Journal Logo

Clinical Investigations

Surge and Mortality in ICUs in New York City’s Public Healthcare System

Toth, Alexander T. BA1; Tatem, Kathleen S. MPH2; Hosseinipour, Nicole MPH2; Wong, Taylor BA1; Newton-Dame, Remle MPH2; Cohen, Gabriel M. MD3; George, Annie PhD, RN4; Sessa, Thomas BS5; Postelnicu, Radu MD1; Uppal, Amit MD1; Davis, Nichola J. MD2; Mukherjee, Vikramjit MD1

Author Information
doi: 10.1097/CCM.0000000000004972


Colder weather and the holiday season have led to record numbers of coronavirus disease 2019 (COVID-19) cases that threaten to overwhelm hospitals across the United States. During previous surges in the Northeast and Sun Belt, national COVID-19 hospitalizations reached peaks of around 60,000 patients, but as of December 2, 2020, over 100,000 people were hospitalized with COVID-19 in the United States (1). Further, as of December 12, 2020, one in eight U.S. hospitals had little to no available ICU space, and as record numbers of COVID-19 cases continue to be reported, experts predict the number of ICUs struggling to increase (2). The effects of surging patient numbers may be particularly harmful in critical COVID-19 patients, where optimal treatment relies on the availability of specialized spaces, staff, and equipment. These patients also often require prolonged care, which places further strain on limited resources (3,4). Modest ICU strain has been associated with mortality (5), but the impact of ICU surge strain on outcomes in critical COVID-19 patients is not yet well documented. As ICUs across the country face record occupancy, it is critical to understand how surges impact ICU mortality. We explore the experience of 11 ICUs in the largest public healthcare system in the United States during the initial surge of COVID-19 cases in New York City (NYC). We investigate the impact of ICU surge on mortality and explore differences in mortality risk based on clinical and sociodemographic factors.


Study Population

We conducted a retrospective observational study of adults with severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection requiring ICU-level care at one of NYC Health + Hospital’s (H+H) 11 acute care hospitals from March 24, to May 12, 2020 and discharged or deceased by June 12, 2020. SARS-CoV-2 infections were confirmed by real-time reverse transcriptase-polymerase chain reaction (RT-PCR) tests. Incarcerated patients, pregnant women and patients still admitted at study end were excluded. ICU care encompassed admission to traditional and “flex” ICU beds. Flex beds were in hospital spaces converted to ICU-capable environments. The study was approved by the Institutional Review Board (IRB) at Biomedical Research Alliance of New York, IRB number 20-12-193-373.

Data Collection

Demographic and clinical data were extracted from the electronic medical record (EMR), surge level was recorded daily, and American Community Survey (ACS) data provided neighborhood poverty levels.

Sociodemographic variables included gender, age, race/ethnicity, primary payer, marital status, primary spoken language, borough of residence, and neighborhood poverty level. Age was categorized as 18–44, 45–64, 65–74, and 75+ years. Gender was captured as male or female. Race and ethnicity were combined into a composite measure (Hispanic, non-Hispanic Black/African American, non-Hispanic White, non-Hispanic Asian/Pacific Islander, non-Hispanic other, declined/missing/unknown) (6). Primary payer was categorized into medicaid, medicare, commercial/other, and uninsured which included emergency medicaid, self-pay, and COVID-19 Health Resources and Services Administration uninsured/provider relief fund (7). Marital status included married/life partner, single, divorced/separated/widowed, or unknown. Primary spoken language was dichotomized into English and non-English. Patient zip codes determined borough of residence and neighborhood poverty, defined as low (0–9%), medium (10–19%), high (20–29%), and very high (30%+) percentage of the population living at or below the federal poverty level per ACS (8).

Diabetes, hypertension, asthma, cancer, heart disease, congestive heart failure (CHF), and chronic kidney disease (CKD) International Classification of Diseases, 10th Edition (ICD-10) codes were extracted from the EMR problem list. The Clinical Classification Software diagnosis grouper was used for disease prevalence (9), and Elixhauser comorbidity index used for chronic condition count (10). Laboratory values were from initial ICU presentation. If unavailable, laboratory values in the 7 days prior to a positive SARS-CoV-2 RT-PCR test were used.

From March 24, 2020, to May 12, 2020 hospitals reported ICU surge levels daily as an indicator of system strain. An internally developed quantitative scale scored each ICU from no surge (Level 0) to no space available for ICU patients or no ventilators available in hospital (Level 5) (Fig. 1). Surge levels were calculated and reported daily by ICU directors as part of a critical care council that managed strain across H+H. Patients were assigned the surge level of their admitting facility from the first day of ICU admission.

Figure 1.
Figure 1.:
Surge status category definitions and snapshot of daily surge status reporting from March 24, 2020, to April 16, 2020 (A) and April 17, 2020 to May 12, 2020 (B). Dates for surge scores are shown with dates expressed as month/d. Facility names have been replaced with random letter designations. CCU = critical care unit, NeuroICU = neurologic ICU, PACU = postanesthesia care unit, SICU = surgical ICU.

The primary outcome was mortality among ICU patients, defined as in-hospital mortality recorded in Epic. ICU length of stay (LOS), dialysis orders, and use of mechanical ventilation were secondary study outcomes. ICU LOS included multiple ICU stays if the patient returned to the ICU after discharge. Dialysis orders and need for renal replacement therapy (RRT) were captured by canceled, sent, or completed orders from March 24, to June 12. Orders for mechanical ventilation or extubation and ventilator flow sheets captured patients on mechanical ventilation.

Statistical Analysis

The primary outcome was in-hospital mortality. We explored differences in baseline sociodemographic characteristics, comorbidity, laboratory, treatment, and surge level variables between discharged and deceased patients via Pearson’s chi-square and Wilcoxon rank-sum tests (p < 0.05). Categorical variables were presented as percent and number of discharged or deceased and continuous variables as medians and interquartile range (IQR). Missing data were excluded.

We also stratified on surge level to further assess its impacts on mortality rates for key demographics, including age, race/ethnicity, and neighborhood poverty. Surge levels were stratified by high/very high level (≥ 3) and low/medium surge level (< 3).

Univariable and multivariable generalized estimating equation (GEE) models incorporating facility-level clustering were used to identify risk factors associated with mortality. The final model was selected using step-wise backward selection (threshold p < 0.1). Variables strongly associated with mortality in univariable analyses or of significance in the literature were included in the GEE multivariable step-wise backward selection. Covariates included sex, age, race/ethnicity, primary payer, marital status, primary spoken language, borough, neighborhood poverty level, hypertension, diabetes, asthma, cancer, heart disease, CHF, and CKD, surge level (high/very high vs low/medium), mechanical ventilation, dialysis order, and initial laboratory values of WBC count, platelet, hemoglobin, hematocrit, and creatinine. Odds ratios (ORs) and 95% CIs were generated. We also examined demographics and clinical characteristics of patients still admitted and excluded from the study population. p values of less than 0.05 were considered statistically significant. All analyses were conducted with SAS Enterprise Guide 7.1 (SAS Institute, Cary, NC).


Study Population

Between March 24, and May 12, 2020 2,233 patients with confirmed SARS-CoV-2 infection were admitted to ICUs across the NYC H+H system and were discharged or had died by June 12. Patients were majority male (n = 1,479; 66.0%) with median age of 63 years (IQR, 52–72 yr). Most patients were Black (25.8%) or Hispanic (35.3%), and the majority resided in the Bronx (31.4%), Queens (29.1%), or Brooklyn (28.4%). Patients in low, medium, high, or very high poverty neighborhoods comprised 3.4%, 34.6%, 27.7%, and 34.2%, respectively. English was the primary language (58.8%). The most common primary payer was medicare (41.0%) followed by medicaid (20.6%), uninsured (20.6%) and commercial/other (17.8%) (Table 1).

Comorbidities and select laboratory values are displayed in Table S1 ( A total of 92.5% of patients had at least one chronic condition, with 71.7% having hypertension, 62.6% diabetes, and 67.2% heart disease (Table S1, Most patients were vented (85.7%), and 25.3% had a dialysis order during their admission. Discharged patients had a median ICU LOS of 4.8 versus 6.4 days for deceased patients (Table 1).

TABLE 1. - Baseline Characteristics of Coronavirus Disease 2019 Patients With an ICU Stay
Characteristicsa All Patients, n Discharged Patients, n (%) Deceased Patients, n (%) p
Total, n 2,233 771 (34.5) 1,462 (65.5)
 Female 757 281 (37.1) 476 (62.9) 0.065
 Male 1,476 490 (33.2) 986 (66.8)
Age (yr)
 18–44 338 198 (58.6) 140 (41.4) < 0.0001
 45–64 883 336 (38.1) 547 (62)
 65–74 575 135 (23.5) 440 (76.5)
 75+ 437 102 (23.3) 335 (76.7)
 White 188 76 (40.4) 112 (59.6) 0.375
 Black/African American 574 200 (34.8) 374 (65.2)
 Hispanic 788 262 (33.3) 526 (66.8)
 Asian or Pacific Islander 146 43 (29.5) 103 (70.6)
 Non-Hispanic other 405 144 (35.6) 261 (64.4)
 Declined, missing, unknown 132 46 (34.9) 86 (65.2)
Primary payer
 Medicaid 459 194 (42.3) 265 (57.7) < 0.0001
 Medicare 916 244 (26.6) 672 (73.4)
 Commercial or other 397 172 (43.3) 225 (56.7)
 Uninsured 461 161 (34.9) 300 (65.1)
 Bronx 702 278 (39.6) 424 (60.4) < 0.0001
 Brooklyn 636 217 (34.1) 419 (65.9)
 Manhattan 213 71 (33.3) 142 (66.7)
 Queens 649 185 (28.5) 464 (71.5)
 Staten Island 3 2 (66.7) 1 (33.3)
 Non-NYC 30 18 (60.0) 12 (40.0)
Neighborhood poverty level
 Low poverty (0–9%)/non-NYC 77 41 (53.3) 36 (46.8) 0.001
 Medium poverty (10–19%) 771 255 (33.1) 516 (66.9)
 High poverty (20–29%) 620 196 (31.6) 424 (68.4)
 Very high poverty (30%+) 765 279 (36.5) 486 (63.5)
ICU surge status
 Low level 0–1 73 33 (45.2) 40 (54.8) < 0.0001
 Medium level 2/2.5 475 202 (42.5) 273 (57.5)
 High level 3/3.5 953 288 (30.2) 665 (69.8)
 Very high level 4+ 732 248 (33.9) 484 (66.1)
ICU length of stay, median (interquartile range) 5.8 (2.6–12.1) 4.7 (2.2–10.0) 6.4 (2.8–12.8) < 0.0001
NYC = New York City.
aData are presented as number and row percentages unless otherwise noted. Percentages may not add up to 100 due to rounding.

Predictors of In-Hospital Mortality

Overall, 1,468 patients (65.5%) died, and 772 (34.5%) were discharged during the study period (Table 1). The percent who died varied across age, primary payer, borough of residence, neighborhood poverty, surge status, marital status, most comorbidities (except asthma, cancer, and CHF), being ventilated, having a dialysis order, and abnormal WBC and platelet counts (Table 1) (Table S1,

In a comparison of crude mortality rates by select sociodemographic characteristics during periods of high and very high surge, the mortality rate was variable across racial/ethnic groups (Table 2). The mortality rate in Hispanic patients was significantly higher during periods of high/very high surge compared with low/medium surge (69.6% vs 56.4%; p = 0.0011); other groups were not found to be significantly different. Patients 65 years old and older had higher rates of mortality across all surge levels (Table 2). Patients 18–44 and 45–64 years old saw significant reductions in mortality from high/very high surge levels to low/medium surge levels (18–44: 46.4% vs 27.3%; p = 0.0017 and 45–64: 64.9% vs 53.2%; p = 0.002) (Table 2). Patients from medium, high, and very high poverty neighborhoods showed significant reductions in mortality from high/very high surge to low/medium surge (medium poverty: 69.5% vs 60.7%; p = 0.019 and high poverty: 71.2% vs 59.7%; p = 0.0078 and very high poverty: 66.6% vs 50.7%; p = 0.0003) (Table 2).

TABLE 2. - Mortality Rate by Demographics and Surge Status
ICU Surge Status: Low/Medium Level < 2 and 2/2.5 ICU Surge Status: High/Very High Level 3/3.5 and 4+
Characteristics Discharged Patients, n Deceased Patients, n Mortality Rate, % Discharged Patients, n Deceased Patients, n Mortality Rate, % p a
Total 235 313 57.1 536 1,149 68.2 < 0.0001
Age (yr)
 18–44 64 24 27.3 134 116 46.4 0.0017
 45–64 103 117 53.2 233 430 64.9 0.002
 65–74 38 101 72.7 97 339 77.8 0.2176
 75+ 30 71 70.3 72 264 78.6 0.0847
 White 23 28 54.9 53 84 61.3 0.4257
 Black/African American 61 96 61.1 139 278 66.7 0.216
 Hispanic 75 97 56.4 187 429 69.6 0.0011
 Asian or Pacific Islander 19 26 57.8 24 77 76.2 0.239
 Other 43 50 53.8 101 211 67.6 0.142
 Declined, missing, unknown 14 16 53.3 32 70 68.6 0.122
Neighborhood poverty level
 Low poverty (0–9%)/non-New York City 12 10 45.5 29 26 47.3 0.8851
 Medium poverty (10–19%) 88 136 60.7 167 380 69.5 0.019
 High poverty (20–29%) 62 92 59.7 134 332 71.2 0.0078
 Very high poverty (30%+) 73 75 50.7 206 411 66.6 0.0003
aχ2 assessing differences across rows.

In the univariable models, gender, primary spoken language, and hemoglobin (< 14 g/dL) did not reach significance to be included in the final model (Table S2, Marital status, borough of residence, hypertension, and CKD were significantly associated with mortality in univariable models but did not reach significance in backward step-wise elimination for final model inclusion (Table S2,

In the multivariable model, higher surge levels were associated with mortality (Level 3+ vs 0–2 OR, 1.4; 95% CI, 1.2–1.8) (Table 3). Sociodemographic variables such as race (Black vs White OR, 1.5; 95% CI, 1.1–2.0 and Asian vs White OR, 1.5; 95% CI, 1.0–2.3 and other vs White OR, 1.5; 95% CI, 1.0–2.3), age (45–64 vs 18–44 OR, 2.0; 95% CI, 1.6–2.5 and 65–74 vs 18–44 OR, 5.1; 95% CI, 3.3–8.0 and 75+ vs 18–44 OR, 6.8; 95% CI, 4.7–10.1), primary payer (uninsured vs commercial/other OR, 1.7; 95% CI, 1.2–2.3 and medicaid vs commercial/other OR, 1.3; 95% CI, 1.1–1.5), and neighborhood poverty level (medium vs low OR, 1.6; 95% CI, 1.0–2.4 and high vs low OR, 1.8; 95% CI, 1.3–2.5) were also independently associated with higher odds of death. Coexisting conditions including diabetes (OR, 1.6; 95% CI, 1.2–2.0), asthma (OR, 1.4; 95% CI, 1.1–1.8), and heart disease (OR, 2.5; 95% CI, 2.0–3.3) increased odds of death (Table 3). Adjusted odds of mortality were highest in patients who required mechanical ventilation (OR, 8.8; 95% CI, 6.1–12.9; p < 0.001) or had dialysis orders (OR, 3.0; 95% CI, 1.9–4.7; p < 0.001) compared with patients who did not have either intervention. Marital status and history of cancer remained in the final model but were insignificant.

TABLE 3. - Univariable and Final Adjusted Multivariable Generalized Estimating Equation Model for Odds of Mortality for Coronavirus Disease 2019 ICU Patients (n = 2,233)
Characteristicsa Univariable Models OR (95% CI) p Multivariable Model OR (95% CI) p
Age (yr)
 45–64 (reference: 18–44) 2.3 (1.8–2.9) < 0.001 2.0 (1.6–2.5) < 0.001
 65–74 (reference 18–44) 4.5 (3.2–6.5) < 0.001 5.1 (3.3–8.0) < 0.001
 75+ (reference 18–44) 4.7 (3.5–6.3) < 0.001 6.8 (4.7–10.1) < 0.001
 Black/African American (reference: White) 1.3 (1.1–1.6) 0.003 1.5 (1.1–2.0) 0.006
 Hispanic (reference: White) 1.3 (1.1–1.5) 0.002 1.4 (0.9–2.0) 0.107
 Asian or Pacific Islander (reference: White) 1.5 (1.1–2) 0.011 1.5 (1.0–2.3) 0.032
 Other (reference White) 1.2 (1–1.5) 0.016 1.5 (1.0–2.3) 0.038
 Declined, missing, unknown (reference: White) 1.4 (1–2) 0.072 1.1 (0.7–1.8) 0.615
Primary payer
 Medicaid (reference: commercial/other) 1 (0.9–1.2) 0.553 1.3 (1.1–1.5) 0.002
 Medicare (reference: commercial/other) 2.1 (1.7–2.7) < 0.001 1.1 (0.9–1.5) 0.244
 Uninsured (reference: commercial/other) 1.4 (1–1.8) 0.028 1.7 (1.2–2.3) 0.002
Marital status
 Single (reference: married, life partner) 0.8 (0.6–1) 0.018 0.9 (0.7–1.2) 0.435
 Divorced, separated, widowed (reference: married, life partner) 1.2 (0.9–1.5) 0.144 1 (0.7–1.3) 0.954
 Unknown (reference: married, life partner) 1.4 (1.1–1.9) 0.016 1.4 (0.9–2.0) 0.107
Neighborhood poverty levelb
 Medium poverty (10–19%) (reference: low poverty [0–9%]) 2.2 (1.6–3.1) < 0.001 1.6 (1.0–2.4) 0.041
 High poverty (20–29%) (reference: low poverty [0–9%]) 2.3 (1.6–3.3) < 0.001 1.8 (1.3–2.5) < 0.001
 Very high poverty (30%+) (reference: low poverty [0–9%]) 1.9 (1.3–2.8) 0.001 1.5 (0.9–2.5) 0.153
 Diabetes 1.7 (1.5–1.9) < 0.001 1.6 (1.2–2.0) < 0.001
 Asthma 1 (0.9–1.2) 0.744 1.4 (1.1–1.8) 0.001
 Cancer 0.9 (0.7–1.2) 0.5 0.8 (0.6–1.0) 0.082
 Heart disease 2.5 (2–3.1) < 0.001 2.5 (2–3.3) < 0.001
 Congestive heart failure 1.1 (0.9–1.4) 0.491 0.7 (0.6–1.0) 0.028
Elixhauser (count of comorbidities) 1.1 (1–1.1) < 0.001 0.9 (0.8–0.9) < 0.001
ICU surge status
 High/very high level (3/3.5 and 4+) (reference: low/medium level [< 2 and 2/2.5]) 1.5 (1.3–1.8) < 0.001 1.4 (1.2–1.8) < 0.001
 Ever vented (mechanical) 10 (7.5–13.4) < 0.001 8.8 (6.1–12.9) < 0.001
 Dialysis order 3.9 (2.9–5.3) < 0.001 3.0(1.9–4.7) < 0.001
 WBC (> 10,000 per mm3) 2.5 (2–3.2) < 0.001 1.8 (1.4–2.4) < 0.001
 Platelet count (<150,000 per mm3) 1.6 (1.3–2.1) < 0.001 1.9 (1.7–2.2) < 0.001
 Hematocrit (< 42%) 0.9 (0.7–1.2) 0.4693 0.8 (0.6–1) 0.044
 Creatinine (> 1.5 mg/dL) 3.2 (2.5–4) < 0.001 2.1 (1.7–2.7) < 0.001
OR = odds ratio.
aOR and 95% CI for mortality.
bLow poverty (0–9%) = low poverty (0–9%)/non-New York City.


We observed a 65.5% mortality rate among COVID-19 patients admitted to NYC H+H ICUs between March 24, and May 12, 2020, and discharged or deceased by June 12. ICU surge status at admission, race/ethnicity, age, neighborhood poverty level, comorbidities, and insurance status were all significantly associated with mortality in our final multivariable model, with patients admitted during periods of highest surge experiencing 1.4 times greater odds of dying.

Surge and Mortality

Various indicators have been used to measure the impact of surge strain on ICU mortality; however, bed availability, patient acuity, and number of admissions, stand out as key strain indicators associated with adverse outcomes (5). In a systematic review by Eriksson et al (11), nine of 12 studies on strain impact, in ICU settings, showed an increased risk of death. Additionally, in a large multicenter, observational cohort study of 149,310 patients across 213 hospitals in the United Kingdom, mortality risk decreased during lower strain periods as measured by bed census (5). Similarly, we found that increased strain, captured by our surge score, reflecting ICU capacity and resource availability, was associated with higher mortality risk.

A critical care council, composed of the H+H ICU directors, worked to reduce surge strain through the reallocation of staff, ventilators, resources, and interhospital patient transfers (12–14). During the study period, 189 critically ill COVID-19 patients were transferred to relieve strain with all but two originating from hospitals in Brooklyn and Queens. The prevalence of transfers from these hospitals demonstrates the disproportionate strain faced by certain hospitals and the steps needed to keep them from being completely overwhelmed.

During the highest surge periods, older patients, Hispanic patients, and those from higher poverty neighborhoods were admitted to NYC H+H ICUs at higher frequency than non-Hispanic, younger, and lower poverty neighborhood residing patients. As surge levels decreased, we observed a significant reduction in unadjusted mortality odds among patients 64 years or younger. The general association of older age with increased mortality may in part explain why the same changes in mortality were not observed in older adults.

Similar to younger patients, unadjusted mortality risk in Hispanic patients also improved with lower surge levels, whereas significant differences in mortality by surge status were not observed in other racial and ethnic groups. Many of the Hispanic patients in our sample were treated in H+H ICUs in Queens. Prior to the COVID-19 pandemic, H+H facilities in Queens had a total of 2.5 ICU beds per 10,000 residents compared with 2.8 in Brooklyn and 6.9 in the Bronx (15,16). Yet, during the study period, H+H ICUs cared for similar numbers of patients in Queens (n = 651; 29.1%), Brooklyn (n = 637; 28.4%), and the Bronx (n = 704; 31.4%). Initial ICU capacity was not proportional to the numbers of patients received and may help explain why Hispanic patients were more impacted by surge than other racial and ethnic groups. This highlights the need to understand and anticipate where surges will occur, so strain can be alleviated through the targeted supplementation of ICU capacity.

Poverty and surge level associations were also notable. Patients from neighborhoods with medium, high, and very high poverty levels saw significant reductions in unadjusted mortality risk as surge levels decreased, whereas the mortality rate in patients from lower poverty neighborhoods remained steady. Patients from lower poverty neighborhoods consistently had the lowest mortality rates and admissions numbers in our study.

Since the initial surges and first wave of the pandemic, the mortality rate in COVID-19 patients has improved by around 30%, but surge strain threatens to erase these improvements (17). In the months ahead, as hospitals and ICUs across the United States contend with record numbers of cases, steps must be taken to understand, prevent, and mitigate surge strain in order to reduce mortality.

Insurance Status and Mortality

Significant differences in mortality were observed based on insurance status. Uninsured critical patients have higher in-hospital mortality rates compared with insured patients (18–20), but there is little research on insurance status and outcomes in critical COVID-19 patients. One study of 102 patients at a safety net hospital in Brooklyn explored insurance status in critically ill patients but found no significant associations, likely due to small sample size (21). In our model, patients on medicaid, and those categorized as uninsured higher had higher odds ratios for mortality, compared with insured patients.

Immigration status may be relevant to our findings on insurance status. There are over 500,000 undocumented immigrants in NYC, with approximately one million residents living in a mixed-status household with at least one undocumented member (22). The uninsured population included subsets of the designated payers of COVID-19 relief as well as emergency medicaid, both of which provide medicaid coverage to those who meet income but not immigration requirements for full medicaid (23). As such, emergency medicaid is commonly used as a proxy for temporary visitors, and undocumented immigrants and increased risk for death observed in patients with emergency medicaid could highlight the increased burden of COVID-19 on immigrants and the undocumented.

Racial and Ethnic Identity and Mortality

Previous studies have demonstrated that differences in mortality risk between racial and ethnic groups become insignificant when controlling for comorbidities, poverty, and other factors (24–26). The persistence of race/ethnicity as a mortality risk factor in our multivariable model does not point to inherent differences in patients from varying racial and ethnic groups. Instead, racial and ethnic identity may serve as a proxy for additional confounding factors that were not accounted for in our model. For example, high SARS-CoV-2 plasma viral load has been associated with worse disease severity, leading to higher mortality rates amongst hospitalized patients (27). Members of racial and ethnic minority groups may have been exposed to greater viral inoculum due to higher rates of poverty and socioeconomic factors resulting in increased representation in frontline jobs, inability to social distance, and living in high density areas or multigenerational homes, which may have resulted in higher viral load and worse disease severity (28,29). Studies have also suggested that racial and ethnic disparities in COVID-19 mortality risk may be due to underlying systemic and structural racism that exists in the United States (30). Unequal exposure and risks due to racial gaps in wealth and health are also significant factors to racial and ethnic disparities seen in COVID-19 mortality (31).


We present one of the largest studies evaluating critically ill COVID-19 patients at public hospitals and include, ICU surge status, insurance status, and neighborhood poverty as explanatory variables in multivariable models alongside other factors such as age and comorbidities in predicting mortality.

Limitations of the study include that patients still admitted as of June 12 (n = 210) (Table S3, and seven patients with missing laboratory values were excluded from analysis, which may have resulted in a slightly higher reported mortality rate, as 92.1% of the deceased patients in our study died within 30 days of hospital admission. Comorbidity data were obtained via ICD-10 codes, which may be subject to errors (32). There may have been differential documentation of chronic conditions, especially for patients new to our hospitals. The number of patients who received RRT was unknown with orders for dialysis used as a proxy which may have attenuated the OR. We were unable to report the acuity of our patients at admission due to pandemic related issues with reliability and availability of acuity metrics.


ICU strain caused by COVID-19 surges was captured using a quantitative surge score that reflected space, staffing, and critical resource availability. Patients admitted to ICUs with higher surge scores at the time of admission were at greater risk of death. As surge scores decreased, significant improvements in mortality rate were observed in younger, middle aged, Hispanic, and higher poverty level residing patients. Patients on medicaid, emergency medicaid, self-paying, or covered by COVID-19 relief funding had higher mortality rates compared with commercially/other insured patients. More research is needed to explore disparities in the impact of surge on patient outcomes and differences in mortality risk among patients with varying demographics across different locations in a large public healthcare system. The impact of surge on mortality reaffirms the vital importance of continued public health, community, and individual efforts to limit disease transmission and prevent new COVID-19 surges.


We wish to thank Professor of Biostatistics, Mengling Liu, PhD, from the Department of Population Health and Department of Environmental Medicine at New York University School of Medicine for her statistical expertise and advice during the peer review process.


1. Meyer R, Alexis CThe U.S. Has Passed the Hospital Breaking Point. 2020The Atlantic,
2. Alltucker K, Aleszu BWe’re Not Winning This Battle: Relentless COVID-19 Surge Fills 1 in 8 hospital ICU Units2020USA Today,
3. Arabi Y, Venkatesh S, Haddad S, et al.: A prospective study of prolonged stay in the intensive care unit: Predictors and impact on resource utilization. Int J Qual Health Care 2002; 14:403–410
4. Filardo TD, Khan MR, Krawczyk N, et al.: Comorbidity and clinical factors associated with COVID-19 critical illness and mortality at a large public hospital in New York City in the early phase of the pandemic (March-April 2020). PLoS One 2020; 15:e0242760
5. Wilcox ME, Harrison DA, Patel A, et al.: Higher ICU capacity strain is associated with increased acute mortality in closed ICUs. Crit Care Med 2020; 48:709–716
6. Azar KMJ, Shen Z, Romanelli RJ, et al.: Disparities in outcomes among COVID-19 patients in a large health care system in California. Health Aff (Millwood) 2020; 39:1253–1262
7. HRSA: COVID-19 Claims Reimbursement to Health Care Providers and Facilities for Testing and Treatment of the Uninsured.2020. Available at: Accessed December 17, 2020.
8. United States Census Bureau: American Community Survey, 2016. Available at: Accessed December 17, 2020.
9. Agency for Healthcare Research and Quality: Clinical Classifications Software Refined (CCSR), 2020. Available at: Accessed December 17, 2020.
10. Elixhauser A, Steiner C, Harris DR, et al.: Comorbidity measures for use with administrative data. Med Care 1998; 36:8–27
11. Eriksson CO, Stoner RC, Eden KB, et al.: The association between hospital capacity strain and inpatient outcomes in highly developed countries: A systematic review. J Gen Intern Med 2017; 32:686–696
12. Uppal A, Silvestri DM, Siegler M, et al.: Critical care and emergency department response at the epicenter of the COVID-19 pandemic. Health Aff (Millwood) 2020; 39:1443–1449
13. Mukherjee V, Toth AT, Fenianos M, et al.: Clinical outcomes in critically ill coronavirus disease 2019 patients: A unique New York City public hospital experience. Crit Care Explor 2020; 2:e0188
14. Bhatt A, Nair S, Postelnicu R, et al.: Building the pyramids: A perspective on creating and upscaling a critical care workforce at a public hospital during the coronavirus disease 2019 pandemic in New York City. Chest 2020; 158:884–886
15. New York State Department of Health: NYS Health Profiles: Find and Compare New York Health Care Providers.2020. Available at: Accessed September 11, 2020.
16. United States Census Bureau2020. QuickFacts, New York City, New York; United StatesAvailable at:,US/PST045219. Accessed December 17, 2020.
17. The Institute for Health Metrics and Evaluation2020. COVID-19 Results Briefing for Massachusetts, Available at: Accessed December 17, 2020.
18. Dillman J, Mancas B, Jacoby M, et al.: A review of the literature: differences in outcomes for uninsured versus insured critically ill patients: Opportunities and challenges for critical care nurses as the Patient protection and affordable care act begins open enrollment for all americans. Dimens Crit Care Nurs 2014; 33:8–14
19. Fowler RA, Noyahr LA, Thornton JD, et al.; American Thoracic Society Disparities in Healthcare Group: An official American Thoracic Society systematic review: The association between health insurance status and access, care delivery, and outcomes for patients who are critically ill. Am J Respir Crit Care Med 2010; 181:1003–1011
20. Gerry JM, Weiser TG, Spain DA, et al.: Uninsured status may be more predictive of outcomes among the severely injured than minority race. Injury 2016; 47:197–202
21. Capone S, Abramyan S, Ross B, et al.: Characterization of critically Ill COVID-19 patients at a Brooklyn safety-net hospital. Cureus 2020; 12:e9809
22. Affairs OoI (Ed). Office NYCMs: State of Our Immigrant City.2018
23. Medicaid for the Treatment of an Emergency Medical Condition Fact Sheet.Available at: Accessed September 4, 2020.
24. Kabarriti R, Brodin NP, Maron MI, et al.: Association of race and ethnicity with comorbidities and survival among patients with COVID-19 at an Urban Medical Center in New York. JAMA Netw Open 2020; 3:e2019795
25. Price-Haywood EG, Burton J, Fort D, et al.: Hospitalization and mortality among black patients and white patients with Covid-19. N Engl J Med 2020; 382:2534–2543
26. Yehia BR, Winegar A, Fogel R, et al.: Association of race with mortality among patients hospitalized with coronavirus disease 2019 (COVID-19) at 92 US hospitals. JAMA Netw Open 2020; 3:e2018039
27. Fajnzylber J, Regan J, Coxen K, et al.; Massachusetts Consortium for Pathogen Readiness: SARS-CoV-2 viral load is associated with increased disease severity and mortality. Nat Commun 2020; 11:5493
28. Tai DBG, Shah A, Doubeni CA, et al.: The disproportionate impact of COVID-19 on racial and ethnic minorities in the United States. Clin Infect Dis 2021; 72:703–706
29. Vahidy FS, Nicolas JC, Meeks JR, et al.: Racial and ethnic disparities in SARS-CoV-2 pandemic: Analysis of a COVID-19 observational registry for a diverse US metropolitan population. medRxiv. 2020
30. Egede LE, Walker RJ, Garacci E, et al.: Racial/ethnic differences in COVID-19 screening, hospitalization, and mortality in southeast wisconsin: Study examines racial/ethnic differences in COVID-19 screening, symptom presentation, hospitalization, and mortality among 31,549 adults tested for COVID-19 in Wisconsin. Health Affairs 2020; 39:1926–1934
31. Basett MT: COVID-19 and Structural Racism. 2020In: Sixth Annual Health Disparitis Symposium, NYU Langone, New York,
32. O’Malley KJ, Cook KF, Price MD, et al.: Measuring diagnoses: ICD code accuracy. Health Serv Res 2005; 40:1620–1639

coronavirus disease 2019; healthcare disparities; intensive care units supply and distribution; mortality; public hospitals; severe acute respiratory syndrome coronavirus-2

Supplemental Digital Content

Copyright © 2021 by the Society of Critical Care Medicine and Wolters Kluwer Health, Inc. All Rights Reserved.