ABSTRACT
Objective To develop a predictive model for pulmonary embolism(PE)based on the related clinical symptoms,signs,and the labo-ratory index,so as to improve the positive rate of CTPA. Methods The model was developed from a database of 119 patients with suspected PE. The risk factors of suspected PE were analyzed by logistic regression analysis ,which included significant differences in the prevalence of PE be-tween non-diseased and non-diseased groups. Receiver operating characteristic(ROC)curves was draw to determine the cut-off value of the clini-cal probability. It was validated in an independent sample of 106 patients with suspected PE. Results According to the univariate analysis ,17 of 51 variables show a significant difference between PE and non-PE patients. The model comprised 4 variables:age,dyspnea,D-dimer and unilater-al leg swelling. The area under the ROC curve is 0.776,and the cut-off value is supposed to be 0.38. In the validation sample,27% patients had PE confirmed by CTPA. The prevalence of PE was 54%when the clinical probability was above 0.38. Conclusion The proposed predictive mod-el in this study can improve the positive rate of CTPA ,simplify the diagnosis process of suspected PE patients.