Abstract: OBJECTIVE To establish a multi-factor early warning diagnosis model for sepsis by comparing the clinical data between patients with sepsis and common infection. METHODS A total of 215 patients admitted to the PLA General Hospital from Jun. 2010 to Jun. 2015 were divided into the sepsis group（86 cases） and non-sepsis infection group（also as the control group, 129 cases） according to their disease outcomes. The clinical data of patients were collected and analyzed to find the risk factors of sepsis, and the early warning and diagnosis model of sepsis was constructed. The receiver operating characteristic（ROC） of the early warning diagnosis model was drawn, and the area under the ROC curve（AUC）, sensitivity and specificity were calculated in order to evaluate the diagnostic efficiency of the early warning diagnosis model. RESULTS The risk factors of sepsis were skin petechiae（OR=0.023,P=0.013）, platelet（OR=0.900,P＜0.001）, procalcitonin（OR=0.717,P=0.024）, creatinine（OR=1.076,P=0.001） and international standard ratio（OR=9.842,P=0.009）. The area under the ROC curve（AUC） was 0.972, and the sensitivity and specificity were 98.8% and 74.8%, respectively. Preliminary verification test of 30 infected patients showed that the sensitivity of the model was 72.7%, the specificity was 89.5%, the positive predictive value was 80.0%, the negative predictive value was 85.0%, and the diagnostic coincidence rate was 83.3%. CONCLUSION Skin petechiae, platelet count, procalcitonin, creatinine and international standardized ratio are the risk factors of sepsis in patients with infection. The model based on these risk factors has a good performance on early warning and diagnosis of sepsis, which can be used for the early clinical identification of patients with infection and sepsis.