P261 DIAGNOSTIC STRATEGIES FOR PULMONARY EMBOLISM IN PATIENTS HOSPITALIZED FOR COVID–19: ROLE OF CLINICAL PREDICTION RULES

Abstract

BackgroundDiagnostic strategies for pulmonary embolism (PE) in patients already facing respiratory failure due to COVID–19 is challenging. The use of the conventional diagnostic algorithm and clinical prediction rules (CPR) for PE is controversial in these patents.MethodsThe aim to evaluate the accuracy of currently available CPRs to assess the risk for venous thromboembolism (VTE) in medically ill patients with COVID–19 and algorithms for the diagnosis of PE. Consecutive patients >18 years hospitalized at Santa Maria della Misericordia Hospital (Perugia, Italy) from March 2th, 2020, to September 29th, 2021 were included if they had: 1) confirmed diagnosis of COVID–19 with a molecular testing; 2) chest CT angiography performed for clinical suspicion of PE during the hospital stay. The study outcome was the accuracy of currently available CPRs for PE diagnosis (Wells and Geneva) and for VTE–risk stratification in medically ill patients (IMPROVE, IMPROVEDD and Padua score) to predict the diagnosis of PE as confirmed by a contrast–enhanced CT lung scan.ResultsDuring the study period, 74 COVID–19 patients who had CT angiography for PE clinical suspicion were included (mean age 68 years, male 64.9%). Thirteen patients (17.6%) had PE confirmed at CT. No significant differences were observed for comorbidities, antithrombotic treatment and mortality between the two groups. D–dimer resulted significantly higher in patients with compared to patients without PE. Poor discrimination was observed for Wells and Geneva scores (AUC 0.596, 95% CI 0.413–0.779, and AUC 0.603, 95% CI 0.439–0.767, respectively), without substantial differences adding d–dimer at conventional cut–off (Table). The IMPROVEDD score had the highest discriminative power among CPRs for VTE (AUC 0.699, 95% CI 0.539–0.860). Scores’ performance improved by increasing the D–dimer cut–off at level of 2000 ng/ml: among diagnostic scores, Wells showed the best discrimination (AUC 0.806, 95% CI 0.674–0.939, negative predictive value 97%); among CPRs for VTE the IMPROVEDD confirmed its accuracy (AUC 0.769, 95% CI 0.633–0.904, negative predictive value 94%).ConclusionsThe accuracy of the currently used diagnostic and predictive scores for PE or VTE in COVID–19 patients is poor. D–dimer improves the diagnostic accuracy of these scores; most of all, it seems to allow a diagnostic strategy with a high negative predictive value, so we can rule out a consistent part of the patients with a low risk of PE.