This proposal aims to devise algorithms for ad hoc (non-real-time) query of heterogeneous electronic patient record systems (EPRSs) that store patient data in at least three forms: Conventional relational structure, with an attribute represented by a column in a table, Entity-Attribute-Value (EAV) form, where one or more tables record information that includes entity (Patient, Date/Time of Event, etc.), the name or ID of the Attribute on which information is to be stored, and the value corresponding to that attribute, Free text or quasi-structured textual information, such as discharge summaries, and surgical operation notes. The proposed work will evaluate these algorithms in at least two production systems, ACT/DB and a pilot data warehouse at the West Haven, CT, Veterans Administration Medical Center. The user interface will be Web-based and will be designed for expressive power as well as ease of use by analysts who are non-database professionals. Additional research issues will involve integration of the query system with one or more controlled medical vocabularies, integration of information-retrieval (IR) technology with relational database (RDBMS) technology to provide advanced query functionality, and support of complex temporal queries on the data. Finally, the proposed work will explore architectural issues impacting the efficiency of the query process. In particular, research will be performed on utilization of newer database technologies, such as new indexing methods, as well as parallel database implementations.
Showing the most recent 10 out of 15 publications