Title:
Outlier detection for clinical monitoring and alerting
Speaker:
Prof. Milos Hauskrecht
Professor of Computer Science
University of Pittsburgh
Abstract:
Outlier detection has been successfully applied to identification of unusual data instances, unusual behaviors and outcomes. In this work, I present an outlier detection framework for clinical monitoring and alerting that aims to identify unusual patient management actions in electronic health record data. Our hypothesis is that patient-management actions that are unusual with respect to past patients may be due to errors and that it is worthwhile to raise an alert if such a condition is encountered prospectively. Our methodology was tested on data derived from electronic health records, and the quality of alerts was evaluated using a panel of clinicians. Our results support that outlier-based alerting can lead to reasonably low false alert rates and that stronger outliers are correlated with higher alert rates
Bio:
Dr. Milos Hauskrecht is a professor of computer science at the University of Pittsburgh. He received his PhD from MIT in 1997. His research interest are in machine learning, data mining and their applications in medicine, in particular, clinical monitoring and alerting, and real-time clinical data analysis. He has authored or co-authored over 100 publications in these areas. His research work is funded by grants from NIH. He serves regularly on program committees of the top artificial intelligence and biomedical informatics conferences. He is the recipient of Homer R. Warner award for 2010 for his work on outlier-based clinical monitoring and alerting.