Outcomes in patients discharged with extended venous thromboembolism prophylaxis after hospitalization with COVID-19

This retrospective analysis included adult patients (⩾ 18 years) who needed hospitalization and had an index severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) polymerase chain reaction (PCR) positive test between March 2020 and December 2020, within a health system comprising of 11 hospitals in Minnesota and Western Wisconsin. Patients were excluded for any of the following: pregnancy at the time of hospitalization, death, diagnosis of VTE, major bleeding event during hospitalization, on therapeutic or prophylactic anticoagulation for any indication before hospitalization, and patients who declined use of their medical records for health research. Data were abstracted from electronic medical records to include demographics as well as clinical characteristics by members of the organization’s clinical research informatics team. This study was approved by the Allina Heath Institutional Review Board (study #1661097).

Patients were placed in two study groups based on prescribed therapy at the time of discharge (extended prophylactic anticoagulation, no anticoagulation). Extended prophylactic anticoagulation was defined as any of the following medications: unfractionated heparin 5000 units subcutaneous twice or three times per day, enoxaparin 40 mg subcutaneous once or twice per day (if body mass index > 40), and rivaroxaban 10 mg once per day. The primary clinical outcomes of interest were VTE at 30 and 90 days posthospital discharge. Secondary outcomes of interest were mortality at 30 and 90 days, superficial venous thrombosis (SVT), acute myocardial infarction (MI), acute ischemic stroke, clinically significant bleeding (defined as any bleeding needing clinic, urgent care, or emergency department visit, a procedure to control bleeding, or any amount of packed red blood cell transfusion), and inpatient readmission all within 30 days of discharge. Demographic characteristics included age, sex, patient-reported race and ethnicity, and history of tobacco use. Additional clinical characteristics included weight, major comorbid conditions, medications prior to hospitalizations, anticoagulation use during hospitalization, in-hospital supplemental oxygen needs, admission to intensive care unit (ICU), extracorporeal membrane oxygenation (ECMO), laboratory values of D-dimer, C-reactive protein (CRP), ferritin within 72 hours of hospitalization, and the Severity of Illness (SOI) and Risk of Mortality (ROM) modifiers assigned by the All Patient Refined Diagnosis Related Groups (APR-DRG). The SOI is defined as ‘the extent of organ system loss of function or physiologic decompensation’ whereas the ROM is the likelihood of in-hospital mortality based on age, diagnoses, and procedures (both are coded as minor, moderate, major, and extreme).^15,16

Patients were compared to examine differences in outcomes of interest between those prescribed prophylactic anticoagulation and those not prescribed prophylactic anticoagulation at discharge. Descriptive statistics were used to summarize baseline and hospital-related characteristics among all study participants. Chi-squared, Fisher’s exact, and Wilcoxon rank sum tests were used, as appropriate, to compare differences in baseline and hospital-related characteristics between study groups.

To examine the association between prophylactic anticoagulation at discharge and clinical outcomes of interest, multivariable Firth’s logistic regression models were used to produce odds ratios (OR) with 95% CI. Firth’s logistic regression was used to reduce biases in predictions due to the small number of events occurring in models and separation of data points.^17,18 Additionally, we examined differences in all-cause mortality between study groups. Survival was calculated from the date of discharge to the date of death within the first 30 and 90 days. Multivariable Cox proportional hazard models were used to examine the association between prophylactic anticoagulation at discharge and risk of death within 30 and 90 days from hospitalization with COVID-19.

Covariates for all the aforementioned multivariable models included age, sex, race, severity of illness, in-hospital anticoagulation, remdesivir use, steroid use, any oxygen supplementation during hospitalization, and discharge disposition. The variables age, sex, race, severity of illness index, in-hospital anticoagulation, and discharge disposition were selected as covariates a priori.^19–22 The severity of illness index incorporates comorbidities related to the diagnosis and hospital course of care, and therefore individual comorbidities were not included in multivariable models.¹⁵ In-hospital remdesivir and steroid use were included due to the significant differences in proportions of patients receiving these therapies between study groups and any potential effect these therapies may have on clinical outcomes.^23,24 Oxygen supplementation was included as a potential marker for severity of COVID-19 infection and due to the potential effect on outcomes.^20,25,26 Owing to a very small number of patients in our study who received tocilizumab and convalescent plasma, these variables were not included in multivariable models. Analyses were performed using SAS for Windows, Version 9.4 (Cary, NC, USA) and p-values less than 0.05 were considered statistically significant.

There are limitations of this study that should be noted. The nonrandomized design and retrospective nature is a major limitation. Although we had a large sample size of patients, the outcomes of interest were relatively rare. Future studies with larger sample sizes should continue to examine these outcomes. Our study spans over a time when our understanding of COVID-19 was evolving, and universal testing of all hospitalized patients was not done at our institution due to scarcity of tests. Additionally, some therapies (e.g., steroids) were not shown to improve survival yet. As such, some patients with COVID-19 may have been inadvertently excluded due to unknown status, and in-hospital treatments may have changed over time. After some consideration, we decided to include all hospitalized patients with index SARS-CoV-2 PCR positivity during the study period, which could have led to the inclusion of patients where COVID-19 positivity was asymptomatic. It is unknown if asymptomatic COVID-19 positivity may contribute to any additional VTE risk in hospitalized patients during and after hospitalization. Though education sessions and discharge order guidance (Figure 1) were available to providers about when to consider extended VTE prophylaxis, it was up to the discharging provider regarding to which patients to prescribe extended VTE prophylaxis and, as such, selection bias in prescribing extended VTE prophylaxis should be considered. Providers were guided to provide extended VTE prophylaxis based on the IMPROVE VTE score and D-dimer values outlined in the Introduction, but we did not measure whether patients actually received prophylaxis based on advice – so again, selection bias remains a limitation. Furthermore, despite our study being generally representative of the United States’ racial and ethnic populations, our study population was primarily non-Hispanic White, and future work with oversampling of patients from other racial and ethnic backgrounds will enhance the generalizability of results.⁴⁶ Finally, this study included subjects treated at the time when milder COVID-19 variants were not prevalent and when vaccination was not widely available for health care workers and high-risk patients. The impact of vaccination on the outcomes cannot evaluated.