Timely detection of severe patient conditions or concerning events and their mitigation remains an important problem in clinical practice. This is especially true in the critically ill patient. Typical computer-based detection methods developed for this purpose rely on the use of clinical knowledge, such as expert-derived rules, that are incorporated into monitoring and alerting systems. However, it is often time-consuming, costly, and difficult to extract and implement such knowledge in existing monitoring systems. The research work in this proposal offers computational, rather than expert-based, solutions that build alert systems from data stored in patient data repositories, such as electronic medical records. Briefly, our approach uses advanced machine learning algorithms to identify unusual clinical management patterns in individual patients, relative to patterns associated with comparable patients, and raises an alert signaling this discrepancy. Our previous studies provide support that such deviations indicate clinically important events at false alert rates below 50%, which is very promising. We propose to further improve the new methodology, and build a real-time monitoring and alerting system integrated with production electronic medical records. We propose an evaluation of the system using physicians'assessment of alerts raised by our real-time system for intensive-care unit (ICU) patient cases. The project investigators comprise a multidisciplinary team with expertise in critical care medicine, computer science, biomedical informatics, statistical machine learning, knowledge based systems, and clinical data repositories.

Public Health Relevance

There remain numerous opportunities to reduce medical errors in the intensive care unit (ICU). This project develops and evaluates a new clinical monitoring and alerting framework that uses electronic medical records and machine-learning methods to send alerts concerning clinical decisions in the ICU that are unexpected given the clinical context and may represent medical errors.

Agency
National Institute of Health (NIH)
Institute
National Institute of General Medical Sciences (NIGMS)
Type
Research Project (R01)
Project #
2R01GM088224-04A1
Application #
8641014
Study Section
Biomedical Computing and Health Informatics Study Section (BCHI)
Program Officer
Marcus, Stephen
Project Start
2009-09-01
Project End
2018-05-31
Budget Start
2014-08-15
Budget End
2015-05-31
Support Year
4
Fiscal Year
2014
Total Cost
$580,180
Indirect Cost
$203,440
Name
University of Pittsburgh
Department
Biostatistics & Other Math Sci
Type
Schools of Arts and Sciences
DUNS #
004514360
City
Pittsburgh
State
PA
Country
United States
Zip Code
15213
Nguyen, Quang; Valizadegan, Hamed; Hauskrecht, Milos (2014) Learning classification models with soft-label information. J Am Med Inform Assoc 21:501-8
Hauskrecht, Milos; Batal, Iyad; Valko, Michal et al. (2013) Outlier detection for patient monitoring and alerting. J Biomed Inform 46:47-55
Valizadegan, Hamed; Nguyen, Quang; Hauskrecht, Milos (2013) Learning classification models from multiple experts. J Biomed Inform 46:1125-35
Batal, Iyad; Valizadegan, Hamed; Cooper, Gregory F et al. (2011) A Pattern Mining Approach for Classifying Multivariate Temporal Data. Proceedings (IEEE Int Conf Bioinformatics Biomed) 2011:358-365
Nguyen, Quang; Valizadegan, Hamed; Seybert, Amy et al. (2011) Sample-efficient learning with auxiliary class-label information. AMIA Annu Symp Proc 2011:1004-12
Amizadeh, Saeed; Wang, Shuguang; Hauskrecht, Milos (2011) An Efficient Framework for Constructing Generalized Locally-Induced Text Metrics. IJCAI (U S) :1159-1164