Research

Current Projects
 
 
Leveraging Informatics to Improve Mental Health Services

This project aims to (1) develop innovative data science methods and tools toward a better understanding of data generated in mental health services; (2) enhance prevention and intervention of mental health problems for the vulnerable population through informatics.

A sub-project that aims to improve mental health services to LGBT-identified community members is sponsored by a grant from the Living Well Foundation. PI: Dena Abbott, PhD. Co-PI: Chen Liang, PhD.

Previous Projects
 
 
Distributional Semantics to Characterize Cognitive Patterns in HIV/AIDS Online Community

In 2015, Louisiana was in 2nd place in the HIV infection rate (24.2%). To promote HIV prevention, shared stories and conversations from HIV/AIDS online communities provide valuable information to extend public health knowledge. However, exploiting such information presents significant challenges. For example, the available data contains a great portion of highly unstructured data, i.e., free text, introducing notable linguistic complexity. In addition, such data may be better understood if they could be interpreted into human cognitive and behavioral patterns. Building on distributional semantics, we extracted data from HIV/AIDS online communities to portray users’ cognitive and behavioral patterns, e.g., false belief about HIV/AIDS, behavioral change. These cognitive and behavioral patterns are discussed further to enhance health promotion.

This project is sponsored by a grant from the College of Applied and Natural Science at Louisiana Tech University. PI: Chen Liang, PhD.

Understanding Patient Safety Reports via Multi-label Text Classification and Semantic Representation

Medical errors are the results of problems in health care delivery. One of the key steps to eliminate errors and improve patient safety is through patient safety event reporting. A patient safety report may record a number of critical factors that are involved in health care when incidents, near miss, and unsafe conditions occur. Clinicians and risk management team can generate actionable knowledge by harnessing useful information from the reports. Through the nationwide patient safety event reporting initiative, the increasing volume of reports has been driving improvement in quantitative measures of medical errors, such as statistical distributions of errors types. However, the number of preventable medical incidents is still climbing up. In our view, inspired by the Swiss Cheese Model, identifying system vulnerabilities in healthcare is the key to learning lessons from recurrent incidents because human (individual) errors are inevitable.

We found that the narrative data (e.g., event description in free text) in patient safety event reports carry valuable information that may disclose system vulnerabilities. These data contain clues about interconnected procedures, shared responsibilities, contributing factors, etc. but are challenging to be captured by structured data entry (e.g., the AHRQ Common Formats; the International Classification for Patient Safety) or processed by the manual review in a timely manner. Therefore, we incorporated semantic web ontology, natural language processing, machine learning, and human evaluation to facilitate data processing and knowledge extraction of these narrative data. The extracted knowledge can help clinicians and risk management team to amend clinical protocols, regulations, and health information technologies with the goal of reducing medical errors. Our approach also holds potentials to reduce the workload of commonly used aggregate analysis by incorporating with existing event reporting systems.

This project is partially sponsored by an R01 grant from NIH and a grant from the University of Texas System. PI: Yang Gong, MD, PhD.