Abstract: Objective To compare the detection situation of influenza-like illnesses with different symptoms and analyze the impact on the influenza surveillance by adopting different case definitions of influenza-like illness. Methods Data was collected from 2 national influenza surveillance sentinel hospitals in Shanghai, 2015. We compared the positive rate of influenza virus among patients with different symptoms (with cough and sore throat, with cough only and with sore throat only), and utilized Logistic regression model to analyze the influencing factors of the detection rate of influenza virus. Results Among 2 010 influenza-like illnesses, 1 105 patients were with cough and sore throat, 270 patients were with cough, 635 patients were with sore throat, and the positive rate of influenza was 36.2%, 39.3% and 15.9% respectively. The patients with cough and sore throat or with cough only had a higher positive rate of influenza than patients with sore throat (all P<0.05). For patients with specimens collected within 3 days, patients with cough and sore throat, or with cough only had a higher positive rate than patients with sore throat only (all P<0.05). Logistic regression results showed that fever (body temperature ≥ 39℃) (OR=1.719, 95% CI:1.389-2.127) and cough (OR=3.046, 95% CI:2.377-3.905) were associated with the detection of influenza virus. Conclusions We suggested that we can adopt the case definition of influenza-like illness 'fever (body temperature ≥ 38℃) and cough' in the influenza surveillance system.