This study evaluated the diagnostic value of D-dimer, CRP, and leucocytes count to detect an underlying pulmonary embolism (PE) in patients with pneumonia. A predictive model of an underlying PE, based on laboratory markers and clinical symptoms, was our ultimate objective. Overall 100 patients underwent a computed tomography angiography (CTA) of the lung: 54 with coexistence of PE and pneumonia (cases) and 46 with pneumonia without PE (controls). Cases and controls were matched 1 : 1. Symptoms and paraclinical findings were registered on admission. Receiver operating characteristic (ROC) curves, search for an optimal threshold, and conditional logistic regression analysis were conducted. D-dimer has a moderate ability to detect PE in pneumonia. Sensitivity of D-dimer was estimated at 97.78% and specificity at 11.11%. No optimal cut-point has acceptable diagnostic ability. After excluding patients with sepsis, sensitivity was reduced to 96.97%, whereas specificity increased to 16.13%. Consolidation in chest X-ray and positive D-dimer predict better an underlying PE as D-dimer itself. Thus, discriminatory power of the prediction model (AUC of 0.740) is not much greater than D-dimer (AUC of 0.703). No threshold that could increase the diagnostic value of D-dimer or a prediction model which is significantly better than D-dimer itself was identified.
