key: cord-0929396-ik0dvsha
authors: Bhavani, Sivasubramanium V.; Verhoef, Philip A.; Maier, Cheryl L.; Robichaux, Chad; Parker, William F.; Holder, Andre; Kamaleswaran, Rishikesan; Wang, May D.; Churpek, Matthew M.; Coopersmith, Craig M.
title: Coronavirus Disease 2019 Temperature Trajectories Correlate With Hyperinflammatory and Hypercoagulable Subphenotypes
date: 2022-02-04
journal: Crit Care Med
DOI: 10.1097/ccm.0000000000005397
sha: 9e8e909d930b02badf8a44fb1d9e844b09d9be8a
doc_id: 929396
cord_uid: ik0dvsha

Body temperature trajectories of infected patients are associated with specific immune profiles and survival. We determined the association between temperature trajectories and distinct manifestations of coronavirus disease 2019. DESIGN: Retrospective observational study. SETTING: Four hospitals within an academic healthcare system from March 2020 to February 2021. PATIENTS: All adult patients hospitalized with coronavirus disease 2019. INTERVENTIONS: Using a validated group-based trajectory model, we classified patients into four previously defined temperature trajectory subphenotypes using oral temperature measurements from the first 72 hours of hospitalization. Clinical characteristics, biomarkers, and outcomes were compared between subphenotypes. MEASUREMENTS AND MAIN RESULTS: The 5,903 hospitalized coronavirus disease 2019 patients were classified into four subphenotypes: hyperthermic slow resolvers (n = 1,452, 25%), hyperthermic fast resolvers (1,469, 25%), normothermics (2,126, 36%), and hypothermics (856, 15%). Hypothermics had abnormal coagulation markers, with the highest d-dimer and fibrin monomers (p < 0.001) and the highest prevalence of cerebrovascular accidents (10%, p = 0.001). The prevalence of venous thromboembolism was significantly different between subphenotypes (p = 0.005), with the highest rate in hypothermics (8.5%) and lowest in hyperthermic slow resolvers (5.1%). Hyperthermic slow resolvers had abnormal inflammatory markers, with the highest C-reactive protein, ferritin, and interleukin-6 (p < 0.001). Hyperthermic slow resolvers had increased odds of mechanical ventilation, vasopressors, and 30-day inpatient mortality (odds ratio, 1.58; 95% CI, 1.13–2.19) compared with hyperthermic fast resolvers. Over the course of the pandemic, we observed a drastic decrease in the prevalence of hyperthermic slow resolvers, from representing 53% of admissions in March 2020 to less than 15% by 2021. We found that dexamethasone use was associated with significant reduction in probability of hyperthermic slow resolvers membership (27% reduction; 95% CI, 23–31%; p < 0.001). CONCLUSIONS: Hypothermics had abnormal coagulation markers, suggesting a hypercoagulable subphenotype. Hyperthermic slow resolvers had elevated inflammatory markers and the highest odds of mortality, suggesting a hyperinflammatory subphenotype. Future work should investigate whether temperature subphenotypes benefit from targeted antithrombotic and anti-inflammatory strategies.

dysregulated immune responses, coagulation abnormalities, and cardiac and renal dysfunction (2) (3) (4) (5) (6) . Identification of COVID-19 subphenotypes could lead to targeted treatment of these heterogeneous manifestations.

To date, studies have used static measurements of labs and vitals to identify subphenotypes (7) (8) (9) (10) (11) (12) . However, the host response to SARS-CoV-2 and other infections is a dynamic process with physiologic responses that evolve over hours (13) (14) (15) . We have previously shown that longitudinal body temperature trajectories can identify sepsis subphenotypes with distinct clinical characteristics and outcomes (16) . Importantly, these subphenotypes correlate with dynamic immune responses to infection over the course of hospitalization (17) . Temperature abnormalities are common in viral syndromes such as COVID-19 (18) (19) (20) , and the thermoregulatory response to infection operates at the intersection of the immunological, neurologic, cardiovascular, and other body systems (21) . Therefore, the universally available bedside measurement of temperature could provide key insights into the host response in COVID- 19 .

In a single-center pilot study, we found that temperature trajectories in COVID-19 were associated with different mortality rates (22) . The objectives of this current study are 1) to determine the best-fit temperature trajectory model for a multicenter cohort of hospitalized COVID-19 patients, 2) to evaluate the inflammation and coagulation biomarker profiles of the resulting subphenotypes, 3) to evaluate the association of these subphenotypes with hypercoagulable manifestations such as venous thromboembolic events (VTEs) and cerebrovascular accidents (CVAs), as well as the association with outcomes such as shock, respiratory failure, and 30-day inpatient mortality, and 4) to investigate whether the prevalence of the hyperthermic slow resolvers subphenotype (in our prior work, found to have elevated inflammatory markers) has decreased over the course of the pandemic with the widespread use of the anti-inflammatory dexamethasone.

We included all adult patients admitted to four academic hospitals in the Emory Healthcare system. Patients who were admitted between March 1, 2020, and February 28, 2021, were included if they had laboratory-confirmed COVID-19 and/or had a primary or secondary International Classification of Diseases, 10th Version (ICD-10) diagnosis of COVID-19 (U07.1). We excluded patients who were discharged or died within 24 hours of hospitalization, as we have shown in prior work that trajectory fit drops in patients who are discharged/die early (22) . We excluded patients who were transferred to a different hospital given incomplete encounter data. We included temperature measurements taken in the first 72 hours of hospitalization in the algorithm. Given that temperature measurements vary by site of measurement (e.g., oral, tympanic, etc.), only oral measurements were used for the analysis. The temperature measurements for study patients were standardized to the mean and sd of values for the subjects in the study. Patients with VTEs were identified using the following ICD-10 codes: I26, I80.1, I80.2, and I80.3 (23) . Patients with CVAs were identified using the following ICD-10 codes: I61, I62, I63, I69, and I67 (24) . Based on general impracticability and minimal harm, the Emory University Institutional Review Boards granted a waiver of consent for this study (STUDY00001627).

Our previously validated temperature trajectory model was derived on a cohort of 12,413 hospitalized patients with all-cause infection (e.g., viral, bacterial, fungal) and identified four distinct temperature patterns: 1) hyperthermic slow resolvers-patients with persistently elevated temperatures, 2) hyperthermic fast resolvers-patients with elevated presenting temperature with defervescence over the ensuing 72 hours, 3) normothermic-patients with normal body temperatures, and 4) hypothermic-patients with low body temperatures (16) . Each of the four temperature trajectories is a unique quadratic function describing temperature as a function of time from presentation to the hospital (i.e., Temperature = β 0 + β 1 × Time + β 2 × Time 2 ). As done in prior work, the study patients were classified to the trajectory that results in the lowest mean squared error (16, 17) . In addition to the validated model, we tested de novo models with varying number of classes (2-, 3-, 4-, 5-, and 6-class models) and selected the optimal de novo model using Bayesian information criterion, Akaike information criterion, and group membership. The selected de www.ccmjournal.org February 2022 • Volume 50 • Number 2 novo model's goodness-of-fit was compared with the validated model. The goodness-of-fit was calculated using the average mean squared error for the patients in each subphenotype. Between the de novo model and validated model, the model with the lower error was used for the remainder of analyses. After patients were classified into subphenotypes, the differences in demographics, comorbidities, and clinical characteristics between the subphenotypes were compared using analysis of variance (ANOVA) or chi-square tests, as appropriate.

Laboratory biomarkers were selected a priori for comparison between subphenotypes and were categorized as inflammatory, coagulopathic, or markers of organ dysfunction:

1) Inflammatory-C-reactive protein (CRP), erythrocyte sedimentation rate, WBCs, absolute lymphocytes, procalcitonin, ferritin, and interleukin-6 (IL-6). 2) Coagulopathic-d-dimer, fibrinogen, platelets, prothrombin fragment 1.2, thrombin-antithrombin complex, and fibrin monomers. 3) Organ dysfunction-Creatinine, total bilirubin, troponin, brain natriuretic peptide (BNP), and lactic acid.

If a patient had multiple measurements of a biomarker, the maximum value of that biomarker from the first 72 hours of hospitalization was used. No imputation process was used for patients with missing biomarkers. Non-normally distributed biomarkers were log-transformed. Biomarker levels were compared between subphenotypes using ANOVA. Given the prespecified set of 18 biomarkers, all tests of significance were corrected for multiple testing using Bonferroni correction.

The subphenotypes were evaluated for association with hypercoagulable outcomes: VTEs and CVAs. The subphenotypes were also evaluated for association with renal replacement therapy, vasopressors, mechanical ventilation, and 30-day inpatient mortality. Logistic regression was performed to evaluate these outcomes controlling for age, sex, race, comorbidities (congestive heart failure, chronic pulmonary disease, diabetes mellitus, hypertension, chronic renal disease, liver disease, and metastatic cancer), and maximum Sequential Organ Failure Assessment (SOFA) score in the first 72 hours of hospitalization. The hyperthermic fast resolvers were selected as the reference group based on our prior work, as this subphenotype has consistently demonstrated the lowest mortality rates (16, 17, 22) .

Given the widespread use of the anti-inflammatory dexamethasone after publication of the Randomized Evaluation of COVID-19 Therapy (RECOVERY) trial in July 2020 (25) , we hypothesized that we would see a contemporaneous decrease in the prevalence of the hyperthermic slow resolvers (in our prior work in sepsis, this subphenotype had elevated inflammatory markers).

To evaluate whether dexamethasone was associated with decreased odds of membership in the hyperthermic slow resolvers subphenotype, we employed propensity scores with inverse probability weighting. The outcome was membership in the hyperthermic slow resolvers subphenotype. The publication of the RECOVERY trial served as a natural experiment, with the control group consisting of patients admitted pre-RECOVERY trial (March 2020 to June 2020) who did not receive dexamethasone. The treatment group was patients admitted post-RECOVERY trial (August 2020 to February 2021) who received dexamethasone during the first 72 hours of hospitalization. Patients who received other corticosteroids were excluded. Patients who received dexamethasone may have confounders that are significantly different from patients who did not receive dexamethasone. To account for these measurable confounders, we created a propensity score for treatment with dexamethasone based on pretreatment characteristics that would be associated with subphenotype membership: age, sex, race, ethnicity, comorbidities, and admission SOFA score. We evaluated for distributional overlap of propensity scores to ensure common support between treatment and control. Inverse probability weighting was applied to the propensity scores to weigh patients such that the measured confounding variables were balanced between the treatment (dexamethasone) and control groups. We checked covariate balance to ensure that after adjustment, all absolute standardized differences between confounding variables were below the threshold of 0.1. Finally, we evaluated the probability weights to ensure there were no abnormally weighted patients (weight > 10) (26) (27) (28) .

To estimate the effect of dexamethasone treatment on subphenotype membership, a logistic regression model of treatment on hyperthermic slow resolvers subphenotype membership was fitted with application of the probability weights.

We tested whether the temperature trajectory subphenotypes could be identified using temperature data from the first 24 hours of hospitalization. Temperature measurements from each study patient in the first 24 hours were compared with the predicted measurement of each trajectory function, and patients were classified to the subphenotype resulting in the lowest mean squared error. The accuracy was evaluated using agreement in classification between the 24-hour and 72-hour models. Propensity score analyses were performed using Stata Version 16 (StataCorp, College Station, TX). All other analyses were performed using R Version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria).

The study cohort consisted of 5,903 hospitalized patients with COVID-19 (Appendix Fig. 1 , http:// links.lww.com/CCM/G926). The median age was 61 years (49-73 yr), with 7.1% prevalence of VTEs and 7.7% prevalence of CVAs during hospitalization. Furthermore, 12% required mechanical ventilation, 12% required vasopressors, and the 30-day inpatient mortality rate was 7.4% ( Table 1) . Our previously validated temperature trajectory model was compared with a de novo model and was found to have better fit (Appendix Table 1 and Appendix Fig. 2 , http://links. lww.com/CCM/G926). All trajectory analyses in the article are based on the previously validated model.

The distribution of subphenotype membership was: hyperthermic slow resolvers (n = 1,452, 25%), hyperthermic fast resolvers (1,469, 25%), normothermics (2,126, 36%), and hypothermics (856, 15%) (Fig. 1) . Individual patient trajectories are also presented in Figure 1 . Age was significantly different between subphenotypes (p < 0.001): hyperthermic slow resolvers were the youngest (median age, 58 yr; interquartile range [IQR], 46-69 yr), while hypothermics were the oldest (69 yr; IQR, 56-80 yr). Comorbidities were significantly different, with hypothermics having the highest burden of several comorbidities: 22% having congestive heart failure (p < 0.001), 18% chronic pulmonary disease (p = 0.02), 64% hypertension (p = 0.02), and 26% chronic kidney disease (p < 0.001).

The hyperthermic slow resolvers had the highest rates of mechanical ventilation (17%), renal replacement therapy (9.7%), and vasopressors (17%). Hyperthermic slow resolvers had the lowest prevalence of VTEs (5.1%), while the hypothermic patients had the highest prevalence (8.5%, p = 0.005). Hypothermics also had the highest prevalence of CVAs (10%, p = 0.001). Inpatient 30-day mortality rate was significantly different between subphenotypes (p = 0.02): 8.6% mortality rate for hypothermics, 8.5% for hyperthermic slow resolvers, 7.2% for normothermics, and 5.9% for hyperthermic fast resolvers ( Table 1) .

There were significant differences in levels of inflammatory markers between subphenotypes. The hyperthermic slow resolvers had the highest CRP, with a mean of 137 mg/L (95% CI, 132-142 mg/L), followed by hyperthermic fast resolvers (121; 95% CI, 117-126), normothermics (97; 95% CI, 93-101), and lowest levels in hypothermics (95; 95% CI, 89-101) (p < 0.001). The hyperthermic slow resolvers also had the highest levels of IL-6 and ferritin, while normothermic patients had the lowest levels (p < 0.001) ( Fig. 2; and Appendix Table 2 , http://links.lww.com/CCM/G926). The mean non-log-transformed levels of IL-6 were 19 pg/mL (13.9-24 pg/mL) in hyperthermic slow resolvers, 14 pg/mL (10.5-17.5 pg/mL) in hyperthermic fast resolvers, 12 pg/mL (8.6-15.5 pg/mL) in normothermic, and 13.5 pg/mL (8.4-18.6 pg/mL) in hypothermic. The missingness of the biomarkers by subphenotype and over the course of the study are presented in Appendix Tables 3 and 4 (http://links.lww.com/CCM/G926).

d-dimer levels were significantly different between the subphenotypes (p < 0.001), with the hypothermic patients having the highest levels, followed by normothermics, and then the hyperthermic subphenotypes. Hypothermic patients also had the highest levels of fibrin monomers (p < 0.001) (Fig. 2; and Appendix 

The markers of end-organ dysfunction were significantly different between the subphenotypes. The hypothermic subphenotype had the highest BNP and lactic acid (p < 0.001) ( Fig. 2; and Appendix Given the significant association between sex and trajectory subphenotype, we tested the interactions between sex and subphenotype in association with outcomes. We found that in hyperthermic slow resolvers, males had significantly decreased OR for mechanical ventilation (0.49; 95% CI, 0.26-0.90; p = 0.02) and decreased OR of vasopressor use (0.43; 95% CI, 0.23-0.80; p = 0.008) compared with females.

The percentage of patients who were classified as hyperthermic slow resolvers decreased over the course of the pandemic, from 53% of patients in March 2020; 23% of patients by August 2020; and less than 15% by 2021 (Fig. 4) . There was a contemporaneous increase in dexamethasone use: less than 5% receiving dexamethasone from March to May 2020, 14% in June 2020, and over 60% in the subsequent months. We hypothesized that increasing dexamethasone use was associated with the decreasing prevalence of hyperthermic slow resolvers, and we evaluated this association through propensity score matching with inverse probability weighting. The 2,217 treated patients (i.e., patients in the post-RECOVERY trial period who received dexamethasone in the first 72 hr of hospitalization) were matched to 899 control patients (i.e., patients in the pre-RECOVERY trial period who did not receive dexamethasone). The propensity model had acceptable distributional overlap of propensity scores, well-balanced covariates, and no disproportionally weighted patients (Appendix Figs. 4 and 5 , http://links.lww.com/CCM/ G926). After inverse probability weighting, dexamethasone treatment was associated with a significant reduction in probability of hyperthermic slow resolvers membership (27% reduction; 95% CI, 23-31%; p < 0.001). We performed the following sensitivity analyses: 1) We included patients treated with any corticosteroid within the first 72 hours and found that treatment with any corticosteroid was associated with a significant reduction in probability of hyperthermic slow resolver membership (26%, p < 0.001). 2) We restricted the treated and control cohort to patients requiring supplemental oxygen and found that dexamethasone was associated with a significant reduction in hyperthermic slow resolver membership (29%, p < 0.001). 3) We limited the treatment cohort to patients receiving dexamethasone in the first 24 hours of hospitalization and found a significant reduction in hyperthermic slow resolvers membership (31% reduction, p < 0.001).

The changes in proportions of other subphenotypes and treatments over the study period are presented (Appendix Figs. 6 and 7 and Appendix Table 6 , http:// links.lww.com/CCM/G926). Odds ratio (OR) for hospital outcomes compared between temperature trajectory subphenotypes. The temperature trajectory subphenotypes were evaluated for association with venous thromboembolic events (VTEs), cerebrovascular accidents (CVAs), renal replacement therapy (RRT), mechanical ventilation, vasopressors, and 30-d inpatient mortality. Logistic regression was performed to evaluate these outcomes controlling for age, sex, race, comorbidities, and maximum Sequential Organ Failure Assessment score in the first 72-hr of hospitalization. The hyperthermic fast resolvers were selected as the reference group based on our prior work, as this subphenotype has consistently demonstrated the lowest mortality rates. Hypothermics and normothermics had increased OR of CVAs 

Patients were classified into temperature trajectory subphenotypes using temperature data from the first 4 hours (i.e., presenting temperature) and the first 24 hours of hospitalization (Appendix Fig. 8 , http://links. lww.com/CCM/G926). There was poor agreement (51%) between the presenting temperature and the 72-hour model. There was 74% agreement in subphenotype classification between the 24-hour and 72-hour models. The accuracy incrementally improved with higher frequency of temperature measurements within the 24-hour period (Appendix Fig. 9 , http://links.lww. com/CCM/G926).

We present the novel discovery of a hypercoagulable temperature trajectory subphenotype and a hyperinflammatory subphenotype in COVID-19. Using longitudinal body temperature, we discovered that the hypothermic subphenotype had marked abnormalities in coagulation biomarkers and the highest prevalence of VTEs and CVAs. As far as we know, this is the first study to demonstrate a direct clinical association between body temperature and coagulation abnormalities in COVID-19. The hyperthermic slow resolvers had elevated CRP, ferritin, and IL-6 and higher frequency of cardiovascular failure requiring vasopressors and respiratory failure requiring mechanical ventilation. The hyperthermic slow resolvers had significantly higher mortality compared with hyperthermic fast resolvers. Interestingly, we also observed that the prevalence of the hyperthermic slow resolvers drastically decreased over the course of the pandemic. COVID-19 disease is associated with significant perturbations to the hematological system and often results in a hypercoagulable state (29) . However, it is still unclear which subgroups of COVID-19 patients would benefit most from higher intensity anticoagulation (30, 31) . Prior studies have suggested a link between lower temperatures and coagulation abnormalities (32) (33) (34) . In our study, the hypothermic subphenotype had elevated d-dimer and fibrin monomer levels. Importantly, this subphenotype had significantly higher unadjusted prevalence of VTEs and CVAs. The prevalence of CVAs remained significant even when controlling for age, comorbidities, and severity of illness. Given the marked elevation in levels of d-dimer and fibrin monomers, we hypothesize that hypothermic patients may need more aggressive screening for VTEs and that the true prevalence may be underdiagnosed. Further research is needed to evaluate whether hypothermic patients benefit from therapies targeting specific perturbations in hemostasis.

The host response to COVID-19 is heterogeneous, and patients may demonstrate both pro-and anti-inflammatory responses (3, (35) (36) (37) . Given this heterogeneity, anti-inflammatory therapies may benefit certain groups, while harming others (38) . Although there is no gold standard definition of hyperinflammation, hyperthermic slow resolvers demonstrated several abnormalities consistent with a hyperinflammatory Therapy (RECOVERY) trial in July 2020, we hypothesized that we would see a contemporaneous decrease in the prevalence of the hyperinflammatory temperature subphenotype (the hyperthermic slow resolvers). Dexamethasone use (as a percentage of admitted patients per month that received the medication) and the prevalence of hyperthermic slow resolvers per month are presented over the course of the study. Dexamethasone use steadily increased over the course of the pandemic with a large rise in July coinciding with the publication of the RECOVERY trial (marked by the dashed black line). The percentage of patients with the hyperthermic slow resolver temperature pattern steadily decreased over the course of the pandemic, from over 50% in March 2020 to less than 15% by 2021.

state (39) (40) (41) (42) (43) . CRP and ferritin have been commonly used to define COVID-19-associated hyperinflammation, and hyperthermic slow resolvers had the highest levels of both biomarkers (40) . Sinha et al (10) have reported on using IL-6 and vasopressor need to identify an acute respiratory distress syndrome hyperinflammatory subphenotype. Hyperthermic slow resolvers had elevated IL-6 levels and the highest frequency of vasopressor use. These findings are consistent with our prior observations in hyperthermic slow resolvers with Staphylococcus aureus bacteremia and all-cause septic shock, in which we found that this subphenotype had persistent elevations in inflammatory cytokines (17) . Future research should investigate whether this subphenotype would benefit from anti-inflammatory therapies including tocilizumab. Interestingly, there were significant interactions between sex and the hyperthermic slow resolver subphenotype. Specifically, females had increased OR for mechanical ventilation and vasopressor use in this subphenotype. This suggests that the increased risks of poor outcomes associated with hyperthermic slow resolvers may be predominantly in females following this temperature pattern. Further research is needed to study the complex interactions between sex and body temperature responses to infection.

Over the course of the pandemic, we have seen drastic shifts in standards of care. In our center, we observed that the percentage of inpatients receiving dexamethasone increased from less than 5% in the beginning of the pandemic to over 60% after publication of the RECOVERY trial. Interestingly, we found a contemporaneous decrease in the prevalence of hyperthermic slow resolvers. Prior studies have shown that dexamethasone can change temperature patterns and may benefit hyperinflammatory phenotypes (44) (45) (46) . Given that the hyperthermic slow resolvers are a hyperinflammatory subphenotype, we hypothesize that the use of dexamethasone is contributing to the shifting of this temperature pattern. However, the decreasing prevalence of hyperthermic slow resolvers began before the publication of the RECOVERY trial, and thus cannot be completely explained by dexamethasone usage. This shift in temperature patterns could also be due to other changes in clinical practices (such as the use of remdesivir, which increased in parallel with the use of dexamethasone), changes in demographics of infected patients, and changes in the pathogen itself.

Further research is necessary to investigate whether corticosteroids affect temperature patterns through immunological changes.

The study has several limitations. First, the retrospective nature limits the capacity for causal inferences. For instance, our results suggest that dexamethasone contributed to the observed changing temperature patterns over the course of the pandemic, but there could be other interventions such as IV fluids that contribute to changes in temperature trajectories. Second, the biomarkers were collected as clinically indicated resulting in missing data. Importantly, missing data varied by subphenotype, suggesting that the clinical decision to check the biomarker may have been influenced by the differential presentations of these subphenotypes. Third, the temperature trajectory model currently requires 72 hours of temperature data and cannot be accurately determined by the initial measurements at presentation, which may limit immediate clinical implementation. As supported by our data, future work investigating higher frequency temperature measurements may allow early identification of these subphenotypes.

Using subphenotypes derived from the universally available measurement of temperature, we found that hyperthermic slow resolvers were hyperinflammatory and at high risk for respiratory failure, shock, and mortality, while hypothermics were hypercoagulable with marked coagulation abnormalities. Future work should investigate whether temperature trajectory subphenotypes have differential responses to treatment.

World Health Organization: Novel Coronavirus (COVID-19) Situation

The trinity of COVID-19: Immunity, inflammation and intervention

For the B, Translational Science Committee of the Research Section for the Society of Critical Care M: Severe acute respiratory syndrome-associated coronavirus 2 infection and organ dysfunction in the ICU: Opportunities for translational research

Clinical phenotypes of SARS-CoV-2: Implications for clinicians and researchers

Coagulation dysfunction in COVID-19: The interplay between inflammation, viral infection and the coagulation system

Cytokine storm syndrome in coronavirus disease 2019: A narrative review

Phenotypes and subphenotypes of patients with COVID-19: A latent class modeling analysis

COVID-19 group and COVID@HULP groups: Identification and validation of clinical phenotypes with prognostic implications in patients admitted to hospital with COVID-19: A multicentre cohort study

Clinical phenotypes of critically ill COVID-19 patients

Prevalence of phenotypes of acute respiratory distress syndrome in critically ill patients with COVID-19: A prospective observational study

Data Science Collaborative Group: Differences in clinical deterioration among three sub-phenotypes of COVID-19 patients at the time of first positive test: Results from a clustering analysis

COVID-19 SEMICYUC Working Group: Deploying unsupervised clustering analysis to derive clinical phenotypes and risk factors associated with mortality risk in 2022 critically ill patients with COVID-19 in Spain

A dynamic COVID-19 immune signature includes associations with poor prognosis

Longitudinal characteristics of lymphocyte responses and cytokine profiles in the peripheral blood of SARS-CoV-2 infected patients

Stages or phenotypes? A critical look at COVID-19 pathophysiology

Identifying novel sepsis subphenotypes using temperature trajectories

Temperature trajectory subphenotypes correlate with immune responses in patients with sepsis

Fever in the ICU: A predictor of mortality in mechanically ventilated COVID-19 patients

National Centre for Infectious Diseases COVID-19 Outbreak Research Team: Fever patterns, cytokine profiles, and outcomes in COVID-19

Body temperature correlates with mortality in COVID-19 patients

Central circuitries for body temperature regulation and fever

Novel temperature trajectory subphenotypes in COVID-19

Thromboembolism and the Oxford-AstraZeneca COVID-19 vaccine: Side-effect or coincidence?

Association of stroke among adults aged 18 to 49 years with long-term mortality

Dexamethasone in hospitalized patients with Covid-19

Using balance statistics to determine the optimal number of controls in matching studies

Comparison of propensity score methods and covariate adjustment: Evaluation in 4 cardiovascular studies

Alternative approaches for confounding adjustment in observational studies using weighting based on the propensity score: A primer for practitioners

Supported by the ESC Working Group on Pulmonary Circulation and Right Ventricular Function: COVID-19 and thrombotic or thromboembolic disease: Implications for prevention, antithrombotic therapy, and follow-up: JACC stateof-the-art review

Effect of intermediate-dose vs standard-dose prophylactic anticoagulation on thrombotic events, extracorporeal membrane oxygenation treatment, or mortality among patients with COVID-19 admitted to the intensive care unit: The INSPIRATION randomized clinical trial

National Heart, Lung, and Blood Institute: ATTACC, ACTIV-4a & REMAP-CAP Multiplatform RCT: Results of Interim Analysis. 2021. Available at

Hypothermia: Effects on platelet function and hemostasis

Effects of low temperature on shear-induced platelet aggregation and activation

Sustained hypothermia accelerates microvascular thrombus formation in mice

Immunotherapies for COVID-19: Lessons learned from sepsis

Severe immunosuppression and not a cytokine storm characterizes COVID-19 infections

Pro-and anti-inflammatory responses in severe COVID-19-induced acute respiratory distress syndrome-an observational pilot study

Should we stimulate or suppress immune responses in COVID-19? Cytokine and anti-cytokine interventions

Temple University COVID-19 Research Group: Preliminary predictive criteria for COVID-19 cytokine storm

COVID-19-associated hyperinflammation and escalation of patient care: A retrospective longitudinal cohort study

Delving beneath the surface of hyperinflammation in COVID-19

Association of inflammatory markers with the severity of COVID-19: A meta-analysis

COREG Investigators: Inflammatory biomarkers as independent prognosticators of 28-day mortality for COVID-19 patients admitted to general medicine or ICU wards: A retrospective cohort study

COVID-19 and non-COVID ARDS patients demonstrate a distinct response to low dose steroids-a retrospective observational study

Corticosteroid therapy is associated with improved outcome in critically ill COVID-19 patients with hyperinflammatory phenotype

Impact of corticosteroids in hospitalised COVID-19 patients