Venous or arterial thrombosis in COVID-19 cases in the North Carolina COVID-19 Community Research Partnership (NC-CCRP)

Background Although the incidence of venous and arterial thrombosis after a COVID-19 diagnosis and hospitalization has been well described using data available from electronic health records (EHR), little is known about their incidence after mild infections. Objectives To characterize the cumulative incidence and risk factors for thrombosis after a COVID-19 diagnosis among those identified through the EHR and those with a self-reported case. Methods We calculated the cumulative incidence of thromboembolism diagnoses after EHR-identified and self-reported cases in the North Carolina COVID-19 Community Partnership, a prospective, multisite, longitudinal surveillance cohort using a Kaplan-Meier approach. We performed Cox regression to estimate the hazard of a thromboembolism diagnosis after COVID-19 by comorbidities, vaccination status, and dominant SARS-CoV-2 variant. Results Of a cohort of comprising more than 39,500 participants from 6 North Carolina sites, there were 6271 self-reported or EHR-diagnosed cases of COVID-19 reported between July 1, 2020, and April 30, 2022, of which 46 participants were diagnosed with a new-onset thromboembolism in the 365 days after their reported case. Self-reported cases had a lower estimated cumulative incidence of 0.15% (95% CI, 0.03-0.28) by day 90 and 0.64% (95% CI, 0.30-0.97) by day 365 compared with EHR-based diagnoses that had cumulative incidences of 0.73% (95% CI, 0.36-1.09) and 1.78 (95% CI, 1.14-2.46) by days 90 and 365 (log-rank test P value <.001). Those hospitalized and with pre-existing pulmonary and cardiovascular diseases were associated with the highest risk of a thromboembolism. Conclusion We observed a higher cumulative incidence of thromboembolism after EHR-identified COVID-19 than self-reported cases.


| I N T R O D U C T I O N
Severe COVID-19 has been associated with higher rates of thromboembolic diseases, especially among those individuals who require hospitalization [1][2][3]. However, most SARS-CoV-2 infections, the causative agent of COVID-19, do not progress to hospitalization, especially among the vaccinated [4][5][6]. Recent work by Burn et al. [7] found that the estimated risk of venous and arterial embolism was nearly 1% for all COVID-19 cases, not limited to hospitalizations, in a retrospective study of more than 900,000 electronic health records (EHR) from 5 European countries after September 1, 2020. Similarly, a large study by Knight and colleagues using EHR records of 48 million adults in England and Wales found among those diagnosed with COVID-19 were 34% and 80% more likely to be diagnosed with arterial and venous thromboses, respectively, by weeks 27 to 49 when compared with those without a COVID-19 diagnosis. Because these findings relied on EHR records, patients with pauci-symptomatic and mild infections who did not seek treatment were likely underrepresented.
In this study, we describe the cumulative incidence of thromboembolic diseases after COVID-19 infections among those who selfreported a COVID-19 infection vs those who were identified through the EHR using the North Carolina COVID-19 Community Research Partnership (NC-CCRP), a well-defined, prospective cohort of more than 39,500 adult participants [8]. The NC-CCRP provides insight into the incidence and associated sequelae of COVID-19 cases that may not be captured in the EHR. We explored the association of comorbidities, dominant variant, and vaccination status with the risk of a thrombotic disease after infection. An additional secondary analysis included a self-controlled case series design among those in the cohort with thromboembolism diagnosis since January 1, 2020, to estimate the incidence ratio before and after COVID-19 diagnoses with each participant serving as their own control.

| Study design
The NC-CCRP is prospective cohort of more than 39,500 adult participants who completed daily health and symptom logs from April 1, 2020, through April 30, 2022 [9]. Adults [10]. We identified diagnoses of arterial and venous thromboembolisms using ICD-10 codes available within the EHR, consistent with previous studies [7,11,12]. A full list of the codes used to identify venous and arterial thromboembolic events are included in the appendix (Supplementary Table 1

Essentials
• High rates of newly diagnosed thromboembolic diseases (TD) have been reported after COVID-19.
• We analyzed the rate of TD in self-reported and electronic health record (EHR)-identified cases.
• Self-reported COVID-19 cases have a lower incidence of TD than EHR-identified cases.
• Risk is highest in hospitalized and those with history of pulmonary and cardiovascular diseases. diagnosis in the previous year, after January 1, 2020. The exposure period was defined as the date of COVID-19 report up to 365 days.
Participants who could not be associated with their EHR were excluded from the analysis.

| Statistical methods
We used a Kaplan-Meier approach to estimate the cumulative incidence of new thromboembolism diagnoses overall and by diagnosis source. A log-rank test was used to detect whether a statistically significant difference was observed between the sources of COVID-19 diagnosis. We performed adjusted and unadjusted Cox regressions to estimate the association of comorbidities, vaccination status at time of the infection, and variant period on the risk of new thromboembolism.
Age, sex, and hospitalization status were included as covariates in the adjusted model as potential sources of confounding. Schoenfeld residuals were plotted to visually confirm that proportional hazards assumption was not violated. Incidence rate ratios (IRR) were calculated using conditional Poisson regression models to compare the intraperson incidence rates of thromboembolism after COVID-19 in days 1 to 365 after diagnosis with the baseline (pre-COVID-19) period. We considered P values <.05 statistically significant.

| R E S U L T S
Among the 6271 self-reported or EHR-diagnosed cases of COVID-19 reported between July 1, 2020, and April 30, 2022, 46 study participants were diagnosed with a new-onset thromboembolism in the 365 days after their reported case date ( Table 1). Those that were diagnosed with a thromboembolism tended to be older than those who were not with a median age of 58 years (IQR, 48-67) vs 48 years (IQR, 38-59). The estimated overall cumulative incidence was 0.35% (95% CI, 0.20-0.50) by day 90 and 1.09% (95% CI, 0.75-1.41) by day 365 after reported COVID-19 case ( Supplementary Fig. 1).
When considering the source of COVID-19 diagnosis, self-reported cases had a lower estimated cumulative incidence of 0.15% (95% CI, 0.03-0.28) by day 90 and 0.64% (95% CI, 0.30-0.97) by day 365 compared with EHR-based diagnoses which had cumulative incidences of 0.73% (95% CI, 0.36-1.09) and 1.78 (95% CI, 1.14-2.46) by days 90 and 365, respectively ( Figure 1, panel A). These cumulative incidences translate to 1996 events per 100 000 person years for those cases identified through the EHR and 676 events per 100,000 person years among the self-reported cases for an overall rate of 1188 events per 100,000 person years (Supplementary Table 2). The log-rank test indicates that there was a statistically significant difference detected between those who self-reported COVID-19 vs those who were identified through the EHR (P <.001, Figure 1

| D I S C U S S I O N A N D C O N C L U S I O N
In this study, we find that the cumulative incidence of venous and arterial thromboembolisms is lower among those who self-reported a COVID-19 diagnosis than those who had a COVID-19 diagnosis identified through the EHR. These findings supplement those of Burn and colleagues as well as those by Knight and colleagues by including those COVID-19 cases that may have been underascertained when using EHR data. These findings continue to underline the association of thromboembolic disease after COVID-19 and support a growing body of literature that suggest higher incidence of thrombotic disease postinfection [1,7,12,[14][15][16][17].
Those who were diagnosed with a thromboembolism tended to be older than those who were not in agreement with studies suggesting the association of increased age with baseline risk of thromboembolism [18]. Underlying comorbidities, especially those with underlying cardiovascular and pulmonary diseases and diabetes, have been associated with progression to hospitalization in COVID-19 cases DEWITT ET AL.
-3 of 7 [19][20][21][22] The associations of these comorbidities with a subsequent thromboembolism diagnosis remained even after adjusting for age and sex. Hospitalization for COVID-19 was associated with the highest risk of a subsequent thromboembolism. Those with EHR-identified cases may have been more likely to have more severe disease and thus were at higher risk for adverse outcomes. Furthermore, analysis that relies on EHR-identified cases may generate biased estimates of the rate of adverse outcomes because of this association. Our  Hazard ratio E adjusted unadjusted F I G U R E 1 The estimate cumulative incidence of thromboembolism diagnosis by source of reported COVID-19 (A). Estimated hazard ratios for thromboembolism diagnosis by comorbidity condition (B). Vaccination status, reference is unvaccinated (C). Predominant SARS-CoV-2 variant at the time of infection, reference is pre-Delta (D). Adjusted hazards account for the contribution of age, sex, and hospitalization. Estimated hazard ratios for thromboembolism diagnosis by hospitalization status (E), reference is nonhospitalized; adjusted hazards account for the contribution of age and sex DEWITT ET AL. sequela. Self-reported infections were associated with a lower cumulative incidence of thrombotic disease than EHR-identified diagnoses for COVID-19, and a self-controlled case series approach failed to detect a statistically significant difference in the incidence ratio of thrombotic diseases among those self-reporting COVID-19 compared with the pre-COVID-19 control period. drafted the manuscript. All authors contributed to the final draft.

RELATIONSHIP DISCLOSURE
All authors have no relevant conflicts of interest to disclose.