Shifting the context from the emergency department to the department of medicine, we compared different scores to diagnose deep vein thrombosis (DVT) in patients with several comorbidities, hospitalized in a department of internal medicine. We prospectively recruited 178 consecutive hospitalized patients in whom clinical suspicion of DVT was assessed by Wells modified score for DVT, Hamilton, Kahn, and St Andrè Hospital scores. Deep vein thrombosis was confirmed in 85 (48%) patients by both echocolor Doppler and angiocomputed tomography scan. The use of risk scores based on symptoms and clinical signs was weakly useful (area under the curve [AUC]: 0.69, positive predictive value: 59%, and negative predictive value: 74%). Patients with DVT had significantly (P < .0001) lower serum albumin and protein S levels compared to those without DVT. Moreover, serum protein S (AUC: 0.82) and albumin in percentage (AUC: 0.80) showed a better accuracy than clinical scores (P < .001) in assessing the diagnosis of DVT. Therefore, serum albumin and protein S improved the accuracy of clinical scores for the diagnosis of incident DVT in patients hospitalized in a department of medicine.