PB0072 - Predictors of venous thromboembolism in hospitalised COVID-19 positive patients

Background: Inpatients with COVID-19 have a high rate of venous thromboembolism (VTE), yet those that are most unwell have been shown to exhibit excess bleeding following thromboprophylaxis. Risk profiling of those at highest thrombotic risk may therefore improve outcomes.

Aims: To derive and validate a risk assessment model for VTE in COVID-19 inpatients.

Methods: Electronic health records were used to assess all patients admitted for ≥1 night with laboratory-confirmed COVID-19 between March 2020 and July 2021 to Barts Health NHS Trust in East London. The primary event of interest was VTE within 28 days from diagnosis.

The study population was split into derivation (n=4655) and validation sets (n=1844). Potential predictors of VTE included demographic and lifestyle variables, clinical characteristics, and biomarkers. A logistic regression model was developed with predictors identified using least absolute shrinkage and selection operator (LASSO) methodology.

Results: The study population comprised 6499 patients (45% women, median age 60). 394 patients (6.1%) were diagnosed with ≥1 VTE event (30 DVT, 364 PE +/- DVT) within 28 days of diagnosis. D-dimer on admission was the strongest predictor for VTE. The risk of VTE was associated with increasing D-dimer up to 10 mg/L. Further rises in D-dimer above this level did not confer additional risk. Chronic cardiac disease, chronic obstructive pulmonary disease, and oxygen flow rate were also independently associated with increased risk. High peripherally measured oxygen saturations, ischaemic heart disease and supraventricular arrhythmias were associated with a reduced risk of VTE (Figure 1). The risk assessment model offered a strong discriminatory value (c-index 0.77) and achieved good calibration in both the derivation and validation set (Table 1).

Conclusion(s): The proposed model was robust in predicting VTE risk in successive waves of COVID-19 infection (original, alpha and delta variants) and supports the use of the D-dimer level for guiding thromboprophylaxis.