key: cord-1018930-op8cbg14
authors: Meizlish, Matthew L.; Goshua, George; Liu, Yiwen; Fine, Rebecca; Amin, Kejal; Chang, Eric; DeFilippo, Nicholas; Keating, Craig; Liu, Yuxin; Mankbadi, Michael; McManus, Dayna; Wang, Stephen Y.; Price, Christina; Bona, Robert D.; Ochoa Chaar, Cassius Iyad; Chun, Hyung J.; Pine, Alexander B.; Rinder, Henry M.; Siner, Jonathan M.; Neuberg, Donna S.; Owusu, Kent A.; Lee, Alfred Ian
title: Intermediate‐dose anticoagulation, aspirin, and in‐hospital mortality in COVID‐19: A propensity score‐matched analysis
date: 2021-02-22
journal: Am J Hematol
DOI: 10.1002/ajh.26102
sha: 4c04a13fca800003c5fa0bb22440a1528b49362e
doc_id: 1018930
cord_uid: op8cbg14

Thrombotic complications occur at high rates in hospitalized patients with COVID‐19, yet the impact of intensive antithrombotic therapy on mortality is uncertain. We examined in‐hospital mortality with intermediate‐ compared to prophylactic‐dose anticoagulation, and separately with in‐hospital aspirin compared to no antiplatelet therapy, in a large, retrospective study of 2785 hospitalized adult COVID‐19 patients. In this analysis, we established two separate, nested cohorts of patients (a) who received intermediate‐ or prophylactic‐dose anticoagulation (“anticoagulation cohort”, N = 1624), or (b) who were not on home antiplatelet therapy and received either in‐hospital aspirin or no antiplatelet therapy (“aspirin cohort”, N = 1956). To minimize bias and adjust for confounding factors, we incorporated propensity score matching and multivariable regression utilizing various markers of illness severity and other patient‐specific covariates, yielding treatment groups with well‐balanced covariates in each cohort. The primary outcome was cumulative incidence of in‐hospital death. Among propensity score‐matched patients in the anticoagulation cohort (N = 382), in a multivariable regression model, intermediate‐ compared to prophylactic‐dose anticoagulation was associated with a significantly lower cumulative incidence of in‐hospital death (hazard ratio 0.518 [0.308–0.872]). Among propensity‐score matched patients in the aspirin cohort (N = 638), in a multivariable regression model, in‐hospital aspirin compared to no antiplatelet therapy was associated with a significantly lower cumulative incidence of in‐hospital death (hazard ratio 0.522 [0.336–0.812]). In this propensity score‐matched, observational study of COVID‐19, intermediate‐dose anticoagulation and aspirin were each associated with a lower cumulative incidence of in‐hospital death.

covariates in each cohort. The primary outcome was cumulative incidence of inhospital death. Among propensity score-matched patients in the anticoagulation cohort (N = 382), in a multivariable regression model, intermediate-compared to prophylactic-dose anticoagulation was associated with a significantly lower cumulative incidence of in-hospital death (hazard ratio 0.518 [0.308-0.872]). Among propensity-score matched patients in the aspirin cohort (N = 638), in a multivariable regression model, in-hospital aspirin compared to no antiplatelet therapy was associated with a significantly lower cumulative incidence of in-hospital death (hazard ratio 0.522 [0.336-0.812]). In this propensity score-matched, observational study of COVID-19, intermediate-dose anticoagulation and aspirin were each associated with a lower cumulative incidence of in-hospital death.

Thrombosis is among the most devastating complications of COVID- 19 . In multiple studies, venous thromboembolism (VTE), arterial thrombosis, and microvascular thrombosis have all been described. [1] [2] [3] [4] [5] [6] High VTE rates have been reported in critically ill COVID-19 patients despite the use of prophylactic anticoagulation. 1, 6, 7 The development of pulmonary microvascular thrombosis may be central to the pathogenesis of COVID-19 in the lungs. 5 An elevated D-dimer, a breakdown product of fibrin clots, is one of the strongest predictors of mortality from COVID-19. 8, 9 A common global practice has been to administer escalated intensities of antithrombotic therapy beyond standard prophylactic-dose anticoagulation in hospitalized COVID-19 patients. [10] [11] [12] To date, there has been little evidence to support this practice. 13, 14 Some retrospective studies have observed lower mortality rates with therapeutic-dose anticoagulation compared to either prophylacticdose anticoagulation or no anticoagulation, while others comparing therapeutic-and prophylactic-dose anticoagulation have found no mortality difference. [15] [16] [17] [18] [19] [20] [21] [22] Recent findings from the ACTIV-4a randomized controlled trial showed that therapeutic-compared to prophylactic-dose anticoagulation improved outcomes among non-critically ill patients but not among critically ill COVID-19 patients. 23, 24 To date, however, no large-scale study has compared the effects of intermediate-versus prophylactic-dose anticoagulation.

Some investigators have also proposed a potential role for aspirin and other antiplatelet therapies in light of the high burden of microvascular thrombosis and emerging models of endotheliopathy, platelet activation, and immunothrombosis in COVID- 19. 5,25-29 One retrospective study reported improved outcomes with aspirin therapy but did not account for disease severity between treatment groups, making its conclusions difficult to interpret. 30 A major limitation in retrospective studies is bias in the likelihood of patients to receive the treatments being studied. In unadjusted observational studies, disease severity is a confounding factor affecting treatment decisions and outcomes, often precluding accurate analysis of potential treatment effects. To address this, propensity score matching for disease severity and other variables has been utilized in some observational studies, leading to findings compatible with those obtained from randomized controlled trials. 31, 32 The use of propensity score matching in a few landmark observational studies in COVID-19

has yielded key insights about potential treatment effects by enabling treatment groups with balanced covariates to be reliably compared. 33, 34 We sought to examine the impact of intermediate-dose anticoagulation and aspirin on in-hospital mortality in COVID-19. To account for variations in treatment, we utilized propensity score matching and multivariable regression analysis incorporating markers of disease severity and other clinical covariates. One scoring system for assessing disease severity in use at many hospitals is the Rothman Index (RI), a composite score of 26 distinct clinical, laboratory, and nursing variables, which has been shown to have prognostic value in some surgical and critical care studies, although its utility in COVID-19 is presently unknown. [35] [36] [37] We hypothesized that the RI might be a useful tool for evaluating disease severity in COVID-19, both for clinical care and for the purposes of propensity score matching. In this observational study, we first analyzed a large, multisite cohort of hospitalized COVID-19 patients by multivariable regression analysis and found a novel prognostic role for the admission RI in predicting inhospital mortality. Then, incorporating the RI and other measures of illness severity, we performed propensity score matching and multivariable regression analyses and observed significant reductions in in- Demographic, clinical, and laboratory data were extracted from each patient's medical record by JDAT. Established population health registries were used to identify patients with diabetes, hypertension, coronary artery disease (CAD), and congestive heart failure (CHF) (Table S1 ). Inclusion into a population health registry required an encounter diagnosis of the referenced disease state and at least a single, non-abstract patient encounter within the health system in the preceding 3 years; in addition, either the patient problem list was required to contain the referenced diagnosis, or the patient had to have a minimum of two face-to-face encounters within the previous 12 months. For disease states without established population health registries, ICD-10 codes were used. We defined cardiovascular disease as any of the following: hypertension, diabetes, CAD, myocardial infarction, CHF, atrial fibrillation, stroke, or transient ischemic attack.

We categorized body mass index (BMI) according to the U.S. Centers for Disease Control definitions. 38 We categorized the first RI on admission into four quartiles (quartile 1, RI -33 to 43; quartile 2, RI 42 to 65; quartile 3, RI 66 to 79; quartile 4, RI 80 to 99), with the lowest and highest quartiles representing patients with the greatest and least illness severities, respectively. Patients who received any other dose of enoxaparin and who did not receive a direct oral anticoagulant (DOAC) or any other therapeuticdose anticoagulant were categorized as "Alternative enoxaparin dose". Patients who received a DOAC and no other type of therapeutic-dose anticoagulation were categorized as "DOAC". All other patients were categorized as "No documented anticoagulation".

Manual chart review was performed in cases with ambiguous data regarding anticoagulation dosing.

The primary outcome in this study was in-hospital death, measured as cumulative incidence of in-hospital death, with cumulative incidence of hospital discharge as a competing risk. Univariable and multivariable regression modeling of subdistribution hazard functions for the primary outcome was performed in all cohorts; we also reported hazard ratios (HR) from competing risks regression. 39 Variables incorporated into the modeling included demographic factors, medical history, and clinical and laboratory features reflecting disease severity.

Propensity score matching was performed on the different cohorts to achieve balance in covariates between patients treated with intermediate-versus prophylactic-dose anticoagulation in the anticoagulation cohort, and separately between patients treated with inhospital aspirin versus no antiplatelet therapy in the aspirin cohort.

Cumulative incidence curves were estimated for nonparametric visualization of in-hospital death and discharge events and tested using Gray's test in the propensity score-matched anticoagulation and aspirin cohorts 40 ; for clarity, only curves for in-hospital death are displayed in the figures.

Propensity scores were calculated within each cohort using multivariable logistic regression models. Propensity scores included covariates that may affect both the likelihood of patients to receive the treatment of interest and the outcome of interest, and that were unbalanced between treatment groups before matching. These variables included a number of patient characteristics as well as markers of disease severity. Matching based on propensity scores incorporating different sets of covariates was performed using a 1:1 nearest-neighbor algorithm, either with a caliper width of 0.25 (anticoagulation cohort) or without a caliper (aspirin cohort). In each analysis, the approach that yielded the best-matched cohort was identified based on the most balanced distribution of propensity scores and the best balance in individual covariates between the two treatment groups.

In March 2020, our hospital system established antithrombotic guidelines for management of hospitalized COVID-19 patients (Table S2) . The overall study cohort consisted of 2785 patients (Table S3) .

Half of patients were male (50.1%; N = 1396). The majority were over 60 years old (58.4%; N = 1627). Among all patients, 13.8% (N = 383) died in the hospital; 83.7% (N = 2330) were discharged alive, while 2.6% (N = 72) remained in the hospital at the time of data abstraction. We sought to identify variables significantly associated with disease severity in COVID-19 for use in propensity score matching. To achieve this, we performed multivariable analyses of the overall study cohort, examining associations of inhospital death with different variables (Table 1) . We observed a novel association of low admission RI quartile with increased cumulative incidence of in-hospital death in a model accounting for the competing risk of hospital discharge. Age > 60, male sex, obesity, and the maximum D-dimer level during hospitalization (DDmax)

were also significantly associated with in-hospital death, in keeping with prior studies. 41 (Table S4 ). Using this group of propensity score-matched patients, we fit a competing risks multivariable regression model adjusting for age, aspirin and antiplatelet therapy T A B L E 1 Multivariable analysis of in-hospital death in the overall study cohort Table 2) . Cumulative incidence curves also showed a significant reduction in in-hospital death among propensity score-matched patients in the anticoagulation cohort who were treated with intermediate-compared to prophylactic-dose anticoagulation (Figure 1(A) ).

Next, we explored the effects of in-hospital aspirin use. For this analysis, we established the "aspirin cohort", a nested cohort of patients from the overall study cohort who were not on home antiplatelet therapy prior to admission and received either aspirin or no antiplatelet therapy during their hospitalization (N = 1956). Within the aspirin cohort, we performed propensity score matching for age, DDmax, admission RI, and male sex, which achieved the most balanced distribution of covariates between patients treated with in-hospital aspirin compared to those who received no antiplatelet therapy ( Figure S2 ; Table S5 ).

Using this propensity score-matched group of 638 patients, we fit a competing risks multivariable regresson model adjusting for age, anticoagulation other than prophylactic dose, male sex, obesity, cardiovascular disease, African-American race, DDmax, and admission RI; in addition, we included ICU admission as a covariate based on a tendency of providers at our institution to administer aspirin preferentially to critically ill patients earlier on in the pandemic, before aspirin was added onto our hospital's treatment guidelines.

On multivariable analysis of propensity score-matched patients in the aspirin cohort, the use of in-hospital aspirin was associated with a lower cumulative incidence of in-hospital death (HR 0.522

[0.336-0.812]) (Table 3) .

Separately, we also analyzed outcomes of patients in the aspirin cohort who were admitted after May 18, the date on which our hospital system's antithrombotic guidelines added a recommendation to T A B L E 2 Multivariable analysis of in-hospital death in the propensity-score matched anticoagulation cohort administer aspirin to all hospitalized COVID-19 patients (Table S2) . For this analysis, we applied propensity score matching for age, DDmax, and admission RI score, which together yielded the most balanced distribution of covariates between aspirin-and non-aspirin-treated patients admitted after May 18, among the different combination of variables tested ( Figure S3 ). The final group of 140 propensity score-matched patients was well-balanced between aspirin-and non-aspirin-treated patients with respect to all variables except BMI ( (Table S7 ). Cumulative incidence curves showed a significant reduction in in-hospital death among propensity score-matched patients in the aspirin cohort admitted after May 18 who were treated with in-hospital aspirin compared to those who did not receive any antiplatelet therapy (Figure 1(B) ).

In our large observational study of hospitalized COVID-19 patients, we report, for the first time, a significantly lower cumulative incidence 48 Recently, one other retrospective study reported improved in-hospital mortality in COVID-19 patients who received aspirin within one week before or 24 h after admission. 30 Despite drawing similar conclusions, our study adopted a more rigorous methodological approach through the use of propensity score matching to account for differences in illness severity among patients, enabling us to more accurately compare different treatment effects. In addition, we excluded patients on home aspirin in order to minimize confounding effects from underlying cardiovascular disease. Presently, randomized controlled trials are underway to definitively address the question of whether aspirin may have an impact on outcomes in hospitalized COVID-19 patients.

Our analysis also reveals a novel role for the admission RI as a prognostic tool for evaluating the risk of in-hospital mortality in COVID-19. The RI, which synthesizes multiple clinical, laboratory, and nursing assessment variables into a single score, has been shown to have predictive value for assessing mortality and readmission rates in some critical care and surgical studies, although its applicability to T A B L E 3 Multivariable analysis of in-hospital death in the propensity-score matched aspirin cohort Note: Multivariable regression analysis was performed among propensity score-matched patients within the aspirin cohort to examine the association of in-hospital death with covariates. Cumulative incidence of in-hospital death was evaluated in a competing risks model with hospital discharge, and hazard ratios (HR) for in-hospital death were reported. For the maximum D-dimer level during hospitalization (DDmax), the hazard ratio represents the effect of an increase of one fibrinogen equivalent unit. Abbreviations: CI, 95% confidence interval; DDmax, maximum D-dimer level during hospitalization; DOAC, direct oral anticoagulant; ICU, intensive care unit; RI, Rothman Index.

COVID-19 has not been previously tested. [35] [36] [37] with potential disease-modifying effects outside of those included in our analysis, such as remdesivir or dexamethasone, may also have impacted clinical outcomes, although during the three-month span of our study, which represented the early phase of the pandemic, only a minority of patients in our overall study cohort would have been expected to receive either of these two therapies.

In summary, in our large, observational study of hospitalized patients with COVID-19, using propensity score matching and multivariable regression analyses, we observed a mortality benefit with intermediate-compared to prophylactic-dose anticoagulation and, separately, with in-hospital aspirin compared to no antiplatelet therapy. Our findings suggest that increased-intensity anticoagulation and antiplatelet therapy may be beneficial in the treatment of COVID-19.

We await the results of several randomized clinical trials to more definitively elucidate the impact of these therapies in COVID-19.

The authors would like to thank all providers, health care workers, and staff for their tireless dedication to the care of patients with COVID- 19 . We dedicate this work to all individuals afflicted by the pandemic.

This study was presented at an Oral Session at the 62nd Annual 

This study was approved by the Yale Institutional Review Board (HIC 2000027792).

Patient consent was not mandated for this study. Permission to reproduce material from other sources: No material from other sources is included in this study.

This study was not a clinical trial.

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.

COVID-19 and coagulation: bleeding and thrombotic manifestations of SARS-CoV-2 infection

Severe arterial thrombosis associated with Covid-19 infection

Risk of ischemic stroke in patients with coronavirus disease 2019 (COVID-19) vs patients with influenza

Pulmonary post-mortem findings in a series of COVID-19 cases from northern Italy: a two-Centre descriptive study

Pulmonary vascular Endothelialitis, thrombosis, and angiogenesis in Covid-19

Risk of venous thromboembolism in patients with COVID-19: a systematic review and metaanalysis

Incidence of venous thromboembolism in critically ill coronavirus disease 2019 patients receiving prophylactic anticoagulation

Factors associated with hospital admission and critical illness among 5279 people with coronavirus disease 2019 in new York City: prospective cohort study

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

ISTH interim guidance on recognition and management of coagulopathy in COVID-19: a comment

COVID-19 and its implications for thrombosis and anticoagulation

COVID-19: opening a new paradigm in thromboprophylaxis for critically ill patients?

A comprehensive assessment of the coagulation profile in critically ill COVID-19 patients

Anticoagulation in hospitalized patients with Covid-19

Anticoagulation, bleeding, mortality, and pathology in hospitalized patients with COVID-19

Association of Treatment Dose Anticoagulation with in-Hospital Survival among Hospitalized Patients with COVID-19

Abnormal coagulation parameters are associated with poor prognosis in patients with novel coronavirus pneumonia

Empiric therapeutic anticoagulation and mortality in critically ill patients with respiratory failure from SARS-CoV-2: a retrospective cohort study

The hazard of (sub)therapeutic doses of anticoagulants in non-critically ill patients with Covid-19: the Padua province experience

Increasing doses of anticoagulation are associated with improved survival in hospitalized COVID-19 patients

Thrombosis, bleeding, and the effect of anticoagulation on survival in critically ill patients with COVID-19 in the United States

Anticoagulant and antiplatelet use not associated with improvement in severe outcomes in COVID-19 patients

NIH ACTIV Trial of blood thinners pauses enrollment of critically ill COVID-19 patients

Full-dose blood thinners decreased need for life support and improved outcome in hospitalized COVID-19 patients

Immunothrombosis in severe COVID-19

Severe COVID-19 is a microvascular disease

Endotheliopathy in COVID-19-associated coagulopathy: evidence from a single-centre, crosssectional study

Platelet activation and platelet-monocyte aggregate formation trigger tissue factor expression in patients with severe COVID-19

Platelet gene expression and function in patients with COVID-19

Aspirin use is associated with decreased mechanical ventilation, ICU admission, and in-hospital mortality in hospitalized patients with COVID-19

Observational studies analyzed like randomized experiments: an application to postmenopausal hormone therapy and coronary heart disease

Propensity scores for confounder adjustment when assessing the effects of medical interventions using nonexperimental study designs

Observational study of hydroxychloroquine in hospitalized patients with Covid-19

Renin-angiotensinaldosterone system inhibitors and risk of Covid-19

Application of electronic medical record-derived analytics in critical care: Rothman index predicts mortality and readmissions in surgical intensive care unit patients

Rothman index variability predicts clinical deterioration and rapid response activation

Development and validation of a continuous measure of patient condition using the electronic medical record

Defining Adult Overweight and Obesity

A proportional hazards model for the subdistribution of a competing risk

A class of K-sample tests for comparing the cumulative incidence of a competing risk

Factors associated with death in critically ill patients with coronavirus disease 2019 in the US

D-dimer levels on admission to predict in-hospital mortality in patients with Covid-19

Thromboembolism and anticoagulant therapy during the COVID-19 pandemic: interim clinical guidance from the anticoagulation forum

Prevention, diagnosis, and treatment of VTE in patients with coronavirus disease 2019: CHEST guideline and expert panel report

Scientific and standardization committee communication: clinical guidance on the diagnosis, prevention, and treatment of venous thromboembolism in hospitalized patients with COVID-19

Anticoagulant treatment is associated with decreased mortality in severe coronavirus disease 2019 patients with coagulopathy

Higher intensity Thromboprophylaxis regimens and pulmonary embolism in critically ill coronavirus disease 2019 patients

SUPPORTING INFORMATION Additional supporting information may be found online in the Supporting Information section at the end of this article. How to cite this article