BACKGROUND: Ultrasonography is used routinely for ruling out suspected deep vein thrombosis in hospitalized patients, although most evidence supporting this strategy is derived from the outpatient setting. This study aimed to estimate the rate of venous thromboembolism when anticoagulant therapy was withheld from inpatients with normal findings on whole-leg ultrasonography. METHODS: As part of a prospective multicenter cohort study, 1926 medical and surgical inpatients with clinically suspected deep vein thrombosis during their stay were enrolled. Ultrasonography of all lower extremities was performed by board-certified vascular medicine physicians using a standardized examination protocol. Deep vein thrombosis was detected in 395 patients (20%). Anticoagulant therapy was withheld from patients with normal findings, and 523 of them were randomly selected for follow-up. The main outcome measure was 3-month incidence of symptomatic venous thromboembolism. RESULTS: A total of 513 patients with normal findings on ultrasonography successfully completed 3 months of follow-up, 9 patients were lost to follow-up, and 1 patient received anticoagulant therapy during follow-up. Three patients (0.6%) experienced nonfatal symptomatic venous thromboembolic events confirmed by objective testing. The cause of death was judged to be possibly related to pulmonary embolism for 7 other patients (1.3%). Overall, the 3-month rate of venous thromboembolism was 1.9% (10/513; 95% confidence interval, 0.9-3.5). CONCLUSION: Although withholding anticoagulant therapy after a single negative whole-leg ultrasonography seems to be safe, up to 3.5% of inpatients may nevertheless develop venous thromboembolism in the next 3 months. Further study is warranted to determine whether this strategy is equivalent to serial compression ultrasonography limited to proximal veins.