospitalization with COVID-19 compared to White, non-Hispanic persons. 6 This 241 disproportionate burden is due in part to a higher prevalence of comorbidities as well as to 242 socioeconomic determinants of COVID-19 exposure including household composition and 243 occupational exposure. 11,12 In a study of Massachusetts municipalities, the proportion of 244 foreign-born noncitizens in a community, mean household size, and share of food service 245 workers were strongly associated with higher COVID-19 rates. 3 Additionally, socially vulnerable 246 communities have been shown to have higher rates of acute care hospital readmissions, 13,14 247 and the higher baseline need for acute care hospital services may serve as a second hit, placing 248 already overburdened hospitals in these communities in a position to more likely reach 249 potentially unsafe levels of ICU capacity during the pandemic. 250

There is a growing body of literature assessing the relationship between the SVI and COVID-19 252 case trajectory, suggesting that racial and ethnic disparities in COVID-19 cases may in large part 253 be due to structural inequities in underserved communities. 4,15 Our study updates and extends 254 this work by directly assessing the consequences of social vulnerability on hospitalizations and 255 ICU capacity, which reflect both the underlying burden of COVID-19 as well as the inequitable 256 access to hospital-based acute care in socially vulnerable areas. 16 As ICU capacity has become a 257 major determinant of COVID-19 risk-level for phased re-opening after state-mandated stay-at-258 home measures and lockdowns, our findings have immediate relevance for policymakers to 259 target both pandemic response in the form of testing and vaccinations as well as economic 260 relief to vulnerable communities suffering from the perfect storm of high COVID-19 cases, 261 overburdened hospitals, and economic shutdown. The HHS Protect hospitalization dataset suppresses small capacity and occupancy numbers, 282 which leads to a substantial underrepresentation of rural and critical access hospitals. Our 283 findings may underestimate the impact of rurality on hospital occupancy, and we were unable 284 to analyze whether critical access hospitals had higher odds of reaching critical ICU capacity. 285

Policymakers at HHS should release the full ICU and hospital occupancy data on rural and 286 critical access hospitals, which are currently suppressed, to be able to ascertain the full 287 consequences of overcrowded hospitals for rural communities. This study was conducted 288 during the third wave of the pandemic during the Winter of 2020-2021, and our findings may 289 not reflect patterns observed during later stages of the COVID-19 pandemic. 290 291

In conclusion, we find a substantial shortage of ICU capacity in the US, with approximately 3 in 5 293 hospitals in the US exceeding 90% occupancy for at least two weeks during the current wave of 294 the COVID-19 pandemic. This growing crisis of overburdened ICUs has been driven by surging 295 COVID-19 cases and has led to a reduction in non-COVID ICU care. Hospitals located in areas of 296 high social vulnerability had more than twice the odds of being overburdened. Policymakers 297 should ensure equitable access to hospital care by vulnerable populations during the pandemic. 298

These findings can inform future pandemic preparedness and response efforts to support both 299 vulnerable populations and the hospitals that serve them. 300 

Estimated Demand for US Hospital 302 Inpatient and Intensive Care Unit Beds for Patients With COVID-19 Based on Comparisons With 303 Wuhan and Guangzhou, China

Projecting hospital utilization during the 306 COVID-19 outbreaks in the United States

Associated With Racial And Ethnic Disparities In COVID-19 Rates In Massachusetts

Association of Social Vulnerability with COVID-19 Cases and Deaths in the USA

Variation in COVID-19 Hospitalizations and 315 Deaths Across New York City Boroughs

8. Area Health Resources Files

CDC/ATSDR Social Vulnerability Index Database

Prevention/Agency for Toxic Substances and Disease Registry/Geospatial Research, Analysis, 326 and Services Program

Order of the State Public Health Officer -Hospital Surge 1/5/2021. In: 329 California Department of Public Health, editor. California2021

US-county level variation in intersecting individual, household 331 and community characteristics relevant to COVID-19 and planning an equitable response: a 332 cross-sectional analysis

Thirty-day readmission rates for Medicare beneficiaries by 338 race and site of care

Disparities in surgical 30-day readmission rates for Medicare 340 beneficiaries by race and site of care

Association of Social and Demographic Factors With 343 COVID-19 Incidence and Death Rates in the US

Income Disparities In Access To Critical Care 346

Health Aff (Millwood)

Critical Supply Shortages -The Need for Ventilators and 348 Personal Protective Equipment during the Covid-19 Pandemic

forced major manufacturers to build ventilators. Now they're piling 351 up unused in a strategic reserve

The effect of nurse-to-patient ratios on nurse-354 sensitive patient outcomes in acute specialist units: a systematic review and meta-analysis. Eur 355 J Cardiovasc Nurs

Nurse staffing and patient outcomes in critical care: a concise review

Non-Teaching 2943 (68.9%)