BCIG SPEAKER EVENT: “Applying Visualizations and Data Mining Techniques to Clinical Data”

DESCRIPTION: Physicians and healthcare providers are continually bombarded with new information which drives the medical field and medical decision-making. As a result, information systems have become both very comprehensive over the years and very complex. This complexity is due to the increasing expansion of application areas and the depth of those applications. Data is usually filed in many different database locations, described in multiple data types, and interacted with in a large variety of computer languages. While the capture and automation of this data is better than manual records, it does place a burden on healthcare professionals who must be able to quickly find information, assimilate the data and make critical decisions. This is difficult because typical GUI tools are application-specific and viewing medications, for example may require looking at one page, while lab results may require looking at another, Allergies, problem lists, patient history, etc... can be scattered over various menus and difficult to navigate quickly. Decreasing the time a physician spends moving through menus to gather data can result in faster, more accurate decisions and more time to focus on the patient during an encounter. This talk focuses on two areas currently under development: a command line user interface with a more natural language style basis for interaction, and a timeline-based GUI program. The command line interface program behaves much like a domain specific Google, while the GUI program iconically displays clinical data with or without disease focus. Both consume legacy data and emit functions via web services to allow more universal and modern access methodologies. Also in the talk there will be some discussion on data mining and exploration of clinical data with clinical data sets and experiments to detect the sensitivity to various common data mining algorithms. There are some interesting similarities between molecular exploration and data mining that will be shown. It has become clear that just mapping data to online databases does not allow for sufficient data cleansing and complicates data queries. This part of the talk will describe typical kinds of data that are collected in a large healthcare organization and difficulties with bad input data. How this erroneous data is corrected and standardized for examination and queries will be discussed.

3:00 - 4:30 pm September 13, 2007

Augie Turano, Ph.D.
Department of Veteran Affairs
IT Logistics
Augie.turano@va.gov
Phone: 412-241-0217

Dr. Turano received at BS in chemistry, and Ph.D. in structural biophysics at the University of Pittsburgh. His Ph.D. research involved the molecular structure determination of Vitamin B1 and protein structures, and calculation of the electron density distribution of several molecules. This type analysis led to a great deal of computer programming involving nonlinear equations with large data sets of X-ray and neutron data. Dr. Turano worked as a software engineer in the Veterans Administration for 20 years developing medical application software and acting as CIO for a critical care hospital with several sites. He was recognized nationally on several occasions and received a White House commendation from former President Bush for work on pharmaceutical programs for the VA. After several years as the Federal Healthcare team architect at Microsoft, and VistA architect at HP, he has returned to the VA. Currently, he works with the corporate data warehouse team in Healthcare Informatics and the Information Logistics team in procurement activities. As a solution architect he is focused on database technology and software construction performing data manipulation, reporting and data mining. Dr Turano also is an Adjunct professor at the University Pittsburgh in the School of Health Related Professions, where he teaches graduate technology classes related to healthcare.

Related Links