Excluding pulmonary embolism at the bedside with low pre-test probability and D-dimer: Safety and clinical utility of 4 methods to assign pre-test probability

Less than 35% of patients suspected of having pulmonary embolism (PE) actually have PE. Safe bedside methods to exclude PE could save scarce health care resources if they exclude large proportions of patients with suspected PE and are widely applicable. Non-Elisa D-dimer in combination with pre-test probability of suspected PE can safely exclude PE at the bedside. Pre-test probability can be assigned by gestalt or by using clinical models (Wells, Wicki, Rodger). We combined two databases from studies of patients with suspected PE and retrospectively compared the diagnostic test characteristics of the different methods of assigning pre-test probability. 535 patients were studied. PE was confirmed in 20.8% of study patients. Two clinical predictive models (Rodger and Wells) and overall diagnostic impression have similar sensitivities ranging from 96% (95% confidence interval (CI) 89-99%) to 99% (93-100%). Wicki's model has a sensitivity of 89% (77-96%). The Wells' model with a cutoff of less than 2 points in association with semi-quantitative D-dimer has a specificity of 11% (CI 7-15%). The specificities for the other clinical predictive model are ranging from 21% (17-25%) to 49% (CI 42-55%). Semi-quantitative D-dimer must be combined with safe clinical probability assessment to safely exclude PE in a significant proportion of patients. Wicki's model in association with semi-quantitative D-dimer has the lowest sensitivity and should be used carefully to exclude PE at the bedside. The Wells' model with a cutoff of less than 2 points when combined with semi-quantitative D-dimer excluded very few patients and therefore limits its clinical utility.