COVID-19 health inequities and association with mechanical ventilation and prolonged length of stay at an urban safety-net health system in Chicago

Millions of Americans have been infected with COVID-19 and communities of color have been disproportionately burdened. We investigated the relationship between demographic characteristics and COVID-19 positivity, and comorbidities and severe COVID-19 illness (use of mechanical ventilation and length of stay) within a racial/ethnic minority population. Patients tested for COVID-19 between March 2020 and January 2021 (N = 14171) were 49.9% (n = 7072) female; 50.1% (n = 7104) non-Hispanic Black; 33.2% (n = 4698) Hispanic; and 23.6% (n = 3348) aged 65+. Overall COVID-19 positivity was 16.1% (n = 2286). Compared to females, males were 1.1 times more likely to test positive (p = 0.014). Compared to non-Hispanic Whites, non-Hispanic Black and Hispanic persons were 1.4 (p = 0.003) and 2.4 (p<0.001) times more likely, respectively, to test positive. Compared to persons ages 18–24, the odds of testing positive were statistically significantly higher for every age group except 25–34, and those aged 65+ were 2.8 times more likely to test positive (p<0.001). Adjusted for race, sex, and age, COVID-positive patients with chronic obstructive pulmonary disease were 1.9 times more likely to require a ventilator compared to those without chronic obstructive pulmonary disease (p = 0.001). Length of stay was not statistically significantly associated with any of the comorbidity variables. Our findings emphasize the importance of documenting COVID-19 disparities in marginalized populations.


Introduction
As of April 5, 2021, there have been over 30.5 million cases of COVID-19 and over 550,000 COVID-19 associated deaths reported in the United States (US) [1]. The clinical manifestation of COVID-19 is variable ranging from asymptomatic to life-threatening [2]. Yet, thus far, a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Sinai Chicago includes Mount Sinai Hospital (MSH), Holy Cross Hospital (HCH), and Schwab Rehabilitation (SRH). MSH is a 319-bed acute care community teaching hospital and an adult Level 1 trauma center; HCH is a 264-bed community hospital; SRH is a 102-bed rehabilitation hospital which serves patients that have experienced spinal cord injuries, stroke, brain injury, amputation, and other musculoskeletal and neurological injuries. MSH and HCH have Emergency Departments (ED) and all three facilities offer a range of inpatient and outpatient services. Sinai Chicago also operates 14 clinics across the system's service area and includes a range of services including COVID-19 testing.
The majority of persons in the Sinai Chicago service area, as well as patients seen at Sinai Chicago, are un-or under-insured, non-Hispanic Black or Hispanic, and many are Spanishspeaking only [16].

Study design
We conducted a cross-sectional study of all Sinai Chicago patients 18 years of age and older who were tested for COVID-19 between March 1, 2020 and January 31, 2021 in inpatient and outpatient locations, as well as the ED. The Mount Sinai Hospital Institutional Review Board (FWA#00005088) approved this project (protocol #20-12). Because all of the patients had been discharged at the time of analysis, we were not able to obtain participant consent and the Mount Sinai Hospital Institutional Review Board granted our request for a waiver of HIPAA authorization for research.

Data collection
Data for this study were abstracted from Meditech, Sinai Chicago's electronic medical record (EMR).

Explanatory variables
Explanatory variables included sex, age, race/ethnicity, and patient comorbidities. Sex was a dichotomous variable coded 1 for patients for whom the EMR indicated male sex. Age was a categorical variable coded to groups: 18-24 (referent group), 25-34, 35-44, 45-54, 55-64, and 65+ and calculated as the difference in years between patient date of birth and date of COVID test. Race/ethnicity was a categorical variable coded to groups based on documentation in the EMR: non-Hispanic White (referent group), non-Hispanic Black, Hispanic, non-Hispanic Asian, and Other/Unknown. For patients missing ethnicity data in the EMR, we undertook the following process to recode data. First, of the total records (N = 14171) there were 1234 with Hispanic ethnicity identified in the EMR. For the remaining records (n = 12937), we first applied the surname recode process, which entailed checking the patient's last name against the 1990 US Census list of the 639 most frequently occurring heavily Hispanic surnames [17] and recoding patients with these last names to Hispanic ethnicity (n = 3149). Second, we manually searched for and evaluated last names with the letters "ez", "ll", or "rr" occurring [17] (section 7.1.3) and recoded these to Hispanic ethnicity (n = 107). Finally, for anyone not yet recoded to Hispanic ethnicity, we used the EMR indicator for "patient primary household language is Spanish" to recode to Hispanic ethnicity (n = 208). This brought our total number of records with Hispanic ethnicity to 4698.
Two comorbidity variables were employed: total comorbidities and the Center for Disease Control's (CDC) three-level comorbidity categorization [18]. Total comorbidities was a categorical variable coded to groups: 0 (referent group), 1, 2, 3-5, and 6-9 and was the sum for patients for whom the EMR indicated an ICD-10 code for cancer, CKD, COPD, cardiovascular disease, obesity, smoking & tobacco use, type 2 diabetes, asthma, hypertension, cerebrovascular disease, HIV, type 1 diabetes, and chronic respiratory conditions (chronic respiratory failure and chronic bronchitis).
We then used the CDC's recommendations for categorizing these comorbidities into those that have varying levels of evidence for causing severe COVID-19 illness: strongest and most consistent, mixed, and limited evidence. Strongest and most consistent evidence was defined as consistent evidence from multiple small studies or a strong association from a large study. Mixed evidence was defined as multiple studies that reached different conclusions about risk associated with a condition. Limited evidence was defined as consistent evidence from a small number of studies. The strongest evidence category included: cancer, CKD, COPD, cardiovascular disease, obesity, smoking & tobacco use, and type 2 diabetes. The mixed evidence category included: asthma, hypertension, and cerebrovascular disease. The limited evidence category included: HIV, type 1 diabetes, and chronic respiratory conditions [18].

Outcome variables
The three primary outcome variables of this study were COVID positivity (positivity), COVID-related ventilator use (VENT), and COVID-related length of hospital stay [15].
Positivity was a dichotomous variable coded 1 for patients for whom the EMR indicated a positive COVID test result. Sinai Chicago conducts the SARS-CoV-2 PCR, Rapid, and IgG qualitative tests, which were all included in this analysis. A patient with a positive result for any one or more of those tests was considered positive for COVID-19 (positive). Patients with multiple positive tests were counted only once and for patients with both a positive and negative result, only the positive result was retained in the analysis dataset.
VENT was a dichotomous variable coded 1 for patients for whom the EMR indicated a physician ordered mechanical ventilation of the patient. LOS was a dichotomous variable coded 1 for patients with a length of stay of 3 or more days. A 3-day cutoff point was used as a proxy for severe infection based on the Adaptive COVID-19 Treatment Trial's (ACTT-1) determination that the benefit of Remdesivir, a therapeutic treatment for COVID-19, is most pronounced among patients who have been hospitalized at least 3 days [19]. Because our sample contained patients tested for COVID-19 at various points during their stay, and in order to ensure our length of stay analysis was most closely related to the COVID-19 infection, we calculated length of stay as the difference between the day the patient was tested for COVID-19 and the day the patient was discharged. This continuous variable was then dichotomized as described above for all patients with non-zero and non-negative values. This was only calculated for patients who were admitted and excluded those who expired during their hospital stay.

Statistical analysis
Our first aim was to assess whether positivity differed by sex, race/ethnicity, and age. We performed analysis of variance (ANOVA) tests for each demographic characteristic to determine whether differences in positivity were statistically significant between groups.
Our second aim was to assess whether VENT and LOS were associated with comorbidities. We performed logistic regression to calculate odds ratios (OR) for VENT, LOS, total comorbidities, and the strong, mixed, and limited evidence comorbidity variables. Adjusted models include sex, race/ethnicity, and age as control variables. This analysis was restricted to patients with a positive COVID-19 result and additionally, for LOS, the analysis was restricted to patients who did not expire during their COVID-19-related hospital stay.

Co-morbidities of positive patients
Among the 2286 positive patients, the most prevalent comorbidities were hypertension (26.8%), type 2 diabetes (23.6%), obesity (10.6%), asthma (7.8%), CKD (7.7%), and COPD (7.3%) ( Table 2). Compared to positive males, positive females had a higher prevalence of: COPD, obesity, type 2 diabetes, asthma, hypertension, and chronic respiratory conditions. Positive females also had a higher mean (1.1) total comorbidities compared to positive males (0.8). Compared to positive patients of all other race and ethnicities, non-Hispanic Black positive patients had a higher prevalence of: CKD, COPD, cardiovascular disease, asthma, and chronic respiratory conditions compared to all other race/ethnic groups. Non-Hispanic Black patients (1.3) also had a higher mean total comorbidities compared to positive non-Hispanic  (Table 3). In the adjusted models, only COPD remained significant such that patients with COVID-19 and COPD were 1.9 times more likely to require a ventilator compared to positive patients without COPD (p = 0.001). We also found that in the unadjusted models, ventilator use is statistically significantly associated with having 3-5 and 6-9 total comorbidities, but once adjusted for our covariates, this association is no longer present and the total number of comorbidities is not associated with a higher odds of ventilator use.

Length of stay
Among positive patients who were admitted to the hospital, stayed at least 1 day, and did not expire during their stay, 1033 (77%) had a length of stay � 3 days. A larger proportion of males (79.6%) and patients aged 65+ (82.7%) had a length of stay � 3 days. Length of Stay was not statistically significantly associated with any comorbidity variables. For total comorbidities, we observed an increased odds with each increment of total comorbidities, but while in the expected direction, these results were not statistically significant. None of the unadjusted models for the strong, mixed, and limited evidence comorbidity variables were statistically significant.

Discussion
This study demonstrates the significant disparities in COVID-19 positivity by sex, race/ethnicity and age at a large safety-net hospital system on the south and west sides of Chicago. It is also among the first to present findings on the relationship between comorbidities and two outcomes indicative of severe COVID-19 infection: mechanical ventilation and prolonged length of stay. There are a number of key findings from our analysis. First, our results support the vast disparities in COVID-19 positivity among non-Hispanic Black and Hispanic populations. Similar to other findings in urban populations, our study found that non-Hispanic Black and Hispanic patients have a statistically significantly higher odds of positivity (1.5 and 2.5 respectively) [20][21][22]. Evidence suggests that Hispanic ethnicity is often underreported in healthcare settings, therefore even with our surname recode methodology described above, it is possible this is an underestimate of the impact on the Hispanic community [23]. Some of these disparities can be explained by the long-term disinvestment in these communities which lead to high rates of poverty, poor access to quality healthcare, and distrust of the medical  [24,25]. Of relevance to the COVID-19 pandemic, non-Hispanic Black and Hispanic persons are more likely to have essential jobs that require them to attend work, and increase their potential for exposure, even when stay-at-home orders are in place [26]. In addition, communities of color are more likely to live in inadequate housing conditions which may lead to crowding and prevent adequate social distancing within multi-generational homes [27]. While non-Hispanic Black and Hispanic patients had a statistically significantly higher odds of positivity compared to patients of other races, they also experienced different rates of positivity from each other. The positivity rate for Hispanic patients was almost twice that of non-Hispanic Black patients (14.0% vs. 21.6%). This may be due to the significant barriers Hispanic patients are more likely to experience to access testing, such as lower rates of insurance coverage, language barriers, and farther distances to testing sites [28].
In addition to racial disparities, our study also explored the impact of specific comorbidities, and total number of comorbidities, on two outcomes of interest: mechanical ventilation and COVID-specific length of stay. Our results suggest that even after adjusting for sex, race/ ethnicity, and age, positive patients with a history of COPD have 1.9 times higher odds of requiring mechanical ventilation, an indicator of severe infection. In addition, although not statistically significant, we found that as the total number of comorbidities increases, a patient's odds of requiring mechanical ventilation also increases.
In summary, our findings emphasize the importance of documenting COVID-19 disparities in marginalized populations to understand the profound impact in these communities and plan for equitable vaccine distribution. Whether or not herd immunity is achieved, all individuals should continue to practice COVID-19 safety precautions such as social distancing and masking. This is particularly true of individuals found to be at the highest risk of severe illness: those with COPD.

Limitations
While this study only includes data from one health system in Chicago, the sample is unique in that it captures a predominantly racial/ethnic minority population from the largest private safety-net health system in Illinois. Very few studies to date have been able to explore COVID-19 disparities among racial/ethnic minorities, and of note, this is the first study to go a step further and explore the relationship between comorbidities and severe illness as measured by mechanical ventilation and length of stay. In addition, as described above, race/ethnicity data for Hispanic individuals is historically underreported and we faced similar challenges. However, the surname recode methodology utilized in our data cleaning process allowed us to identify almost 3500 additional Hispanic patients, leading to a better estimate of the impact of COVID-19 in this population. The data analyzed for this study did not include individual-level data on risk factors for COVID-19 infection or barriers to testing. However, patients served come from communities characterized by considerable racial inequities in COVID-19 prevalence, mirroring what we see in our results [29]. We selected comorbidities for inclusion in this analysis based on the CDC's recommendations at the time of analysis. We acknowledge that the literature is constantly changing as we learn more about which comorbidities put patients at higher risk of severe COVID-19 illness. It is also important to note that we did not adjust for having well-controlled comorbidities, either through lifestyle modification or medication. Recent literature suggests that uncontrolled diseases such as hypertension and diabetes are associated with poorer health outcomes [30][31][32]. Additionally, our comorbidity data may not be comprehensive given that patients may seek care at multiple facilities throughout the city or may not seek care at all. As stated above, the variable for mechanical ventilation was coded based on a physician order for mechanical ventilation in the EMR. We did not do a comprehensive chart review to determine if the patient was truly ventilated. Finally, it is important to acknowledge that treatment protocols changed over the course of our study period; therefore it is possible that patients from the early months of the pandemic may have had worse outcomes than if they had access to treatments that became the standard of care as our understanding evolved.

Future research
While this study presents strong data to suggest associations between comorbidities and outcomes (mechanical ventilation and length of stay), future research would benefit from including larger datasets across various age ranges, sex, and race/ethnicity to allow us to better understand the disparities within different settings. This study did not include measures of symptomatology which is another opportunity for future research to explore how comorbidities are associated with symptoms of illness such as fever, dyspnea, fatigue, or new loss of taste or smell, as well as likelihood of death as a result of COVID. As of the writing of this paper, three vaccinations have been approved by the Food and Drug Administration for emergency use approval: BioNTech/Pfizer, Moderna, and Johnson & Johnson. While these are just beginning to be widely distributed in the population, future research should include an exploration of how vaccination may impact the progression of disease.