Lecture Series

Friday, October 9, 2009
11:00 AM to 12:00 PM
Parkvale Building (200 Meyran Avenue)
Classroom M-184 (on the mezzanine level), or via video conference at the UPMC Cancer Pavilion, Room 341

A Temporal Abstraction Framework for Classifying Clinical Temporal Data
Iyad Batal, MS
Doctoral Fellow, Computer Science

Abstract: The increasing availability of complex temporal clinical records collected today has prompted the development of new methods that extend classical machine learning and data mining approaches to time series data.

In this work, we develop a new framework for classifying the patient’s time-series data based on temporal abstractions. The proposed STF-Mine algorithm automatically mines discriminative temporal abstraction patterns from the data and uses them to learn a classification model. We apply our approach to predict HPF4 test orders from electronic patient health records. This test is often prescribed when the patient is at the risk of Heparin induced thrombocytopenia (HIT). Our results demonstrate the benefit of our approach in learning accurate time series classifiers, a key step in the development of intelligent clinical monitoring systems.

Effective communication of Drug-drug interaction knowledge
Richard Boyce, PhD
Post Doctoral Associate

Abstract: Drug-drug interactions (DDIs) are a preventable medication error yet millions of Americans are exposed to clinically important interactions each year. The incorporation of evidence for drug interactions into health technology systems is occurring with marginal improvements of safety. To address these issues, the Agency for Healthcare Research and Quality (AHRQ) is sponsoring a two-day conference focusing on how to improve the DDI evidence base and how it is utilized to ultimately decrease the occurrence of these preventable medical errors. I will be speaking at this conference and would like to share two of the planned objectives for my talk with DBMI colloquium attendees. The first objective will be to identify core knowledge elements for DDI decision support and suggest the possibility of a common model for representing and sharing DDI knowledge. The second will be to suggest how further research on clinical trigger systems could lead to reduced DDI alert fatigue while improving patient safety.

For more information: www.dbmi.pitt.edu or 412.647.7113