key: cord-0923044-ye4dtna5
authors: Garibaldi, B. T.; Fiksel, J.; Muschelli, J.; Robinson, M. L.; Rouhizadeh, M.; Nagy, P.; Gray, J. H.; Malapati, H.; Ghobadi-Krueger, M.; Niessen, T. M.; Kim, B. S.; Hill, P. M.; Ahmed, M. S.; Dobkin, E. D.; Blanding, R.; Abele, J.; Woods, B.; Harkness, K.; Thiemann, D. R.; Bowring, M. G.; Shah, A. B.; Wang, M. C.; Bandeen-Roche, K.; Rosen, A.; Zeger, S. L.; Gupta, A.
title: Patient trajectories and risk factors for severe outcomes among persons hospitalized for COVID-19 in the Maryland/DC region
date: 2020-05-26
journal: nan
DOI: 10.1101/2020.05.24.20111864
sha: 91279a54e6f767669649294e3e10038f466b528c
doc_id: 923044
cord_uid: ye4dtna5

Background: Risk factors for poor outcomes from COVID-19 are emerging among US cohorts, but patient trajectories during hospitalization ranging from mild-moderate, severe, and death and the factors associated with these outcomes have been underexplored. Methods: We performed a cohort analysis of consecutive COVID-19 hospital admissions at 5 Johns Hopkins hospitals in the Baltimore/DC area between March 4 and April 24, 2020. Disease severity and outcomes were classified using the WHO COVID-19 disease severity ordinal scale. Cox proportional-hazards regressions were performed to assess relationships between demographics, clinical features and progression to severe disease or death. Results: 832 COVID-19 patients were hospitalized; 633 (76.1%) were discharged, 113 (13.6%) died, and 85 (10.2%) remained hospitalized. Among those discharged, 518 (82%) had mild/moderate and 116 (18%) had severe illness. Mortality was statistically significantly associated with increasing age per 10 years (adjusted hazard ratio (aHR) 1.54; 95%CI 1.28-1.84), nursing home residence (aHR 2.13, 95%CI 1.41-3.23), Charlson comorbidity index (1.13; 95% CI 1.02-1.26), respiratory rate (aHR 1.13; 95%CI 1.09-1.17), D-dimer greater than 1mg/dL (aHR 2.79; 95% 1.53-5.09), and detectable troponin (aHR 2.79; 95%CI 1.53-5.09). In patients under 60, only male sex (aHR 1.7;95%CI 1.11-2.58), increasing body mass index (BMI) (aHR1.25 1.14-1.37), Charlson score (aHR 1.27; 1.1-1.46) and respiratory rate (aHR 1.16; 95%CI 1.13-1.2) were associated with severe illness or death. Conclusions: A combination of demographic and clinical features on admission is strongly associated with progression to severe disease or death in a US cohort of COVID-19 patients. Younger patients have distinct risk factors for poor outcomes.

The first case of SARS-CoV-2 in the United States was identified January 20 th , 2020 in a returned traveler from Wuhan, China. 1 The US accounted for nearly a third of the world's cases (1,577,758) and deaths (94,729) as of May 22, 2020. 2 After infection with SARS-CoV-2, outcomes range from asymptomatic or mild illness to more severe illness and death. 3, 4 Age, sex, smoking, race, body mass index (BMI), and comorbidities such as hypertension and diabetes are important risk factors for severe outcomes, though to varying degrees.

Elevated inflammatory markers and lymphopenia are also associated with severe outcomes in COVID-19 (the syndrome caused by SARS-CoV-2). [4] [5] [6] While older age is one of the most important risk factors for hospitalization and death, it is increasingly recognized that younger persons may develop severe disease.

The primary data source was JH-CROWN: The COVID-19 PMAP Registry, which utilizes the Hopkins Precision Medicine Analytics Platform. 8 Data in JH-CROWN include demographics, laboratory results, vital signs, respiratory events, medication administration, medical history, comorbid conditions, imaging, electrocardiogram results, and symptoms.

Our primary outcome was severe disease (including death), as defined by the WHO COVID-19 disease severity scale. 9 This is an 8-point ordinal scale ranging from ambulatory (1=asymptomatic, 2=mild limitation in activity), to hospitalized with mild-moderate disease (3=room air, 4=nasal cannula or facemask oxygen), hospitalized with severe disease (5=high flow nasal canula (HFNC) or non-invasive positive pressure ventilation (NIPPV), 6=intubation and mechanical ventilation, 7=intubation and mechanical ventilation and other signs of organ failure (hemodialysis, vasopressors, extracorporeal membrane oxygenation (ECMO)), and 8=death. Peak COVID-19 severity score is reported as the maximum score during the observation period for individual patients. Multi-comorbidity burden was assessed using the Charlson Comorbidity Index (CCI). 10 Diagnosis of COVID-19 was defined as detection of SARS-CoV-2 using any nucleic acid test with an Emergency Use Authorization from the US Food and Drug Administration. Samples predominantly included nasopharyngeal swabs and less commonly oropharyngeal swabs or bronchoalveolar lavage. Selection and frequency of other laboratory testing were determined by treating physicians. Natural Language Processing was used to identify presenting symptoms as described in the supplemental appendix (Table S1 ).

We estimated the cumulative incidence functions of death using the Aalen-Johansen estimator (CITE), with discharge and death as competing risks. 11 To assess the association between patient characteristics and outcomes, a set of 24 demographic and clinical variables were selected based on clinical interest and knowledge. Missing values were imputed using multiple imputation by chained equations (MICE) with predictive mean matching, 12 as implemented in the mice R package (version 3.7.0) 13,14 with 10 rounds of multiple imputation. Cox proportional-hazard models were used to relate the risk of (i) dying and (ii) developing severe disease or dying to baseline patient characteristics. 15 Four patients were discharged and then died. We censored their outcomes at time of discharge to minimize bias from lacking knowledge of deaths outside of the Hopkins system. We excluded the three pediatric patients (age <18 years) from the models. 92% of deaths occurred in patients over the age of 60, and almost half of those patients had a "do not resuscitate/no not intubate" (DNR/DNI) order within 24 hours of admission. We used a cox proportionalhazards model to relate the risk of severe illness or death to baseline characteristics in patients under the age of 60 in order to capture the risks of severe illness in that population.

Models were initially built adding variables in categorized "blocks" (e.g. "demographic") to protect against overfitting. For the composite outcome of severe disease or death, findings were equivalent to those from a model including all covariates. For other models, further variable selection was warranted when including multiple covariate blocks simultaneously. Here, we fit cause-specific proportional-hazards models regularized with an elastic net penalty, as implemented in the glmnet R package (version 3.0.2). 16 The elastic net model was run on each of the 10 imputed datasets, and variables with non-zero coefficients in at least half of the models were chosen for the final model, 17 which was again run on each of the 10 imputed datasets.

Demographic variables were forced to be in the models. No variable selection was done for the time to composite severe outcome or death model, as there were a sufficient number of events to allow for a larger model. Standard error estimates were computed using Rubin's rules (Rubin, 2004) , 18 as implemented in the mice R package. All analysis was done using R Version 3.6.2. 19

A total of 832 adult and pediatric patients were admitted with confirmed SARS-CoV-2 infection from March 4 to April 24 (Figure 1a) 

As shown in Table 1 , Figure 2 and Figure 3 , several characteristics distinguished peak illness states.

Increasing age (aHR 1.54 per 10-year increase; 95% CI 1.28-1.84), admission from a nursing home (aHR 2.13; 95% CI 1.41-3.23) and increasing CCI (aHR 1.13; 95%CI 1.02-1.26) were independently associated with death ( Table 2) . A sensitivity landmark analysis excluding those who achieved an outcome within 6 hours of admission is shown in Table S4 . Most associations were similar to the primary analysis but magnitudes of associations with respiratory and constitutional symptoms weakened. In a sub-group analysis of patients < 60 years of age, we identified male sex (aHR 1.7;95%CI 1.11-2.58), BMI (aHR 1.25 per 5-unit increase; 95%CI 1.14-1.37), CCI (aHR 1.27; 95%CI 1.1-1.46) and respiratory rate (aHR 1.16 per increase of 1 over 18; 95%CI 1.13-1.2) as significantly associated with severe illness or death ( Table 2, Table S5 ).

Our study provides valuable insight into the disease trajectories of hospitalized COVID-19 patients in the US and the risk factors associated with severe outcomes. Patients who developed severe illness and survived had a median length of stay of 15 days with 25% having a 20 day stay or longer. We observed an overall mortality of 14%, with nearly half of all deaths occurring among nursing home residents, many of whom had DNI/DNR orders on admission. Although >60% of patients were non-white, we did not observe statistically significant race/ethnicity associations. Obesity and overall comorbidities were significantly associated with severe illness or death, particularly in persons younger than 60 years. Lastly, we found a few simple-to-measure markers such as respiratory rate, D-dimer, and troponin to be strongly associated with death. Knowledge about disease trajectory and outcome is critical as providers, health systems and public health agencies plan for potentially scarce resources such as ventilators and therapeutics. 20, 21 It is also important to understand disease trajectory to determine appropriate levels of care, and when discussing goals of care with patients and families.

Increasing age was strongly associated with death; each decade increase had a 54% increased hazards of mortality. Similar age associations are now well described. 22, 23 Nursing home residents had a 2.1 fold increased hazards of death independent of age or comorbidity score, illustrating the vulnerability of this population to SARS-CoV-2. 24 There are approximately 1.4 million nursing home residents in the US. 25 It is estimated that one-third of US COVID-19 deaths are among this population. We found that 49% of deaths occurred in nursing home residents, similar to the 48% of deaths in this population that have been reported in Maryland. 26 Many of our older patients, including nursing home residents, had advanced medical directives and were DNR/DNI. This clearly impacted the level of intervention, measurement of severity (e.g. lack of mechanical ventilation) and time to death. The implementation of advanced directives varies substantially globally as does the age of the population. 27 Some of the global differences in SARS CoV-2 mortality are likely due to such differences.

Persons under 60 years of age comprised only 8% of deaths (9 cases) but accounted for the majority of severe illness outcomes among those who were discharged (95 cases; 82%). More than half of those with severe illness were obese, >60% were non-white, and 57% were male. We found that increasing BMI, CCI and male sex but not race/ethnicity were strongly associated with severe disease in those <60 years. The ageadjusted prevalence of obesity in the adult US population is 42.4%. Obesity prevalence is higher among African Americans and Hispanics and linked to socioeconomic status, other comorbidities, and poor health outcomes. 28 It is not surprising that that there is strong association between high BMI and severe COVID-19 in the US, particularly in younger age groups who are more likely to be obese. 29 The association between obesity and poor COVID-19 outcomes has also been reported internationally 30, 31 but the causal link remains unknown. All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . Mechanics of breathing may be impaired in obesity. Inflammation caused by excessive fat cells might worsen the hyperinflammatory response seen in COVID- 19. 32 Future studies of COVID-19 interventions should include this vulnerable population.

We found that a few simple-to-measure, baseline laboratory markers, namely absolute lymphocyte count (ALC), albumin, D-dimer and troponin, were associated with progression to severe disease or death.

Lymphopenia is highly prevalent in COVID-19, but its impact on mortality has been inconsistent across cohorts. 6 In our cohort lymphopenia was associated with increased illness severity or death, but not death alone. Low albumin was also associated with severe disease or death, a finding that has been previously observed. 33 Both elevated D-dimer and troponin were associated with a nearly 3-fold increased risk of death.

An elevated D-dimer is associated with increased risk of death in COVID-19 patients independent of documented thromboembolic disease, but could also indicate an increased risk of thrombosis. 34 An elevated troponin was also an important factor in models of severe disease or death. Whether SARS-CoV-2 leads to direct or indirect cardiac toxicity, it is clear that there is a link between cardiac injury and severe outcomes in COVID-19. 35 The presence of respiratory symptoms was associated with severe disease while the presence of constitutional symptoms seemed to be protective. This suggests that there are distinct phenotypes in COVID-19 that confer differential risk. Magnitudes of associations with respiratory and constitutional symptoms weakened in our landmark analysis potentially highlighting these symptoms as strong indicators of disease severity upon admission.

Lastly, we found that an elevated respiratory rate was associated with severe outcomes. This likely reflects that severity of illness in COVID-19 is tightly linked to pulmonary complications including acute respiratory distress syndrome (ARDS) and thromboembolism. Respiratory rate is included in mortality prediction scores for hospitalized patients 36,37, and has been associated with mortality in COVID-19. 38 The fact that such an easily measured parameter associates more strongly with severe outcome than several All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . https://doi.org/10.1101/2020.05.24.20111864 doi: medRxiv preprint inflammatory markers (e.g. CRP, ferritin) suggests that inexpensive and immediately available metrics can provide valuable information about disease trajectory.

There are some limitations to our study. Ten percent of the patients in our cohort did not yet have an observed outcome and such incompleteness could lead to bias. However, since we adopted time-to-event approaches which handled censored survival data, our analyses remain unbiased and not affected by incompleteness. Our data are derived from a single health system and may not be representative of COVID-19 populations across the US. Care practices may have differed between our 5 hospitals. We may have under ascertained the number of COVID-19 positive cases in our health system due to testing challenges. 39 We may not have captured all comorbidities since some patients may not have had robust documentation in the electronic health record. Post-discharge outcomes are not currently captured if they occur outside of the health system. Lastly, we had to impute a considerable percentage of missing values in several laboratory tests as there is no clear standard of care for laboratory testing in COVID-19.

In conclusion, we identified several important demographic and simple to assess factors associated with severe COVID-19 outcomes including age, nursing home status, BMI, D-dimer, troponin, ALC and respiratory rate. We also identified specific subgroups with a higher risk of disease progression including the elderly, nursing home residents, and younger patients with obesity.

The data utilized for this publication were part of the JH-CROWN: The COVID PMAP Registry, which is based on the contribution of many patients and clinicians. JH-CROWN received funding from Hopkins inHealth, the Johns Hopkins Precision Medicine Program.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . https://doi.org/10.1101/2020.05.24.20111864 doi: medRxiv preprint 1d. Patient trajectories among those who died. This figure illustrates the trajectory of each patient who died.

Each horizontal line represents a single patient. The black diamonds represent the day that a code status including either a do not resuscitate (DNR) or do not intubate (DNI) was entered into the medical record. Fiftyfive patients (49%) had a DNR/DNI order placed within 24 hours of admission (n=113 deaths).

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . https://doi.org/10.1101/2020.05.24.20111864 doi: medRxiv preprint Figure 2 . Cumulative incidence curves for death for key patient characteristics. This figure shows the cumulative incidence curves for death for key patient characteristics. In multivariate analyses increasing age (A), increasing BMI (C), Charlson Comorbidity Index (D), increasing respiratory rate greater (E), a D-dimer greater than 1 mg/L (G) and a detectable troponin (H) and were significantly associated with death. Decreasing absolute lymphocyte count (F) was significantly associated with progression to severe disease or death in a separate multivariate analysis. Male sex (B) was associated with severe disease or death in patients under the age of 60.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. WHO maximum disease state grouped by BMI for patients under 60 (n=364).

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . https://doi.org/10.1101/2020.05.24.20111864 doi: medRxiv preprint All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . https://doi.org/10.1101/2020.05.24.20111864 doi: medRxiv preprint

In order to identify symptoms at presentation, we first created a meta-lexicon of four symptom categories (organized into 11 sub-categories) based on the guidelines provided by the CDC, WHO, and clinical findings. For each symptom category, we generated a set of synonym terms using the Unified Medical Language System (UMLS) Metathesaurus, 1 and we iteratively worked with domain experts to revise the symptom categories and synonyms. Table S4 includes the list of symptom categories and the search terms.

We then selected relevant clinical note types for each patient, including H&P, Critical Care Notes, Progress Notes, and ED Notes, focusing specifically on the notes created within 48 hours before and after admission. Next, we pre-processed the note text and extracted only the relevant narrative parts, particularly the chief complaint and history of the present illness sections.

We then used a COVID-19-customized version of MedTagger, 2 together with our in-house Python tools to (a) identify phrases and synonyms of particular symptoms within the text narratives, (b) determine if these symptom mentions are negated, possible, or positive in their context, (c) classify symptoms into the predefined 11 categories, and (d) map them to their corresponding UMLS Concept Unique Identifiers (CUIs). These NLP pipelines use a combination of machine learning models, including Conditional random fields (CRFs), 3 and contextual rulebased methods, including regular expressions. Finally, we selected only the positive symptom mentions in the notes and aggregated all presenting symptoms for each patient.

To evaluate the performance of our NLP methods, two abstractors manually reviewed over 100 notes from 20 randomly selected patients. For each patient, each symptom was labeled as present or not-present (same label set as the NLP output), resulting in 220 manually labeled symptoms with the inter-rater agreement of 97%. The 3% disagreements were individually adjudicated between the two abstractors. Comparing the created gold standard to the labels generated by the NLP methods, we found that we could achieve the following results: (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Muscular/body aches body ache(s), general muscular aches and pains, generalized (acute) body aches, generalized chronic body aches, muscle aches, muscle discomfort, muscle pains, muscular aches, muscular discomfort, muscular pains, myalgia(s), myalgic, myodynia Sore throat painful swallowing, difficulty swallowing, pharyngeal discomfort, pharyngeal pain, pharynx discomfort, pharynx pain, sore throat, throat discomfort, throat pain, throat soreness All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 1 (0%) 1 (0%) 0 (0%) 0 (0%) Oseltamivir 2 (0%) 1 (0%) 1 (1%) 0 (0%) All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . Remdesivir a 8 (1%) 4 (1%) 4 (2%) 0 (0%) Ritonavir 1 (0%) 0 (0%) 1 (1%) 0 (0%) Rituximab 1 (0%) 0 (0%) 1 (1%) 0 (0%) Statins 271 (33%) 160 (30%) 68 (36%) 43 (38%) Tocilizumab 39 (5%) 4 (1%) 29 (16%) 6 (5%) a Patients were enrolled in a randomized trial. b Heparin and enoxaparin does not distinguish between prophylactic or therapeutic dosing.

All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . https://doi.org/10.1101/2020.05.24.20111864 doi: medRxiv preprint .73)* †Baseline physiologic features and lab findings include the first recorded value assessed within the first 48 hours of admission. Missing values were imputed using multiple imputation prior to model creation. *p < 0.05 All rights reserved. No reuse allowed without permission.

(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 26, 2020. . https://doi.org/10.1101/2020.05.24.20111864 doi: medRxiv preprint

1356) a Mild to moderate include patients with WHO ordinal score of 3 (not on oxygen) and 4 (on nasal cannula or face-mask oxygen). b Severe includes patients with WHO ordinal score of 5 (high-flow nasal cannula or non-invasive positive pressure ventilation), 6 (intubation and mechanical ventilation), and 7 (intubated, mechanical ventilation and other signs of organ failure including ECMO, hemodialysis or vasopressors). c Data was missing for race and ethnicity data in 8 patients, alcohol use in 106 patients, smoking history in 103 patients, and BMI for 16 patients. d Data was missing for admission vital signs for 11 patients, temperature for 13 patients, positive pressure ventilation use for 10 patients, and SaO2/FiO2 for 15 patients. e Data was missing for WBC for 1 patient, ALC for 11 patients, hemoglobin for 1 patient, platelet count for 2 patients, albumin for 13 patients, ALT for 15 patients, AST for 40 patients, bilirubin for 13 patients, creatinine for 2 patients, GFR for 4 patients, CRP for 220 patients, procalcitonin for 382 patients, LDH for 372 patients, D-Dimer for 249 patients, fibrinogen for 603 patients, ferritin for 271 patients, IL-6 for 651 patients, hemoglobin A1c for 580 patients, troponin for 150 patients, and Pro-BNP for 401 patients. References 1

Clinical Characteristics of 138 Hospitalized Patients With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan, China

Presenting Characteristics, Comorbidities, and Outcomes Among 5700 Patients Hospitalized With COVID-19 in the

Case-Fatality Rate and Characteristics of Patients Dying in Relation to COVID-19 in Italy

Risk Factors Associated With Acute Respiratory Distress Syndrome and Death in Patients With Coronavirus Disease

Hospitalization Rates and Characteristics of Patients Hospitalized with Laboratory-Confirmed Coronavirus Disease 2019 -COVID-NET, 14 States

Validation of a combined comorbidity index

An Empirical Transition Matrix for Non-Homogeneous Markov Chains Based on Censored Observations

Multivariate imputation by chained equations in R

Imputing missing covariate values for the Cox model

Regression Models and Life-Tables

Regularization paths for generalized linear models via coordinate descent

How should variable selection be performed with multiply imputed data

Multiple Imputation for Nonresponse in Surveys

The R Project for Statistical Computing. R Foundation for Statistical Computing, 2020

Emergency Use Authorization of Remdesivir: The Need for a Transparent Distribution Process

Fair Allocation of Scarce Medical Resources in the Time of Covid-19

Clinical course and risk factors for mortality of adult inpatients with COVID

China: a retrospective cohort study

Development and Validation of a Clinical Risk Score to Predict the Occurrence of Critical Illness in Hospitalized Patients With COVID-19

One-Third of All U.S. Coronavirus Deaths Are Nursing Home Residents or Workers. The New York Times

Do-not-resuscitate decisions in six European countries

Prevalence of obesity and severe obesity among adults: United States

Obesity could shift severe COVID-19 disease to younger ages

Association of higher body mass index (BMI) with severe coronavirus disease 2019 (COVID-19) in younger patients

High prevalence of obesity in severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) requiring invasive mechanical ventilation

Hypoalbuminemia predicts the outcome of COVID-19 independent of age and co-morbidity

High risk of thrombosis in patients with severe SARS-CoV-2 infection: a multicenter prospective cohort study

Association of Cardiac Injury With Mortality in Hospitalized Patients With COVID-19 in

A severity of disease classification system

The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine

Predictors of Mortality for Patients with COVID-19 Pneumonia Caused by SARS-CoV-2: A Prospective Cohort Study

Table S5: Association of clinical characteristics and severe outcomes, age < 60, demographics not forced into model Clinical characteristic † Severe outcome or death

The Unified Medical Language System (UMLS): integrating biomedical terminology

Desiderata for delivering NLP to accelerate healthcare AI advancement and a Mayo Clinic NLP-as-a-service implementation

All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder