Objective To evaluate the accuracy and efficiency of early warning strategy for cases by real-time healthcare-associated infection(HAI) surveillance system(HAISS), and propose scientific suggestion for the improvement of early warning strategy. Methods By investigating the early warning information and final confirmation of HAI generated by HAISS in a tertiary first-class general hospital in 2017, the sensitivity and positive predictive value of early warning of HAI as well as positive predictive value of early warning of infection were calculated to evaluate the accuracy and efficiency of early warning. Results 832 cases of HAI were confirmed in this hospital in 2017, 715 cases were HAI effectively warned by HAISS, with a sensitivity of 85.94%. A total of 8 468 cases were warned by HAISS in the whole year, 2 817 were infection cases, positive predictive value of early warning of infection cases was 33.27%, 772 cases were HAI, and positive predictive value of early warning of HAI cases was 9.12%. There were 14 857 early warnings in HAISS in the whole year, of which 4 135 were confirmed as infection, positive predictive value of early warning of infection was 27.83%, 1 199 were confirmed as HAI, positive predictive value of early warning of HAI was 8.07%. Conclusion HAISS is an important technical mean for identifying HAI cases, specificity and efficiency of case identification need to be improved, the ability of early warning for high-risk infection cases need to be strengthened.