Abstract: Objective To evaluate the predictive value of different Wells score combined with D-dimer level in the diagnosis of acute pulmonary embolism (PE). Methods A total of 200 patients with suspected PE admitted in our hospital from 2013 to 2016 were enrolled into this study and divided into PE group and non-PE group by using computed tomography pulmonary angiography (CTPA). The risk factors for pulmonary embolism was investigate through statistical analysis. Receiver operating characteristic (ROC) curve analysis was used to compare the different predictive models. Results Of 183 patients included in this study, 101 patients were assigned into PE group and 82 in Non-PE group. The difference in the incident of chronic cardiac insufficiency, vascular diseases, the history of PE or DVT (deep vein thrombosis), difficult breathing, unilateral lower limb pain, lower extremity deep venous tenderness and single limb swelling, elevated levels of D-dimer, high Wells score was statistically significant between PE group and Non-PE group (all P<0.05), and not significant for other variables (all P>0.05). The areas under the ROC curves (AUC) of different assessment Methods were 0.834 (95% CI: 0.758-0.911), 0.775 (95% CI: 0.688-0.862), 0.732 (95% CI: 0.643-0.820), 0.857 (95% CI: 0.783-0.931), and 0.887 (95% CI: 0.820-0.954), respectively. The AUC for the combination of original Wells score version and D-dimer level was greater than that for the combination of simplify Wells score version and D-dimer level, and that for the method alone (all P<0.05). When the cut-off value of the combination of original Wells score version and D-dimer level was 0.96, and the weighted maximum sensitivity was 87.76% and specificity was 85.37%, and it reached the best reliability of diagnosing and eliminating PE. Conclusion Original Wells score version combined with D-dimer level shows a higher value in predicting acute pulmonary embolism than each method alone.