Comorbidities, sequelae, blood biomarkers and their associated clinical outcomes in the Mount Sinai Health System COVID-19 patients

With the continuing rise of SARS-CoV2 infection globally and the emergence of various waves in different countries, understanding characteristics of susceptibility to infection, clinical severity, and outcomes remain vital. In this retrospective study, data was extracted for 39,539 patients from the de-identified Mount Sinai Health System COVID-19 database. We assessed the risk of mortality based on the presence of comorbidities and organ-specific sequelae in 7,032 CoV2 positive (+) patients. Prevalence of cardiovascular and metabolic comorbidities was high among SARS-CoV2+ individuals. Diabetes, obesity, coronary artery disease, hypertension, atrial fibrillation, and heart failure all increased overall mortality risk, while asthma did not. Ethnicity modified the risk of mortality associated with these comorbidities. With regards to secondary complications in the setting of infection, individuals with acute kidney injury and acute myocardial injury showed an increase in mortality risk. Cerebral infarcts and acute venous thromboembolic events were not associated with increased risk of mortality. Biomarkers for cardiovascular injury, coagulation, and inflammation were compared between deceased and survived individuals. We found that cardiac and coagulation biomarkers were elevated and fell beyond normal range more often in deceased patients. Several, but not all, inflammatory markers evaluated were increased in deceased patients. In summary, we identified comorbidities and sequelae along with peripheral blood biomarkers that were associated with elevated clinical severity and poor outcomes in COVID-19 patients. Overall, these findings detail the granularity of previously reported factors which may impact susceptibility, clinical severity, and mortality during the course of COVID-19 disease.


Introduction
During the global COVID-19 pandemic, caused by the severe acute respiratory syndrome Coronavirus 2 (SARS-CoV2), the World Health Organization has reported more than 166 Multiple variables were featured in the database and used in stratification and characterization of the cohort's associated COVID-19 risk. These included age, sex, race, ethnicity, SARS--CoV2 qRT-PCR testing, encounter types, location of care, admission type codes, discharge locations, deceased indicators, comorbidities, sequelae, and varying biomarkers. Consistent with the database, only ethnic groups represented in the US Census were included in our analyses: White, Hispanic, Black or African American, Other, Asian, Native Hawaiian or Pacific Islander, and American Indian or Alaskan native. Black or African Americans will be referred to as Black for the remainder of the manuscript. A combined race and ethnicity variable defined by Mount Sinai's COVID Disparities and Equity Taskforce was used to stratify demographic information to better capture Hispanic/Latinx patients. Considering this combined variable, we will refer to race and ethnicity simply as ethnicity for the remainder of the manuscript.
Data preparation: COVID-19 severity score. To reflect COVID-19 infection severity, we created a scoring system based on the patient's clinical service encounter such that OP visits represented mild cases, and the degree of severity subsequently increased from ED visits to IP admissions and finally to ICU admissions, which represented the most severe clinical course. Unspecified service encounters, including radio-oncology and neuropsychiatric visits (n = 67 patients), were not included. After a severity score was calculated for each encounter, only the first encounter with the maximum severity score was selected to represent an individual patient's clinical course. This scoring system limited the patient's representation in our studied cohort to only one encounter and allowed for a more accurate representation of a patient's From February 28th to June 8th, 2020, this database consisted of 80,108 masked electronic record numbers (representing individual patients) that entered the MSHS with a COVID-19 related encounter, including 109,000 individual service encounters. We analyzed de-identified patient data from 39,625 patients in the Mount Sinai Health System (MSHS), which had recorded metrics with respect to sex, age, and race. We found that 6,537 of them were determined to be CoV2+ by nasopharyngeal qRT-PCR confirmed tests, while 495 patients were presumed positive for a total of 7,032 COVID-19 cases. We excluded all patients for which an outcome was not yet known, limiting the analysis to a total of 6,924 patients-5,629 individuals who survived and 1,295 deceased.
https://doi.org/10.1371/journal.pone.0253660.g001 clinical course. We excluded encounters that included unspecified/un-identified data with respect to age, sex, and ethnicity. This filtration, in addition to the removal of patients with unidentified service, encounters limited our analysis to include 39,539 patients.
SARS CoV-2 qRT-PCR testing. With regards to COVID-19 nasopharyngeal qRT-PCR testing status, the database included individuals who tested positive (CoV2+), presumed positive, negative (COV2-), and individuals who were not tested. Presumed positive indicates that a patient's positive qRT-PCR result has not yet been confirmed by the Center for Disease Control and Prevention (CDC). Within our total cohort of 39,539 patients, 17.19% were CoV2+ (which includes presumed positives), 49.42% were CoV2-, and 32.75% were not tested for the SARS-CoV2 antigen. For our analyses we focused on patients with positive (n = 6,537 patients) or presumed positive (n = 495 patients) qRT-PCR results (i.e., CoV2+ patients). Additionally, we excluded all patients for which an outcome was not yet known, limiting the analysis to a total of 6,924 patients-5,629 individuals who survived and 1,295 deceased. The count of survivors versus deceased individuals is reflective of results as of June 8 th 2020. It is possible that discharge proceeds the event of death or death occurs at home for outpatients. Considering our calculations were performed on the basis of COVID-19 survivors and deceased populations, we further excluded patients with unknown outcomes, i.e., were not discharged or deceased. Considering 93.36% of our deceased patient population is captured in the IP and ICU units, we removed patients whose discharge location was 1) unknown ("blank", n = 101) or 2) involved "admission to inpatient" (n = 7). The latter category represented ED patients with a single encounter at MSHS that did not end up being admitted to IP for unclear reasons. Of all patients discharged to hospice (n = 45), 18 were indicated as deceased, and 27 were indicated as survivors. Despite the poor prognosis of hospice patients, we included these patients in our analysis to avoid an assumption on the outcome-they represent 0.48% of all survivors.

Examination of biomarkers in peripheral blood
We analyzed variance in different lab values between survivors and deceased individuals, including cardiovascular biomarkers [troponin I (TnI) and brain natriuretic peptide (BNP)], markers of inflammation [IL-8, IL-1b, IL-6, TNFα, C reactive protein (CRP)] as well as markers of coagulation [platelets, mean platelet volume (MPV), fibrinogen, d-dimer, and active partial prothrombin time (aPTT)]. The type and frequency of lab results vary per individual. Markers may be missing for certain patients. Results were considered for all individuals with at least one recorded lab result for the marker of interest; average value across all lab results was used.

Statistical analysis
All statistical analyses were conducted using GraphPad Prism version 7.00 for Windows. Relative risk comparison for each comorbidity and sequela was done independently. For example, to assess the relative risk of mortality in patients with CAD, the total CoV2+ population was considered. Within this population, relative risk analysis was done to compare the mortality risk of patients with CAD and those without CAD. The mortality risk of patients with CAD is the number of SARS-CoV2+ patients with CAD who were deceased divided by the number of total SARS-CoV2+ patients with CAD. The mortality risk of patients without CAD is the number of SARS-CoV2+ patients without CAD who were deceased divided by the number of total SARS-CoV2+ patients without CAD. The former mortality risk was then divided by the latter relative risk to obtain the relative risk values. This process was repeated for each comorbidity and each sequela. However, it is possible that these individuals have one or more other comorbidities and/or sequela and therefore were represented in multiple analyses. The H 0 for relative risk analysis states RR = 1.0; neither group faces a greater risk. The Fisher's exact test was used for the computation of the p-value. Two-sided p-value calculations were done at a 95% CI (α = 0.05) by the Koopman asymptomatic score method.
We compared the mean of average lab results in survivors versus deceased individuals for a given biomarker of interest using unpaired two-tailed unpaired t-test with Welch's correction at 95% CI, where H 0 states that there is no statistically significant different in means between the compared groups. ROUT method was used to identify outliers where definitive outliers were removed (Q = 0.1%). The calculations were done at a 95% CI (α = 0.05). Corrections for multiple hypothesis testing were not done due to its unavailability on the GraphPad Prism software used. However, per the Bonferroni correction, the α value could be adjusted by dividing the α value (α = 0.05) by the number of comparisons (n = 12). The new α value would be 0.004. All p values calculated still fall below this threshold and would therefore still be significant.

Ethnicity
We characterized the patient population based on their ethnicity. The CoV2+ patient population (n = 7,032) is represented by 27.3% Hispanics, 28.1% Blacks or African Americans, 25.9% Whites, 13.9% Others, and 4.6% Asians. Note, here and elsewhere in the manuscript, Blacks or African Americans will be referred to as Blacks.

Comorbidities and mortality
The percentage of the CoV2+ population with at least one comorbidity from the corresponding categories are-35.6% metabolic, 28.6% cardiovascular, 8.1% renal, and 6.4% pulmonary (Table 1A). Within these categories, obesity was most prevalent, followed by hypertension (HTN), diabetes mellitus (DM), coronary artery disease (CAD), chronic kidney disease (CKD), heart failure (HF), atrial fibrillation (AFib), asthma, and lastly chronic obstructive pulmonary disease (COPD) ( Table 1B). Of note, an individual may be represented in several categories depending on their number of pre-existing conditions. Mechanisms by which these comorbidities may affect the outcome of COVID-19 disease is unclear. Nonetheless, comorbidities could modify risks of infection, severity of COVID-19 clinical course, and response to administered therapies.
All evaluated comorbidities, with the exception of asthma, were associated with increased risk of mortality (Fig 2A-2D 2D), though it is important to note that the stage of CKD was not indicated in the MSHS database. Overall, the risk of mortality was approximately 2-fold greater in individuals with at least one chronic comorbidity in any category (p < 0.0001) (Fig 2A-2D).
The prevalence (absolute number and percentage) of individuals with each comorbidity within their indicated ethnic group is summarized in Table 2.
We further risk-stratified patients based on their ethnicity and sequelae. The prevalence (absolute number and percentage) of individuals who developed each specific sequela within their indicated ethnic group showed that clinical complications secondary to COVID-19 appeared to have poorer outcomes and were not associated with patient's ethnicity (Table 4).

Serum biomarkers of inflammation, coagulation, and cardiac injury
We assessed whether established inflammatory, coagulation, cardiovascular clinical biomarkers could provide further insight into risk of mortality in CoV2+ individuals (Fig 4A-4C and

PLOS ONE
Comorbidities, sequalae and their associated clinical outcomes in COVID-19 patients Table 5). All values for a given biomarker from a single patient are averaged if more than one result is available. The mean across patients for that biomarker is then calculated within the survivors and deceased patients for comparison. The mean ± standard deviation, the prevalence of patients outside the reference range, and the number of subjects with at least one laboratory test value for a given biomarker are provided in Table 5. Of note, the presence of large deviation in reported values can possibly be attributed to significant inter-patient variability in the context of where in the course of disease the patient is, disease severity, as well as response to treatment. Inflammatory biomarkers and mortality. The levels of several inflammatory biomarkers were elevated in CoV2+ patients, and the degree by which these markers were elevated appears to be associated with poorer clinical outcomes. The levels of IL-8, IL6, and CRP were elevated above the normal range in both survived and deceased patients; however, all three inflammatory markers were approximately 2-4-fold higher in deceased COVID-19 patients compared to patients who recovered (p<0.0001) (Fig 4A and Table 5). Although the deceased patients  had a small 20% increase (p<0.0001) in IL-1β levels compared to the survived patient population, these levels remained within the normal reference range for IL-1β (0-5 pg/mL) in both patient populations (Fig 4A and Table 5). The average levels of TNFα were increased by approximately 30% (p<0.0001) in the deceased patients compared to survivors. In both patient populations, these increases were above the normal range (0-22 pg/mL) of serum TNFα ( Fig  4A and Table 5). It is important to note that levels of interleukins, TNFα, and CRP were tested almost exclusively in hospitalized COVID-19 patients. Thus our inflammatory biomarkers data represent COVID-19 patients with moderate-severe disease, but not asymptomatic or mild cases. Coagulation markers and mortality. Since the beginning of the COVID-19 pandemic, significant evidence had emerged for increased incidence of coagulopathy, endotheliopathy, and various thrombotic and bleeding events in patients with COVID-19 [17][18][19][20][21][22]. We evaluated the levels of several coagulation markers in our CoV2+ patient population to elucidate underlying patterns (Fig 4B). The average mean platelet volume (MPV) levels were mostly within the reference range for both survivors and deceased patients. However, there was a very small 7% increase in MPV levels for deceased patients compared to survivors (Fig 4B and  Table 5). Platelet (Plt) count remained within the lower threshold of normal; however, deceased patients had 12.5% (p<0.0001) lower average Plt count compared to survivors ( Fig  4B and Table 5). The average levels of fibrinogen were increased by approximately 50% in both survivors and deceased patients, approaching the levels (>700 mg/dL) that can promote the formation of clots. D-dimer was elevated above the upper limit of normal in both survived and deceased patients; however, the average d-dimer levels were more than 250% (p<0.0001) higher in the deceased population compared to survivors (Fig 4B and Table 5). The increased

PLOS ONE
levels of fibrinogen and d-dimer above normal thresholds in both groups may suggest an increased VTE risk with presence of macro-and/or micro-thrombi. These findings could suggest that peak values of these markers may be associated with a worse clinical outcome and the need for early administration of anticoagulant therapy as indicated by Paranjpe et al. [23]. Average activated partial prothrombin time (aPTT) values were prolonged by approximately 30% (p<0.0001) in deceased patients compared to those who survived (Fig 4B and Table 5). Taken together, dysregulation of coagulation markers can significantly affect the clotting cascade, and coagulopathies must be considered as one of the major sequelae of COVID-19, considering the reported incidence of thrombi as well as clotting factor deficiencies noted in other studies [23,24]. Cardiovascular markers and mortality. While average cardiac troponin-I (cTnI) levels were at the upper limit of normal in the majority of COVID-19 patients that survived, there was more than a 330% (p<0.0001) increase in average cTnI levels in deceased patients (Fig 4C and Table 5), suggesting that the extent of myocardial injury is associated with worse clinical outcomes. Brain natriuretic peptide (BNP) levels were also significantly increased by more than 200% (p<0.0001) in deceased patients compared to survivors (Fig 4C and Table 5). Although cardiac structural and functional studies are unavailable in this de-identified database study, the substantially elevated levels of cardiovascular biomarkers in deceased patients suggest that significant myocardial injury could predict poorer outcomes in COVID-19 patients.

Discussion
Since the initial reports of COVID-19 in early December 2019, the novel coronavirus outbreak continues to strain the global population, while its pathogenesis and clinical outcomes are not fully understood, and therapies continue to evolve. Using de-identified data from one of the largest US patient cohorts across a range of clinical services including outpatient (OP), emergency department (ED), inpatient (IP), and intensive care unit (ICU) admission, we assessed the associations between various demographic factors including age, sex, ethnicity on SARS--CoV2 qRT-PCR testing and COVID-19 clinical course severity and mortality. Of 39,539 patients who entered the MSHS with a COVID-19 concern as of June 8 th 2020, 7,032 patients (17.9%) were positive or presumed positive for COVID-19 of which 1,295 patients passed away (18.4%) while 5,737 recovered and were discharged (81.6%). Within this confirmed CoV2+ population, we evaluated a large panel of peripheral blood biomarkers, the impact of several comorbidities, and COVID-19 sequelae on patient outcomes.
Multiple epidemiological studies have reported that cardiovascular comorbidities are associated with COVID-19 disease severity and poorer outcomes [25][26][27][28]. Within our CoV2+ patient population, there was a high prevalence of cardiovascular and metabolic comorbidities compared to renal and pulmonary comorbidities. Despite difference in prevalence, comorbidities in these four major tested categories were associated with increased mortality risk, which is consistent with previous reports [12,[29][30][31]. Asthma was the only comorbidity that was not associated with increased mortality risk. Lack of associated mortality risk in patients with asthma may suggest that these individuals have some type of "hypoxic preconditioning" effect due to the occurrence of repetitive acute hypoxic episodes during exacerbations of their disease, a phenomenon recently described in a murine model [32]. In addition, it is also conceivable that chronic management of asthma attacks with inhaled corticosteroids may also provide an enhanced protective effect as a substantial number of patients administered steroids in combination with other therapies had shown improved outcomes in the setting of COVID-19 [33].
Regardless of ethnicities, patients with underlying HTN, CAD, HF, AFib, DM, obesity, COPD, and CKD experienced a higher risk of mortality compared to patients without these comorbidities. Indeed, the presence of at least one cardiovascular, renal, or pulmonary comorbidity increased the probability of poorer outcomes. However, we could not find significant race/ethnicity-associated modifications of mortality risks for most COVID-19 patients with these comorbidities, except for obesity, for which the risk of mortality was unchanged in the Hispanic population. These findings emphasize that in all ethnic groups, these comorbidities pose risks for worse clinical outcomes for COVID-19 patients, as reported before for other patient cohorts nationally and internationally [12,[29][30][31]. Of note, in spite of small differences observed between ethnic groups in terms of the degree by which comorbidity affects mortality rates, our analysis is limited to comparing whether patients have established comorbidity and does not take into consideration the patient's age, stage of disease, or the therapy they have been receiving for their underlying medical condition(s).
Analysis of the prevalence of COVID-19 complications/sequelae revealed that renal and cardiovascular complications were the most prevalent, followed by cerebrovascular and pulmonary. Across all ethnic groups, the highest mortality risk was found in patients that developed AKI, followed by AMI and ARDS. While ethnicity-associated modifications were observed for AMI and ARDS, the AKI risk was not modified by ethnicity. All ethnic groups, except Black patients with AMI, were at increased risk of mortality due to COVID-19. Further analysis of clinical data, which was not available in the MSHS de-identified dataset, would be required to understand the potential causes of these findings. Although, the Black COVID-19 patients had the lowest prevalence for developing ARDS, we found that ARDS was associated with the highest risk of mortality in the Black population which is consistent with prior published studies [34,35]. The worse outcomes due to ARDS in this population may be attributed to a more severe disease course or underlying comorbidities, which could increase predisposition for the development of ARDS.
Surprisingly, AVTE events and CVAs were not associated with an increased risk of mortality, potentially due to the administration of anticoagulation therapy in most of these patients. This may be attributed to early initiation of venous thromboembolism prophylaxis as of April 10 th 2020, with low molecular weight heparin (LMWH) in hospitalized patients in the MSHS with suspected or qRT-PCR confirmed COVID-19 [36].
In our analyses of peripheral blood biomarkers for cardiovascular disease, coagulation, and inflammation, elevation of these markers was consistently higher in the deceased versus recovered COVID-19 patients. Most of these circulating markers were above upper threshold, but IL-1β remained within normal limits for both patient populations. However, even these IL-1β values were statistically higher in deceased patients. A prior study of cytokines in a limited COVID-19 cohort study of 40 ICU patients reported IL-1β levels to be elevated above the upper limit of normal [37]. Considering our sample size and IL-1β's role in the initiation of cytokine release syndrome and induced secretion of IL-6 [38], it may be advisable to consider the ratio of these cytokines (i.e. IL-1β/IL-6) to better understand the cytokine profile in COVID-19 patients and the associated inflammatory state.
CRP, IL-6, IL-8, and TNFα levels were significantly higher in deceased patients, suggesting a greater degree of systemic inflammation perpetuated by a "cytokine storm" which is supported by significant elevations in IL-6, IL-8, and TNFα [2,7,39,40]. Overall, these findings reflect more active systemic inflammation in moderate/severe cases, which may contribute to the development of secondary sequelae, including renal failure or acute liver injury, supporting the concept of a high-risk inflammatory state secondary to COVID-19 and the associated higher risk of mortality.
Coagulation markers, including d-dimer and fibrinogen levels, were significantly elevated above the upper limit of normal in both patient groups suggesting COVID-19 is associated with coagulopathies [41]. Deceased patients had higher circulating d-dimer and fibrinogen compared to surviving patients, which may be reflective of the extent of micro-and macrothrombi formation observed in disseminated intravascular coagulation [42]. An earlier study from NYU also showed 86% of their patients (n = 2,782) had elevated d-dimers (>230 ng/mL) with mean peak of d-dimer at 767 ng/ml, and those with elevated d-dimers were more likely to become critically ill [43]. Thus, our findings support the use of d-dimer as a biomarker for clinical course severity as well as being associated with poor outcomes.
Cardiac injury and stress are consistent with cardiac troponin and BNP elevation in the COVID-19 patient population and could serve as potential predictors for the development of COVID-19 sequelae such as AMI, AKI, HF exacerbation, or pulmonary edema. In our study, the COVID-19 deceased population had 3-fold higher mean cTnI levels compared to survivors. Although there are many potential etiologies of myocardial injury in the setting of COVID-19, including myocarditis [44], septic and/or stress cardiomyopathy [45,46], or secondary hypoxic injury [47], our study did show that elevation in cTnI may be associated with the degree of multisystem organ dysfunction, and thus disease severity. These findings are consistent with an earlier report from the MSHS, where even small amounts of myocardial injury (cTnI > 0.03 to 0.09 ng/ml) were associated with higher risk of mortality [48]. Earlier reports from Wuhan, China also demonstrated elevated troponin in hospitalized patients with levels of greater than 2.14 mg/L predicting in-hospital mortality [49]. Within our patient population, the distribution of BNP also varied between survivors and deceased individuals. Considering the lack of further clinical data, elevation in BNP may also be multifactorial with either intrinsic cardiac dysfunction or secondary etiologies such as pulmonary edema in the setting of acute respiratory distress and renal disease. Earlier studies have also shown elevated cTnI [50] and BNP [51] in patients with COVID-19. However, given the variance in assays and lack of details regarding their upper thresholds, direct comparisons cannot be made for cTnI and BNP values in our studies and those of others. Despite these limitations, our data supports the presence of cardiovascular involvement in COVID-19, and more importantly reiterates the predictive clinical valve of cTnI and BNP for disease severity and outcomes.
In summary, our study is utilizing one of the most diverse and largest COVID-19 patient databases to study various factors that can influence susceptibility to infection and clinical outcomes. We found that-i) underlying metabolic and cardiovascular comorbidities are most prevalent in our COVID-19 patient population and were associated with increased mortality risk along with COPD and CKD; ii) asthma did not modify mortality risk in CoV2+ patients; iii) development of AKI, AMI and ARDS secondary to COVID-19 was associated with increased risk of mortality; iv) d-dimer, CRP, IL-6, TNFα, cTnI, and BNP levels were significantly elevated in deceased individuals. Taken together with previously published results, the finding in our study should assist clinicians and other health care workers in devising strategies to mitigate the risk of infection, assist with determining the timing for initiation of given therapies as well as altering them to optimize medical management.

Study limitations
There are limitations in our study worth highlighting. Although, we included hospice patients in our relative risk analysis, we included hospice patients that were not indicated as deceased in the recovered category. However, this represents a very small portion of survivors or recovered patients (0.48%). Representing them in such a way introduces only a 2.12% reduction in deceased patients. Our analysis also included all emergency department and outpatients. However, their outcomes are not known unless they returned to MSHS for follow-up. Similarly, on discharge, inpatients could have been lost to follow-up, so their outcomes are not known unless they returned to MSHS. While other studies have used hazard functions to address this issue, they have focused on hospitalized populations. Our study facilitates assessment for looking at risks in the overall population entering the health system(s). Risks may still be overstated as many patients remain asymptomatic or have mild symptoms and therefore may not enter hospital systems. The false-negative rate of qRT-PCR tests for CoV2 infection is also a limitation in all current population reports, with studies estimate the sensitivity of most tests to be 70% [52].