Use of data mining surveillance system in real time detection and analysis for healthcare-associated infections
CM Ke, FJ Huang, Shoei‐Sheng Lee, YS Chen, PJ Hsieh, Lin Ye · BMC Proceedings · 2011
HAIs caused by multi-drug resistant organisms are our targets in a medical center in southern Taiwan. We designed an automated mechanism to import laboratory results combined with patient-specific data (DOA, bed #, lab orders, etc..). The moving average and trend of positive cultures were plotted. We also used data mining rules (Apriori, Anomaly, and Time-Series analysis) to determine the potential of undetected HAI and outbreaks. The moving average is a good tool of predicting carbapenem-resistant A. baumannii (CRAB) transmission in ICUs. Our surveillance may also determine the potential index patient in a time-series analysis. The Anomaly analysis was able to point out the potential patient wards to have an outbreak by detecting multi-drug resistant organisms (eg. CRAB and MRSA) or rare organisms (eg. VRE) from laboratory results. Our results showed that a real time rule-based automated infection surveillance system is possible to assist ICPs to detect potential HAIs which saves time and manpower to prevent nosocomial infections.