Trends in Venous Thromboembolism Anticoagulation in Patients Hospitalized With COVID-19

Key Points Question What is the frequency with which patients hospitalized with COVID-19 are treated with venous thromboembolism (VTE) prophylactic- and treatment-dose anticoagulation, and what is the association of anticoagulation with in-hospital and 60-day mortality? Findings In this cohort study of 1351 patients hospitalized with COVID-19 in which 1127 patients received anticoagulation, 34.8% missed 2 or more days of VTE prophylaxis. Use of only prophylactic-dose or treatment-dose anticoagulation was associated with lower in-hospital mortality vs no anticoagulation; however, only prophylactic-dose anticoagulation remained associated with lower mortality at 60 days. Meaning These findings suggest that prophylactic-dose VTE anticoagulation may be optimal therapy for patients hospitalized with COVID-19.


Introduction
Venous thromboembolism (VTE) has been a leading complication of COVID-19. 1 Early publications of high VTE rates likely influenced clinical practice related to VTE prophylactic-and treatment-dose anticoagulation. First, there has been a concerted emphasis on VTE prophylaxis for hospitalized patients with COVID-19. [2][3][4][5] Second, many experts have advocated for escalating doses of prophylactic anticoagulation for some patients hospitalized with COVID-19. 4,6,7 The potential importance of these practices has been highlighted further by a 2020 retrospective study 8 showing a potential mortality benefit with treatment-or prophylactic-dose anticoagulation. More recently, preliminary results from clinical trials have found a decrease in the combined outcome of in-hospital mortality and organ support free days with treatment-dose anticoagulation in patients outside of intensive care. 9,10 Given these findings, we sought to better understand variation in anticoagulation practices for patients hospitalized with COVID and the relationship of anticoagulation strategies with in-hospital and 60-day mortality.

MI-COVID19 Consortium
MI-COVID19 is a statewide collaborative quality initiative (CQI) sponsored by Blue Cross Blue Shield of Michigan and Blue Care Network. In March 2020, hospitals joined to collect patient-level data on COVID-19 patients with a goal of improving patient care. 11 Institutional participation in MI-COVID19 is voluntary, and was arranged through a special collaboration of hospitals participating in other Blue Cross Blue Shield-sponsored CQIs, including those with experience publishing data on VTE outcomes and anticoagulation. 12,13 Of the 92 noncritical access, nonfederal hospitals in Michigan, 38 (41%) elected to participate in MI-COVID19. MI-COVID19 hospitals are located across Michigan and have a median bed size of 391 (interquartile range [IQR], 250-537 beds); 81% are nonprofit and 93% selfidentify as teaching hospitals. There were no standardized treatment protocols provided as part of MI-COVID19. MI-COVID19 received nonregulated status prior to data collection by the University of Michigan institutional review board. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Patient Inclusion
Our primary cohort of interest was adults hospitalized for COVID-19 from March 7, 2020, to June 17, 2020. Patients were excluded if pregnant, aged under 18 years, discharged against medical advice, assigned comfort care on hospitalization, or transferred from another hospital. For patients with multiple hospitalizations, the first was included. As we were interested in patients newly started on empirical anticoagulation, we excluded 203 patients on treatment-dose anticoagulants prior to hospitalization and 28 patients with a VTE diagnosed within 2 days of hospitalization. Given that patients with short hospitalizations are less likely to benefit from anticoagulation, we excluded 224 patients hospitalized for less than 3 days. Hospitals with fewer than 10 cases were also excluded.
Based on available data collection resources, some MI-COVID19 hospitals were able to include all COVID-19-positive patients. Other hospitals (eg, with high volumes) employed a pseudorandom sampling process to select cases. Pseudorandomization involved sorting potentially eligible discharges by timestamp of discharge and reviewing patients for inclusion in ascending order based on the minute in which they were discharged (full pseudorandom sampling procedure available in eAppendix 3 in the Supplement). 11

Data Collection
Experienced, professional abstractors for other Blue Cross Blue Shield CQIs were retrained to collect data on COVID-19 patients for MI-COVID19. 14 Using standardized data templates and detailed data dictionaries, abstractors collected demographic data; comorbidities; daily treatment, laboratory, and stability data (eg, respiratory support); and outcomes (eg, mortality). Outcomes were also collected prospectively up to 60 days after hospitalization via medical record review and a postdischarge patient phone call (see Data Collection Outcomes). As disparities because of patient demographic characteristics may exist, we have reported data on sex, race, and ethnicity obtained from medical records and categorized as noted in eAppendix 2 of the Supplement.

Exposures
Abstractors collected anticoagulant administration data from each day of hospitalization. To ascertain VTE prophylaxis, abstractors were asked, "Was an anticoagulant administered for VTE prophylaxis on the date indicated?" To ascertain treatment-dose anticoagulation, abstractors were asked, "Was a treatment anticoagulant administered on the date indicated?" Abstractors determined whether anticoagulation was prophylactic-vs treatment-dose based on anticoagulant selection and common treatment and prophylactic doses. For example, low molecular weight heparin given 1 or 2 times per day at 30 to 40 mg and any subcutaneous heparin injection was considered prophylactic.
Intravenous heparin infusions could be prophylactic or therapeutic based on prescriber intent (see eAppendix 1 in the Supplement). Dosing of anticoagulation did not explicitly denote prescriber intent, as use of "intermediate" doses of anticoagulation were often recommended for highest-risk patients. 4,6,7

Outcomes
To assess in-hospital and 60-day mortality, abstractors reviewed the medical record 60 days following discharge to determine whether the patient was deceased. If so, date of death and-if available-cause(s) of death were abstracted. If no data were available or the patient appeared to be alive 60 days following discharge, the patient was contacted by phone up to 3 times to obtain additional outcome information. If the telephone respondent noted that the patient had died since hospitalization, they were asked for date and cause of death.
To assess in-hospital and 60-day VTE events, abstractors reviewed the medical record 60 days following discharge to determine whether there was a confirmed or suspected deep vein thrombosis or pulmonary embolism noted in the medical record and, if so, collect related imaging results. Only VTEs confirmed by imaging were considered VTE events. All data were entered into the MI-COVID19 registry using a structured data collection template.

Definitions Exposures
Anticoagulant exposure was categorized into 3 groups: (1) treatment-dose anticoagulation, defined as ever having received treatment-dose anticoagulation (for prophylactic intent) while hospitalized; (2) prophylactic-dose anticoagulation, defined as only receiving prophylactic-dose anticoagulation while hospitalized; and (3) no anticoagulation, defined as receiving neither treatment nor prophylactic anticoagulation while hospitalized. In those who received any anticoagulation, we also evaluated for nonadherence to VTE prophylaxis where any day in which a patient received neither prophylactic-nor treatment-dose anticoagulation was considered nonadherent.

JAMA Network Open | Health Policy
Trends in Venous Thromboembolism Anticoagulation in Patients Hospitalized With COVID-19

Outcomes
Our primary outcomes of interest were in-hospital or 60-day (from hospital admission) all-cause mortality. Sixty-day mortality included mortality captured by medical record review and from telephone calls 60 days following hospitalization, with events censored at 60 days after hospital admission. Our secondary outcome of interest was 60-day VTE events including pulmonary embolism or deep venous thrombosis confirmed by imaging.

Statistical Analysis
Descriptive statistics were used to characterize the cohort. Patient characteristics were compared by anticoagulation strategy using χ 2 , multisample median (ie, Brown-Mood), or t tests, as appropriate.
We assessed change over time in any anticoagulant use, nonadherence to VTE prophylaxis, and treatment-dose anticoagulation without imaging, using logistic regression models adjusted for hospital clustering and patient ICU status. We report adjusted odds ratios (aOR) for the effect of each additional week on anticoagulant use.
To evaluate the association of anticoagulant strategies with in-hospital and 60-day mortality, we used inverse probability of treatment models to control for variables associated with anticoagulant use. Similar to prior studies, 8  models see eTable 1 in the Supplement). 15 We then used IPTW cause-specific hazard models to determine the association of anticoagulant strategy with in-hospital and 60-day mortality. Survival in days was calculated as time from hospital admission to death up to 60 days postadmission. The adjusted hazard ratios (aHR) and their respective 95% CIs are reported for all time-to-event models.
A similar approach was used to evaluate the association of VTE prophylaxis adherence with mortality, where nonadherence was assessed first as a dichotomous exposure (ie, Ն2 days vs <2 days of nonadherence) and then as a continuous exposure (ie, percentage of inpatient days with nonadherence). Because low rates of 60-day VTE events, we report unadjusted event rates only.

Sensitivity Analyses
Because anticoagulant dosing could vary by day of hospitalization, we conducted a sensitivity analysis in which patients were categorized as prophylactic only until they had a confirmed VTE event, at which point they were categorized as treatment-dose anticoagulation. P values <.05 were considered significant in 2-sided tests. Analyses were completed using SAS version 9.4 (SAS Institute).  (Table). Only 162 patients (12.0%) received no prophylactic-or treatmentdose anticoagulation during their hospital stay (Figure 1).

VTE Prophylaxis
Of 1351 total patients, 1127 (83.4%) received pharmacologic VTE prophylaxis at some point during their hospital stay and were evaluated for nonadherence to VTE prophylaxis. Subcutaneous heparin or enoxaparin injections were the most common prophylactic regimens (eTable 2 in the Supplement). Of those who ever received treatment-or prophylactic-dose anticoagulation, approximately one-third (392 of 1127 patients [34.8%]) missed 2 or more days of VTE prophylaxis.

Discussion
In this large, multicenter cohort of patients hospitalized with COVID-19, use of treatment-dose anticoagulation was common, varied widely between hospitals, and increased over time. In addition to confirming recent findings suggesting both prophylactic-and treatment-dose anticoagulation strategies are associated with lower in-hospital mortality, we found that only prophylactic-dose anticoagulation was associated with lower 60-day mortality.
Critically, our study adds to existing literature on how anticoagulation strategies factor into mortality rates. First, it replicates similar findings in a retrospective, single-center New York study 8 that found both prophylactic-and treatment-dose anticoagulation were associated with lower in-hospital mortality. We add to this by including 60-day mortality data, which demonstrated an association with lower mortality for the prophylactic-dose anticoagulation group only. One potential reason for this finding may be that treatment-dose anticoagulation prevents in-hospital death from VTE or microvascular events in some patients who then succumb from severe illness after hospital discharge. Notably, we had far more treatment-dose anticoagulation in our intensive care population (31.1% vs 9.8%) and preliminary trial data suggest treatment-dose anticoagulation improves outcomes only in non-ICU patients. 9,10 Regardless, it appears anticoagulation is critical in hospitalized patients with COVID-19 and that prophylactic dosing may be sufficient to improve mortality.
While we found a low incidence of confirmed VTE (1.3%), the observed rate of therapeutic anticoagulation (16.2% overall, 31.1% in ICU) mirrors early reports of VTE incidence among hospitalized patients. 1 The high and increasing use of treatment-dose anticoagulation we observed (16.2%) likely reflects increased empirical anticoagulant use for suspected VTE and increased  treatment-dose anticoagulation for VTE prophylaxis. Empirical anticoagulation has been recommended by some when imaging cannot be obtained 3,4,16 ; for example, when patients are too critically ill to transport or when access to imaging staff is restricted to minimize infectious exposure.

JAMA Network Open | Health Policy
At the same time, early reports of thrombosis, coagulopathy, and diffuse pulmonary microvascular thrombi as potential contributors to mortality in COVID-19 led many experts to recommend intensification of anticoagulant regimens even without evidence of thrombosis. 17 While some clinicians and health systems use an intermediate dose of prophylaxis, others employ full therapeutic dosing. This approach is, in part, extrapolated from experience with H1N1 influenza. 18 Given the concern about VTE risk for patients hospitalized with COVID-19, national and international efforts have strongly emphasized the need for universal use and administration of VTE pharmacologic prophylaxis. [2][3][4][5] In our study, 12.0% of patients received no anticoagulationpotentially because they were considered low risk or had contraindications to treatment. Given what we know about coagulopathy in COVID-19 and growing evidence that anticoagulation may help hospitalized non-ICU patients the most, it is possible that even patients who seem low risk may benefit from anticoagulation. We also found evidence that adherence may be associated with 60-day mortality; prior studies have shown that missed doses of VTE prophylaxis are common and    associated with higher VTE rates. [19][20][21] Interestingly, the COVID-19 pandemic and focus on VTE prevention may represent one of the most effective implementation efforts of an evidence-based practice. In our multicenter cohort, the average number of patients with 2 or more days of missed prophylaxis medication declined from over 50% to less than 10% over just 3 months. Investigating how that implementation and dissemination effort can be replicated outside of a pandemic warrants further investigation.
Our findings have important implications. First, more data from randomized trials are needed on long-term outcomes of treatment-dose anticoagulation in patients without a confirmed VTE diagnosis. Second, VTE prophylaxis in patients with COVID-19 is standard of care. Hospitals should implement processes to ensure use of VTE prophylaxis for hospitalized patients with COVID-19.
Finally, the variable and increasing use of treatment-dose anticoagulation raises concerns especially given the lack of an association with 60-day mortality. We need better methods to risk stratify and diagnose patients with VTE and a stronger evidence-base on which to decide when to employ prophylactic vs therapeutic doses of anticoagulation for patients hospitalized with COVID-19.
Participation in ongoing clinical trials will help identify whether any patient groups may benefit from therapeutic doses of anticoagulation. Otherwise, given the lack of mortality difference between groups, judicious therapeutic dosing may be necessary.

Limitations
Our findings must be taken in the context of this study's limitations. First, as with all retrospective studies of VTE, particularly during the COVID-19 pandemic, we were limited by incomplete use of diagnostic tests for VTE. Second, we did not have bleeding outcomes, although historically this occurs in only 2% to 3% of patients. 8 Third, we were limited in our classification of treatment-vs prophylactic-dose anticoagulation. For example, while intravenous unfractionated heparin was assumed to be therapeutic, it is possible it was given at a subtherapeutic dose and our data abstraction did not include target or actual coagulation lab tests. Fourth, anticoagulation increased and mortality decreased over time. Although we attempted to control for time, residual confounding may persist. In contrast, key strengths are inclusion of multiple hospitals, ability to assess 60-day mortality and longitudinal trends, and detailed data collection.