Objective: The aim of this study was to assess the incidence of deep vein thrombosis (DVT) of the lower limbs, using serial compression ultrasound (CUS) surveillance, in acutely ill patients with COVID-19 pneumonia admitted to a non-ICU setting. Methods: Multicenter, prospective study of patients with COVID-19 pneumonia admitted to Internal Medicine units. All patients were screened for DVT of the lower limbs with serial CUS. Anticoagulation was defined as: low dose (enoxaparin 20–40 mg/day or fondaparinux 1.5–2.5 mg/day); intermediate dose (enoxaparin 60–80 mg/day); high dose (enoxaparin 120–160 mg or fondaparinux 5–10 mg/day or oral anticoagulation). The primary end-point of the study was the diagnosis of DVT by CUS. Results: Over a two-month period, 227 consecutive patients with moderate-severe COVID-19 pneumonia were enrolled. The incidence of DVT was 13.7% (6.2% proximal, 7.5% distal), mostly asymptomatic. All patients received anticoagulation (enoxaparin 95.6%) at the following doses: low 57.3%, intermediate 22.9%, high 19.8%. Patients with and without DVT had similar characteristics, and no difference in anticoagulant regimen was observed. DVT patients were older (mean 77±9.6 vs 71±13.1 years; p = 0.042) and had higher peak D-dimer levels (5403 vs 1723 ng/mL; p = 0.004). At ROC analysis peak D-dimer level >2000 ng/mL (AUC 0.703; 95% CI 0.572–0.834; p = 0.004) was the most accurate cut-off value able to predict DVT (RR 3.74; 95%CI 1.27–10, p = 0.016). Conclusions: The incidence of DVT in acutely ill patients with COVID-19 pneumonia is relevant. A surveillance protocol by serial CUS of the lower limbs is useful to timely identify DVT that would go otherwise largely undetected.