Performance of pulmonary embolism diagnostic strategies in patients undergoing computed tomography pulmonary angiography

Clinical probability scores and D-dimer-based diagnostic strategies are intended for early evaluation of suspected pulmonary embolism (PE), but their performance after clinician-driven referral for computed tomography pulmonary angiography (CTPA) remains uncertain. We prospectively evaluated PE diagnostic strategies among consecutive emergency department patients aged ≥13 years who underwent CTPA for suspected PE. Following the decision to order CTPA, the modified Wells and simplified revised Geneva scores were calculated, and the YEARS and PEGeD strategies were applied. Clinically significant PE, defined as segmental or more proximal embolism, was the primary outcome; any PE, including isolated subsegmental embolism, was the secondary outcome. Among 408 patients, clinically significant PE was confirmed in 144 (35.3%) and any PE in 154 (37.7%). The Wells score demonstrated better discrimination for clinically significant PE than the simplified revised Geneva score (area under the curve 0.725 [95%CI: 0.671-0.779] vs 0.656 [95%CI: 0.598-0.714]; P = 0.0003). Clinically significant PE was present in 21.6% and 24.4% of patients classified as PE unlikely by the Wells and simplified revised Geneva two-level classifications, respectively. Among 398 patients with D-dimer results, YEARS and PEGeD each demonstrated 96.5% sensitivity but identified only 13.8% and 13.6% of patients, respectively, as potentially eligible for imaging avoidance. Diagnostic performance showed a similar pattern for any PE. In this CTPA-selected cohort, the Wells score demonstrated better discrimination than the simplified revised Geneva score, but both two-level classifications yielded substantial counterfactual missed-PE rates. YEARS and PEGeD maintained high sensitivity but offered limited theoretical imaging-sparing potential.