The low specificity of ventilation-perfusion lung scanning complicates the management of patients with suspected pulmonary embolism. To determine the safety of a clinical model for patients with suspected pulmonary embolism. Prospective cohort study. Five tertiary care hospitals. 1239 inpatients and outpatients with suspected pulmonary embolism. A clinical model categorized pretest probability of pulmonary embolism as low, moderate, or high, and ventilation-perfusion scanning and bilateral deep venous ultrasonography were done. Testing by serial ultrasonography, venography, or angiography depended on pretest probability and lung scans. Patients were considered positive for pulmonary embolism if they had an abnormal pulmonary angiogram, abnormal ultrasonogram or venogram, high-probability ventilation-perfusion scan plus moderate or high pretest probability, or venous thromboembolic event during the 3-month follow-up. All other patients were considered negative for pulmonary embolism. Rates of pulmonary embolism during follow-up in patients who had a normal lung scan and those with a non-high-probability scan and normal serial ultrasonogram were compared. Pretest probability was low in 734 patients (3.4% with pulmonary embolism), moderate in 403 (27.8% with pulmonary embolism), and high in 102 (78.4% with pulmonary embolism). Three of the 665 patients (0.5% [95% CI, 0.1% to 1.3%]) with low or moderate pretest probability and a non-high-probability scan who were considered negative for pulmonary embolism had pulmonary embolism or deep venous thrombosis during 90-day follow-up; this rate did not differ from that in patients with a normal scan (0.6% [CI, 0.1% to 1.8%]; P > 0.2). Management of patients with suspected pulmonary embolism on the basis of pretest probability and results of ventilation-perfusion scanning is safe.