Informatics Research Seminar: Beyond Relational: A Database Architecture for Clinical Data Management

 November 13 @ 4:00 – 5:00 pm

 

Speaker: Ray Hylock, PhD
Presented from ECU

Broadcast Link: Seminar

 

Abstract:

The need to analyze data collected by electronic health record (EHR) systems is becoming an increased reality for many institutions and healthcare providers, whether in response to meaningful use or clinical research. Results have the potential to improve diagnoses, treatments, outcomes, and the cost of patient care. Typically, data is transferred from the online EHR to a secondary database prior to data interrogation. The design of this external storage engine – either relational or entity-attribute-value (EAV) – varies depending upon the anticipated workload of either attribute- or entity-centric queries. Unfortunately, clinical data warehouses are often presented with both forms.

The EAV design maintains many theoretical advantages over relational in the health domain. However, since its inception, relational database management systems have been employed to maintain EAVs, mitigating many of its benefits. In this presentation, the foundation for an EAV database management system is presented along with conversion and query comparisons to state-of-the-art models in this space. The results indicate the proposed executes both entity- and attribute-centric queries as well as or faster than relational and EAV systems and require less conversion time and storage space than all implemented EAV methods. This work represents a portion of the author’s dissertation.

Biosketch:

Dr. Hylock is an Assistant Professor in the Health Services and Information Management department at East Carolina University. He received his Bachelor of Science in Business Administration (High Technology Management) from California State University, San Marcos, and his Master’s and PhD in Informatics (Health Informatics) from the University of Iowa. Broadly, his research foci are clinical data warehousing, federation, optimization, and algorithms.