Abstract: OBJECTIVE To explore a new method for the early warning of hospital-acquired infection outbreak by mining the clinical microbial data, so as to make it more focused on the work, and improve the work efficiency. METHODS Using the mobile percentile method, P95 of the cumulative detection frequency of a certain pathogenic bacteria in one week was served as an early warning threshold (if the P95 was less than 3, then 3 was defined as the threshold). If the cumulative frequency exceeded the threshold, a warning signal would be generated and the timing sequence diagram would be drawn by the SAS program automatically. After screening the timing diagram artificially and reviewing the relevant medical records, the warning could be determined whether it was a suspected hospital-acquired infection outbreak. RESULTS Based on the microbiological data of a hospital in 2013, 126 warning signals were generated by the method and 8 suspected outbreak signals were screened out. Retrospective investigation confirmed that 5 of them were suspected outbreaks. CONCLUSION This method is simple, efficient and with low cost, which can be an effective tool for the early warning of hospital-acquired infection outbreaks.