ALL-RELATIONAL KNOWLEDGE BASE
Park Street's knowledge representation system provides an efficient repository for ontologies that capitalizes on all of the advantages of relational database systems. The knowledge base relies upon proprietary technology for managing directed graphs and semantic networks in the relational environment, allowing developers to implement sophisticated queries using pure declarative SQL rather than complex, inefficient algorithmic solutions.
CLINICAL INFORMATICS DATABASE
Clinical ontologies and coding systems are available from the National Library of Medicine and other sources — but not in a form that can be easily used for any practical purpose. Park Street has defined a complete informatics database using its ONTX knowledge representation system, and has developed the tools required to load and update the database with a full set of essential ontologies.
PATIENT FACT STORE
The Park Street Active Fact Store provides a complete longitudinal record of patient facts using a single, simple representation. In the Park Street database, every patient fact is represented as the intersection of a knowledge base concept with a patient or visit at a point in time. Types of patient facts include administrative data (enrollment, benefits, claims, payments), registration events (admissions, discharges, and transfers), diagnoses, procedures, orders, results, medications, observations, assessments, documents, images, and more.
DATA ACQUISITION TOOLKIT
The Park Street data acquisition system is a high-performance framework for receiving or extracting data, transforming and aligning it to standard ontologies or custom knowledge models, and preparing it for export or bulk load. It includes a complete job execution framework that orchestrates multiple-step ETL jobs, supports serial and parallel execution of sub-tasks and individual task steps, and handles dependencies and job eligibility checking.
The core data model of the Park Street generalized data warehouse is de-identified according to HIPAA standards, so that all analytic queries can be executed without exposing personal health information. Identified data is available for administrative and operational uses, but is segregated so that only users and applications with appropriate rights are allowed to access PHI or re-identify patient facts.
The Park Street architecture includes sophisticated support for the temporal aspects of clinical informatics. Database-resident state machine tools are provided to transform event streams into knowledge base defined maps of patient state over time. Starmaker end-user applications enable enterprise-wide self-service for clinical, administrative, and research users who need data for analysis but lack database navigation skills.
RIGHT TIME BINDING
Traditional repository designs transform data to an enterprise data model as it is received, which can limit flexibility and create operational difficulties. Late binding designs reflect the modeling found in source systems, reducing data acquisition challenges but moving the burden of data transformation and normalization to each and every use of data. Because the Active Fact Store represents the complexity of patient facts using highly flexible knowledge structures, it avoids the brittleness of early binding while offering analysts a query-ready data resource.
Most healthcare institutions already own tools for data analysis and visualization. The most serious challenge involves getting data into the hands of analysts so that they can put their tools to work. Park Street's Starmaker tool allows users to specify data requirements with simple drag-and-drop techniques. Behind the scenes, Starmaker orchestrates a sequence of fact selection algorithms to find and stitch together disparate fact types. Starmaker provides access to the advanced computing capabilities of the Active Fact Store, simplifying its abstractions and adapting it to the perspective of untrained users and common reporting tools.