Pulmonary embolism and deep venous thrombosis after COVID-19: long-term risk in a population-based cohort study

Background Venous thromboembolism (VTE) (pulmonary embolism [PE] or deep venous thrombosis [DVT]) is common during acute COVID-19. Long-term excess risk has not yet been established. Objectives To study long-term VTE risk after COVID-19. Methods Swedish citizens aged 18 to 84 years hospitalized and/or testing positive for COVID-19 between January 1, 2020, and September 11, 2021 (exposed), stratified by initial hospitalization, were compared to matched (1:5), nonexposed, population-derived subjects without COVID-19. Outcomes were incident VTE, PE, or DVT recorded within 60, 60 to <180, and ≥180 days. Cox regression was used for evaluation, and a model adjusted for age, sex, comorbidities, and socioeconomic markers was developed to control for confounders. Results Among exposed patients, 48,861 were hospitalized for COVID-19 (mean age, 60.6 years) and 894,121 were without hospitalization (mean age, 41.4 years). Among patients hospitalized for COVID-19, fully adjusted hazard ratios during 60 to <180 days were 6.05 (95% CI, 4.80-7.62) for PE and 3.97 (CI, 2.96-5.33) for DVT compared with that for nonexposed patients with corresponding estimates among those with COVID-19 without hospitalization 1.17 (CI, 1.01-1.35) and 0.99 (CI, 0.86-1.15), based on 475 and 2311 VTE events, respectively. Long-term (≥180 days) hazard ratios in patients hospitalized for COVID-19 were 2.01 (CI, 1.51-2.68) for PE and 1.46 (CI, 1.05-2.01) for DVT, while nonhospitalized patients had similar risk as nonexposed patients, based on 467 and 2030 VTE events, respectively. Conclusion Patients hospitalized for COVID-19 retained an elevated excess risk of VTE, mainly PE, after 180 days, while long-term risk of VTE in individuals with COVID-19 without hospitalization was similar to that in the nonexposed patients.


| I N T R O D U C T I O N
In most people infected with SARS-CoV-2, the clinical course is relatively mild and without sequelae, but serious pulmonary complications requiring hospitalization, including intensive care, occur in a subset.
Currently, almost 3 years after its first occurrence, complications during follow-up are being increasingly recognized.
Thromboembolic events, such as pulmonary embolism (PE) and deep venous thrombosis (DVT), were promptly recognized as common complications in severe COVID-19 [1][2][3][4][5]. The expected baseline rate of venous thromboembolism (VTE) is 1 to 3 cases per 1000 individuals per year, with a predominance of DVT over PE [6,7], and associated with a number of well-established risk factors such as older age [8][9][10]. Early evaluations rating PE as the most prevalent VTE in COVID-19 [1,2,5,11] were later contested in meta-analyses [4,12], but the distribution between PE and DVT has not yet been reliably described. Also, the increase in PE raised concerns about the risk of persistent symptoms, which may affect around half of PE survivors [13]. Increasing evidence demonstrated an association between the severity of acute COVID-19 and complications, indicating a potential long-time elevated risk of VTE in a considerable number of patients [14,15].
While the short-term high risk of VTE in connection with severe COVID-19 is well known, the extent to which elevated risk persists in nonsevere cases of infection has not yet been reliably established. A prior Swedish matched-cohort study, using the same registers available to us, found high risks of VTE in mild, nonhospitalized cases, although much higher among the severely affected [16], with persistent high rates 2 to 4 months after diagnosis. However, only a limited number of observations were available for long-term analyses. Also, VTE was included as part of a recent British cohort study of cardiometabolic outcomes up to 1 year after COVID-19, with similar results, although with VTE analyzed in conjunction with other cardiovascular diseases (CVDs) [17]. The research group analyzed a large primary care database, including both laboratory-confirmed and clinical diagnoses of COVID-19 but largely excluding patients with prior CVD and accordingly selecting a comparatively healthy cohort.
Further, initially hospitalized, more severe cases could not be identified during follow-up in their database. Later emerging supporting publications from varying populations, limited to a maximum of 1 year of follow-up, point toward a considerably lower, but still elevated, risk of VTE in patients who were never hospitalized [18,19].
Given the uncertainty of long-term risk of VTE after COVID- 19, we identified all recorded infections from national Swedish registers, which were divided into patients initially hospitalized for COVID -19 and followed after discharge and individuals with COVID-19 without hospitalization, and compared them to a non-COVID-19exposed group randomly selected from the Swedish population and matched for age, sex, and week of COVID-19 diagnosis. The aim of this study was to investigate long-term risk of VTE, separately for PE and DVT after COVID-19, focusing on late VTE risk starting 60 days after patients were indexed in the study.

| Data sources
By linking multiple nationwide registries, we collected data on hospitalizations, hospital outpatient visits, and deaths from the Swedish National Patient and Cause of Death Registers, kept by the National Board of Health and Welfare (NBHW). The NBHW also registers data on assisted living (including home care and living in a long-term care facility, as previously described in detail) [20]. Information on education, income, and country of origin was collected from Statistics Sweden. Positive SARS-CoV-2 tests using polymerase chain reaction were retrieved from the surveillance system for communicable diseases in Sweden (SmiNet).
2.2 | Study design and definition of exposed and matched nonexposed populations The exposed groups were derived from all Swedish residents aged 18 to 84 years, alive on January 1, 2020, with no prior diagnosis of VTE, and either a laboratory-confirmed diagnosis of SARS-CoV-2 or a hospital discharge code of U071 or U072 according to the International Classification of Diseases. For each COVID-19 case, 5

Essentials
• The long-term risk of venous thromboembolism (VTE) after COVID-19 is incompletely understood.
• In COVID-19 cases without hospitalization, the VTE risk had returned to baseline after 60 days.
• After hospitalization for COVID-19, risk of pulmonary embolism was still elevated after 180 days.
nonexposed individuals with neither prior VTE diagnosis nor positive polymerase chain reaction test prior to matching were randomly selected from the Total Population Register. Inclusion, exclusion, and recruitment of the cohort are presented in Figure 1. Matching was performed for sex, year of birth, and week of diagnosis of COVID-19 (detailed procedure in Supplementary Figure S1). Weekly matching procedures were performed to correct for potential confounding by seasonal variation in transmission, increasing immunity from infection or vaccination, prevailing regional and national nonpharmaceutical interventions, and availability of testing (which was less systematic in the early months of the pandemic).
Inclusion in the study was between February 1, 2020, and September 11, 2021, with follow-up until November 11, 2021, ie, a minimum of 60 days of observation. Information on baseline diagnostic comorbidities and the outcomes that followed was collected from hospital and inpatient registers, which were registered until December 31, 2019. The detailed International Classification of Diseases codes used are presented in Supplementary Table S1. COVID-19-exposed patients were divided into those initially hospitalized for COVID-19 and those with COVID-19 without hospitalization. Hospitalizations were considered due to COVID-19 if COVID-19 was the principal diagnosis or COVID-19 was a contributory diagnosis with the principal diagnosis likely to be COVID-19related (listed in Supplementary Table S2). For multiple hospitalizations in 1 patient, only data from 1 hospitalization were accounted for in the following order: COVID-19 as a principal diagnosis, COVID-19 as a contributory diagnosis with an acceptable principal diagnosis.
For patients initially hospitalized for COVID-19, the index time was defined from the admission date of the first (initial) hospitalization for COVID-19, and for those with COVID-19 without hospitalization, it was defined from the date of laboratory-confirmed SARS-CoV-2 infection. Subjects were followed until the first event of outcome as defined below, death, end of follow-up, or, in the case of nonexposed individuals, a COVID-19 diagnosis.
The Charlson Comorbidity Index (CCI) for register-based research was calculated as described previously [21] with modifications according to data availability (Supplementary Table S3). Data on country of origin were dichotomized into patients born in a Nordic country (Sweden, Denmark, Finland, Iceland, or Norway) or any other country.
Information on ethnicity and race is not registered in Swedish population records and, therefore, not available. Information on care for the elderly or disabled was dichotomized into either independent or assisted living. Education was categorized as compulsory (≤9 years), 10 to 12 years, or college/university education.

| Statistical analysis
By definition, missing data on specific comorbidities were coded as absence of the relevant comorbidity. Missing data on living conditions were assumed to be independent living. Missing data on the variables born in Nordic countries and education were imputed using multivariate imputation by chained equations [22] before statistical analyses. Variables included in the multivariate imputation by chained equations algorithm were age, sex, education, born in Nordic countries, need for care, and baseline comorbidities. In baseline tables, data are reported without imputation (Table 1; Supplementary Table S4).
Incident rates were calculated as number of events divided by the total time at risk because matching data were not age standardized.
Outcomes were analyzed as VTE and as PE and DVT separately.
Furthermore, PE was analyzed among patients without any concurrent DVT in order to reduce the risk of diagnostic bias for PE F I G U R E 1 Inclusion, exclusion, and recruitment chart for cases and controls. VTE, venous thromboembolism.

| Ethics statement
The study conforms to the principles outlined in the Helsinki Declaration. All data were linked by the NBHW, after which personal identifiers were removed and replaced by a code. The project was approved by the Swedish Ethical Review Authority. Because pseudonymized data were used, consent was not applicable. compared to those without hospitalization, were markedly older, more often men, and born in non-Nordic countries ( Table 1).
Compared to nonexposed patients, they had more comorbidities, were more often in need of care, were born in non-Nordic countries, and were retired/nonworking, with lower education. For individuals with COVID-19 without hospitalization, the exposed and nonexposed groups were well balanced regarding comorbidities as well as occupational and educational levels but with a higher proportion of hospital and school staff among those exposed to COVID-19.

| Overview
We identified 2380 VTE events in the group of patients initially hospitalized for COVID-19 occurring during follow-up within the first 60 days, the majority of which (1828/2380, 76.8%) were diagnosed with PE during the initial hospital stay (

| General
After the early phase (≥60 days), the proportions between incident PE and DVT became more evenly distributed, but a larger excess hazard of PE relative to that to DVT remained, particularly for patients who were initially hospitalized for COVID-19 during the 60 to <180 days of the follow-up period compared with that in the nonexposed patients. The proportional rise in PE over VTE has not been a consistent observation [1,2,4,5,11,12], but in this large observational study, including long-term follow-up, we were able to confirm a strong preponderance of PE in relation to DVT. Mechanistic explanations may involve inflammation of the pulmonary vasculature, endothelial disruption, and activation of coagulation [25][26][27]. Detection bias due to severe respiratory symptoms has been proposed but might equally lead to underdiagnosis as pulmonary involvement may be attributed SJÖLAND ET AL.
T A B L E 2 Thromboembolic events in patients hospitalized for COVID-19 (exposed) and matched subjects without COVID-19 (nonexposed) groups. T A B L E 3 Thromboembolic events in subjects with COVID-19 without hospitalization (exposed) and without COVID-19 (nonexposed) groups. to the natural course of COVID-19 [28][29][30]. In our study, a similar pattern for PE and PE without DVT supports the absence of detection bias for PE. Importantly, the dramatic increase in VTE during hospitalization and protective effects of anticoagulation (AC) [1,31,32] led to prophylactic implementation of in-hospital AC treatment early during the pandemic [1].

| Patients initially hospitalized for COVID-19 vs COVID-19 without hospitalization
Consistent with prior studies, we observed that cardiometabolic comorbidities, older age, and male sex were associated with a more severe initial course of COVID-19, increased risk of hospitalization [33][34][35][36][37], and dramatic increase in the hazard of VTE [38]. Likely, the propensities for VTE and more severe COVID-19 are determined by host factors such as traditional CVD risk factors [15,39]. The rapid reduction in the HR of VTE over time during follow-up from 60 to 4.3 | Course at late (60 to <180 days) and long-term (≥180 days) follow-up A high incidence of intravascular clotting and intense inflammation has been repeatedly reported during acute COVID-19 and has raised concern for long-term pulmonary vascular function and threat of complications relative to VTE [15]. However, we found that the general risk of VTE from day 60 and onward in subjects with COVID-19 without hospitalization was comparable to that in the background population, albeit with some variation in risk between age groups, but without any consistent pattern. By contrast, follow-up after discharge of the patients initially hospitalized for COVID-19 held an extended and markedly heightened hazard for VTE over a lengthy follow-up after adjustment for several comorbidities. The rapidly reduced incidence in the group with COVID-19 without hospitalization may have resulted from the decrease of inflammatory activity after the acute phase of COVID-19 as inflammation with activation of hematologic and immunologic pathways has been linked to VTE [25][26][27]. A fast decrease in risk of VTE after COVID-19 was also observed in a large cohort of patients registered in primary care general practice covering England and Wales [18]. However, even after extensive adjustment,

| Consideration of bias
The risk of bias due to comorbidities, frailty, and vulnerability being   [4]. Finally, only prior in-and out-patient hospital diagnoses were captured, which will have underestimated comorbidities diagnosed in primary care (such as diabetes, hypertension, and obesity),

| Strengths and limitations
and there was no information on body weight or lifestyle.

| C O N C L U S I O N S
Our study confirms an extremely elevated risk of VTE, predominantly PE, in the early phase of severe COVID-19, with rapid tapering within 60 days. In contrast to nonpandemic conditions, PE was the predominant VTE. From 60 days and onward, the overall risk of VTE became similar to the background population rate for the group with COVID-19 without hospitalization but remained elevated during longterm out-patient follow-up for the cohort initially hospitalized for COVID-19, particularly for PE, and in the younger age groups. However, there was also substantial attenuation of long-term risk in patients hospitalized for COVID-19. These data indicate, at most, a transient increase in future risk of VTE in patients with COVID-19 without hospitalization, ie, most affected individuals.

ETHICS STATEMENT
The study conforms to the principles outlined in the Helsinki Declaration. All data were linked by the National Board of Health and Welfare, after which personal identifiers were removed and replaced by a code. The project was approved by the Swedish Ethical Review Authority. Because pseudonymized data were used, consent was not applicable. developed concepts, designed the study, performed statistical analyses, interpreted data, and wrote and revised the manuscript critically. A.R.
developed concepts, designed the study, interpreted data, and wrote and revised the manuscript critically. The manuscript has been read and approved for submission by all the authors.

RELATIONSHIP DISCLOSURE
There are no competing interests to disclose.