BACKGROUND

Venous thromboembolism (VTE) is a frequent complication in critically ill patients with coronavirus disease 2019 (COVID-19) and is associated with mortality. Early diagnosis and treatment of VTE is warranted.

OBJECTIVE

To develop a prediction model for VTE in critically ill COVID-19 patients.

PATIENTS AND METHODS

In this retrospective cohort study, 127 adult patients with confirmed COVID-19 infection admitted to the intensive care unit of two teaching hospitals were included. VTE was diagnosed with either ultrasound or computed tomography scan. Univariate receiver operating characteristic (ROC) curves were constructed for Positive End Expiratory Pressure, PaO/FiO ratio, platelet count, international normalized ratio, activated partial thromboplastin time as well as levels of fibrinogen, antithrombin, D-dimer and C-reactive protein (CRP). Multivariate analysis was done using binary linear regression.

RESULTS

Variables associated with VTE in both univariate and multivariate analysis were D-dimer and CRP with an area under the curve (AUC) of 0.64, P = 0.023 and 0.75, P = 0.045, respectively. Variables indicating hypoxemia were not predictive. The ROC curve of D-dimer and CRP combined had an AUC of 0.83, P < 0.05. Categorized values of D-dimer and CRP were used to compute a mean absolute risk for the combination of these variables with a high positive predictive value. The predicted probability of VTE with a D-dimer > 15 in combination with a CRP > 280 was 98%. The negative predictive value of D-dimer was low.

CONCLUSION

Elevated CRP and D-dimer have a high positive predictive value for VTE in critically ill COVID-19 patients. We developed a prediction table with these biomarkers that can aid clinicians in the timing of imaging in patients with suspected VTE.