Background: The limitations of the current diagnostic standard, ventilation-perfusion lung scanning, complicate the management of patients with suspected pulmonary embolism. We previously demonstrated that determining the pretest probability can assist with management and that the high negative predictive value of certain D -dimer assays may simplify the diagnostic process.
Objective: To determine the safety of using a simple clinical model combined with D -dimer assay to manage patients presenting to the emergency department with suspected pulmonary embolism.
Design: Prospective cohort study.
Setting: Emergency departments at four tertiary care hospitals in Canada.
Patients: 930 consecutive patients with suspected pulmonary embolism.
Interventions: Physicians first used a clinical model to determine patients' pretest probability of pulmonary embolism and then performed a D -dimer test. Patients with low pretest probability and a negative D -dimer result had no further tests and were considered to have a diagnosis of pulmonary embolism excluded. All other patients underwent ventilation-perfusion lung scanning. If the scan was nondiagnostic, bilateral deep venous ultrasonography was done. Whether further testing (by serial ultrasonography or angiography) was done depended on the patients' pretest probability and the lung scanning results.
Measurements: Patients received a diagnosis of pulmonary embolism if they had a high-probability ventilation-perfusion scan, an abnormal result on ultrasonography or pulmonary angiography, or a venous thromboembolic event during follow-up. Patients for whom the diagnosis was considered excluded were followed up for 3 months for the development of thromboembolic events.
Results: The pretest probability of pulmonary embolism was low, moderate, and high in 527, 339, and 64 patients (1.3%, 16.2%, and 37.5% had pulmonary embolism), respectively. Of 849 patients in whom a diagnosis of pulmonary-embolism had initially been excluded, 5 (0.6% [95% CI, 0.2% to 1.4%]) developed pulmonary embolism or deep venous thrombosis during follow-up. However, 4 of these patients had not undergone the proper diagnostic testing protocol. In 7 of the patients who received a diagnosis of pulmonary embolism, the physician had performed more diagnostic tests than were called for by the algorithm. In 759 of the 849 patients in whom pulmonary embolism was not found on initial evaluation, the diagnostic protocol was followed correctly. Only 1 (0.1% [CI, 0.0% to 0.7%]) of these 759 patients developed thromboembolic events during follow-up. Of the 437 patients with a negative D -dimer result and low clinical probability, only 1 developed pulmonary embolism during follow-up; thus, the negative predictive value for the combined strategy of using the clinical model with D -dimer testing in these patients was 99.5% (CI, 99.1% to 100%).
Conclusion: Managing patients for suspected pulmonary embolism on the basis of pretest probability and D -dimer result is safe and decreases the need for diagnostic imaging.