Document History
- Initial Draft (Matvey Palchuk)
- Initial Draft Continued (Philip Trevvett)
- Formatting for web, split content between "SHRINE" and general SHRIMP concepts (Andrew McMurry)
- Added "Supporting Figures" (Andrew McMurry)
Table of Contents
- Demo
- SHRINE Ontology
- Overview
- Fragments
- Concepts
- Mapping Files
- Interface for Mapping
- Mapping Types
- Supporting Figures
Overview
SHRIMP was developed in support of the SHRINE program to translate between local medical vocabularies (ontologies) and nationally recongized standards defined by HITSP, ICD9, and RxNorm. SHRIMP is data agnostic -- it can be used to Map any Terminology to any other Terminology that is a "bipartite graph" (Wikipedia).SHRIMP can also be used to create concept "Hierarchies" to allow queries to occur on high level concepts, thus making it easier to query systems with a large number of concepts and aggregate medically related facts.
Fragments
Individual vocabularies used by SHRIMP are referred to as fragments. Fragments can either be core or local. A core fragment contains the terms from the core ontology to which local terms are mapped. Each data type (age, gender, diagnoses, etc) has one core fragment that represents all terms of that type in the core ontology. A local fragment represents the terms of any data type used by a local institution. Fragments do not display folders that hold sets of terms and therefore do not represent hierarchies of either the SHRINE Core ontology or local institutions.
Concepts
Each fragment is made up of a set of items known as concepts. Concepts are the specific terms or instances between which all mappings in SHRIMP are created. Each concept may have a set of attributes associated with it, and may also be mapped to any number of other concepts. 5 different site concepts may be mapped to a single core concept, and similarly, 5 different core concepts may be mapped to a single site concept, depending on which fragment contains more granular concepts.
Mapping Files
Association mappings are created that represent mappings FROM terms in local fragments TO terms in core fragments. These mappings are stored in a mapping table that contains the unique ID for each term, the relationship type, the mapping source, and any comments about the mapping. Mappings may be created within SHRIMP or imported from an already existing file. The set of all mappings between a single local fragment and a single core fragment can be represented in a single Mapping File.
Interface for Mapping
Mappings between fragment terms can be managed individually through the mapping interface. This interface allows the user to select a term from a local fragment, search the corresponding core fragment for relevant terms, view all mappings already created, destroy old mappings and create new mappings between local and core fragment terms.
Mapping Type
Mappings between local terms and core terms represent specific relationships such as synonym, has ingredient, etc. New relationships can be created within the tool by clicking on the "Manage Mapping Types" link. When creating a new mapping between terms, the user can select what type of relationship the mapping represents. The user can also designate a "Default Mapping Type" to represent a certain relationship whenever mappings are created, unless the user specifies otherwise. The "Default Mapping Type" can be designated by clicking on the "Users" link and selecting "Edit" for the appropriate user.
Supporting Figures
. 
Figure 1: Mapping local concepts to a global (core) concept space.
In the simplest case, one or more local codes map to a single global concept.
In a more concept case, a local code may map to many different global concepts.
Figure 2: Mapping local concepts to a global (core) concept space.
By assigning hierarchies we can allow a single local concept to be searched through higher paths.
Figure 3: Example of selecting hierarchical concepts during a search.
This makes searching easier and can also be used to aggregate medically related concepts for more robust studies.