Pre-CT risk stratification using the D-dimer/pCO₂ ratio in D-dimer–positive emergency department patients: diagnostic accuracy study

Background The diagnostic utility of the D-dimer/pCO₂ ratio for pulmonary embolism (PE) risk stratification has not been fully established. This study evaluated its diagnostic performance among emergency department patients with positive age-adjusted D-dimer results undergoing computed tomography pulmonary angiography (CTPA). Methods This retrospective diagnostic accuracy study included 698 adult patients with positive age-adjusted D-dimer results, venous blood gas (VBG) pCO₂ measurements, and definitive CTPA interpretation. The D-dimer/pCO₂ ratio was calculated, and receiver operating characteristic (ROC) analysis was performed. Optimal and exploratory thresholds were assessed for overall PE detection and for excluding central PE. Robustness was tested using bootstrap validation and subgroup AUC comparisons. Decision curve analysis (DCA) was applied to evaluate clinical utility. Results PE was confirmed in 90 patients (12.9%). The ratio demonstrated good discrimination (AUC: 0.811, 95% CI: 0.775–0.847). At the optimal cut-off (44.91), sensitivity was 82.2% and specificity 71.1%, with a negative predictive value (NPV) of 96.4%. A lower cut-off (18.1) identified 91 patients with no observed PE (0/91; 95% CI upper bound for false negatives ≈ 4.0%). A higher threshold (61.25) identified 515 patients below this value, among whom no central PE was observed (0/515; 95% CI upper bound ≈ 0.7%). Discriminative ability was preserved across age groups (AUC range: 0.737–0.836). DCA showed modest, range-specific net benefit for incorporating the ratio within a low-to-intermediate threshold band. Conclusion In D-dimer–positive ED patients already being considered for CTPA, the D-dimer/pCO₂ ratio is an adjunctive imaging triage indicator rather than a stand-alone test and may help inform the imaging workflow in this defined context. These findings should not be extrapolated to D-dimer–negative patients or those with very high pretest probability.