RIM-Recasted, Value-Added Efficiency Interpolation in the HL7 Development Paradigm

ISSN (e): 2250 – 3005 || Volume, 06 || Issue, 06|| June – 2016 ||
International Journal of Computational Engineering Research (IJCER)
www.ijceronline.com Open Access Journal Page 43
RIM-Recasted, Value-Added Efficiency Interpolation in the HL7
Development Paradigm
M.I. Sabar1
, P. M. Jayaweera2
, E.A.T.A. Edirisuriya3
Department Of Computer Science,University Of Sri Jayawardenapura, Gangodawila, Nugegoda, Sri Lanka.
I. INTRODUCTION
The Reference Information Model (RIM) constitutes the encyclopedia of all healthcare-related terminology
utilized in HL7-based, IT-driven Healthcare Information Systems.
 Release 1 in 2003, ISO certification in 2005 – 2007.
 Release 2 released in 2010.
It is a Reference Ontology consisting of six core classes, ie., four Generic classes and two Helper classes
namely Entity, Role, Participation, Act, Role Relationship, and Act Relationship [1].
The RIM expresses a diverse, expansive space of information content, within a unified framework. It affords
consistent meaning to all usable healthcare-domain related terminology, beyond all local contexts. This
comprehensive source of pertinent system vocabulary covers the spectrum of possible HL7 specifications,
promoting overarching Semantic Interoperability amongst all domains. Indeed, RIM is the source from which all
ABSTRACT
The Medical fraternity and the healthcare service sector have long acknowledged the need for
smart, IT-based healthcare systems, operating globally. Semantic Interoperability is key, which is
the regulated and meaningful exchange of valued healthcare information with homogenous
understanding amongst participating healthcare service providers. Health Level Seven (HL7) is
the predominant interoperability-related global healthcare standard in operation today.
Introduced in 1987, the standard has evolved to its current version 3 and has been embraced by
the National Health Services of the most developed economies in Europe, North and South
America, and Australasia.
However, the standard is not without issues. Version v3 has been found to be difficult to
implement and maintain. A principle component of the HL7 v3 development paradigm is the
Reference Information Model (RIM) which defines the complete language and vocabulary
schema used in the three v3 paradigms of Messages, Clinical Document Architecture, and
Services, and indeed in all v3 implementations. This study determined that the RIM itself has
many documented issues which ultimately affect implementation. Thus, true global semantic
interoperability which is the germinal goal of the HL7 standard is still an illusion.
This study focuses on the belief that the achievement of true global interoperability is rooted at
the labyrinths of specifications development and its associated foundational paradigms. This
study focused on the Reference Information Model (RIM), the foundational semantic and lexical
reference structure which affords vocabulary derivation to be used in all v3 system
implementations. Infusing sequencing and temporal dimensions to the RIM structure and
operation would promote and afford enhanced analytic, design, semantic interoperability, and
two-way traceability, which in turn would suffuse to high-calibre specifications generation and
true global International Interoperability in operation. In addition, multi-faceted interoperability
interpolation in these core processes would promote and enhance numerous allied activities as
well, from domain requirements cross-checking, audit, and consensus, to kindred system
development verification and validation.
This research therefore analyzed many of the prevalent RIM issues indepth, and effected smart,
delicate, and prudent recasting of this encyclopedic vocabulary and language reference
structure, to derive optimal efficiencies in specifications development and implementation.
Keywords: Act, Health Level 7, Object-Orientation, Reference Information Model,
Relationship, Role, Semantic Interoperability.

RIM-Recasted, Value-Added Efficiency Interpolation In The HL7 Development Paradigm
HL7 v3 Information Models are derived, eg., Domain Information Models (DIMs). It defines the healthcare
grammar in any environment.
However, according to [2], the RIM has many logical and ontological flaws, placing severe obstacles to
efficient, RIM-compliant system implementations. It is incoherent and inconsistent in places, leading to
ambiguity. These flaws ripple to other kindred components or phases in the HL7 v3 specifications development
continuum, such as the paradigmic Messages, Clinical Document Architecture (CDA), and Services. These
deficiencies permeate beyond the generated finalized specifications to actualized system implementation
affecting our principle goal of International Interoperability (II) and inclusive network efficiency. Some of the
documented RIM-related issues are presented in the next section.
II. RIM ISSUES
1. NULL and NULL flavoured attributes should be handled with more clarity [3].
NULL attribute – modelled by a missing attribute.
NULL Flavour – modelled using a placeholder.
Greater Semantic identifiability desired.
2. RIM Object-Orientation (OO) violates sacred OO principles [4].
3. RIM does not provide distinction between an Information Model (IM) and a Reference Ontology (RO) [2][4].
This is not exactly true; indeed the RIM is clearly a Domain Ontology from which the specific domain
environment vocabulary is derived, and the allied Information Model represents the data and information
annotated to the RIM-derived information structure. The reference ontology defines the world in structure and
bound, and the information model fills in its instantiated attribute values. However, a clearer demarcation
between the two viewpoints is desirable.
Egs., i. “person is a human being” in RO versus person record object of demographic information in IM.
ii. “checkPressure” in RO versus record pressure reading in IM, or
iii. (existing real-world analysis) versus ( class-objects based information system design)
The RO versus IM confusion can also be more complex, as shown below.
Instruct patient to stretch arm - SPEECH ACT
“TakePressure”ACT ⟹ Fix pressure tubing on arm-PHYSICAL ACT ACT Relationship
Read pressure - OBSERVATIONAL ACT
Above signifies RIM as a REFERENCE ONTOLOGY (RO)
Record Pressure reading
RIM-driven Information Model (IM)
4. ENTITY component not precisely defined in RIM. Information recorded about an ENTITY and the ENTITY
itself not distinctly identified, eg., the RIM class LivingSubject contains person, dog, plant, mammal, etc.,
versus information about such entities. Both bundled together due to Object-Orientation, but clearly deficient
in exuding RIM-related structural, space, and informational benefits during specification generation.
5. ACT component has similar issues. Information about an ACT (recorded) not clearly distinguished from the
ACT itself (occurred).e.g.,
Instance of patientEncounter (subclass of ACT) versus ACTion itself (Physical, Speech or Observation ACT).
Presently RIM switches back-and-forth between the two; thus ambiguity ripples through related processes.
6. Object-Orientation of RIM objects cause consistency issues, eg., Consider person instantiations such as John
Silva, Mary Perera, Justin Alwis, and so on. Currently, object record of John Silva holds his demographic and
medical information. There could be many John Smiths; how can we converge on required John Smith object
record? Use of unique record identifiers like National Identification Card (NIC) or Social Security Number are
recommended in such situations for precision target binding. Obvious distinction between instantiated person
(Reference Ontology) and the record of his/her demographic information (Information System) is desired, but
the object record does not indicate this.
7. Antiquatedness of current HL7 v3 message-driven methodologies is another issue. In keeping with current
trends, web-centric, service-based technology is preferred; the RIM should be recast accordingly [2].

8. RIM expanse is strictly scoped to cover healthcare-related vocabulary; ancillary terminology not directly
affined to core healthcare processes are not included in RIM. To what extent can the RIM be applied for
coherent interoperability in such situations.
9. Currently, the RIM structure consists of a flat representation of six core classes, ranging from static classes
(ENTITY), static relationships (ROLE, ROLE RELATIONSHIPS), to dynamic, action-related representations
(ACT, PARTICIPATION, ACT RELATIONSHIP). Check the relationships given below.
ROLE ROLE RELATIONSHIP: dependency-type link, static as defined at creation
(NOT dynamic or mutative), e.g., accountability, subordinate, peer, etc.
ROLE PARTICIPATION ACT: dynamic, created on-the-fly.
PARTICIPATION binds pending ACTs to ROLEs (ENTITIES-IN-ROLES).
ACT ACT RELATIONSHIP: totally dynamic.
Presently the RIM structure possesses components which vary from statically-defined and statically-related, to
fully-fledged dynamically defined and dynamically related, in the same diagram and space. Thus, ambiguity
abounds; part processing and part instantiations lead to incoherent, incongruous, and inconsistent RIM-based
domain structures. The chance of in-process verification, validation, and traceability is low, and the finalized,
static snapshot of the RIM is required for further processing. This terminal snapshot defines any domain-related
RIM vocabulary with any resulting ambiguity and inconsistency rippling down to subsequent phases of the
specifications development continuum. Clearly, a distinctive demarcation of the static and dynamic facets of the
RIM is desired in order to eliminate ambiguity. Further, sequencing and temporal injection into inherent RIM
processes would enable score boarding of valuable sub-process data which occur ahead of the terminal
snapshot, affording much desired verification, validation and traceability.
10. International Interoperability, a significant enhancement and a principle goal of this research and study, needs
to be infused into the RIM Ontology.
11. In conclusion, the above issues lend enough ambiguity to the original RIM description in [2] as a “credible,
clear, comprehensive, concise, consistent, universally applicable, and extremely stable” referential structure.
III. PROPOSED RIM REFACTORING SOLUTION
The HL7-based RIM (Reference Information Model) is Reference Ontology consisting of six core classes, ie.,
four Generic classes and two Helper classes. They are [1] :
Entity: Static, library-data. Physical things and beings, eg., person, organization, creatures, devices, animals,
etc.
Role: Instantiations of ENTITY in healthcare environments, eg., patient, doctor, investigator, nurse,
CEO(Entity-in-a-Role).Captures competency, capability, certification (3 C’s).
Participation: context or scoping for an ACT, e.g., who, when, where it was done, e.g., practitioner, giver,
recipient.
Act: an action meaning actual performance, e.g., procedure, observation.
Role Relationship: relationship between ROLES, e.g. dependency relationships, e.g., client-of, reporting-to,
assistant-to.
Act Relationship: relationship between ACTs.
Figure 1 Present RIM Structure
The solution approach was two-phased. Firstly, the current Reference Information Model (RIM) was analysed
and delicately recrafted in three aspects; injection of International Interoperability (II) the principle goal of our
research, infusion of a temporal dimension to the hitherto placid RIM with a view to interpolating automation in
the RIM-related processes such as harmonization, access, adoption, and finally the promotion of expansive RIM

applicability and usage as deemed necessary, in allied and ancillary processes of the core healthcare domain.
The last option is afforded herein but optional during implementation given the requirement of RIM conciseness.
Secondly, the present XML-based (Extended Markup Language) RIM representation was remodelled using the
proposed Unified Data Atom (UDA) language accruing abounding intra and inter-process merits from the RIM
harmonization, access, and adoption processes.
The recrafted RIM Ontology is given below in Figure 2, so termed since the figure gives some semblance of
component constituents and inter-relationships.
Core: Current core healthcare-process related ENTITY items, eg., person, disease, organization, animals,
mammals, etc.
Ancillary: Valuable additions in keeping with principle goals, such as the following.
1. Newly infused terminology related to International Interoperability (II), eg., Continent, Country, Service
Provider, Application, Database/Cloud, Patient, UID, OID, etc. Some overlap of terminology with CORE
ENTITY exists.
2. Useful terminology extraneous to core healthcare processes, their scope, extent, and even adoption
dependant on client and implementor, e.g., agent, medical representative, distributor, drug-substitute, etc.
An open-source ANCILLARY component befits this requirement.
Entity: Now a blend of CORE and ANCILLARY, also incorporating infused temporal and sequencing aspects
(as indicated by the *). Each new propogation via the ENTITY component with respect to RIM
harmonization, access, or adoption is sequenced and time-stamped enabling sub-process verification,
validation, and traceability between refinements. Indeed, this also affords seamless automation of these
RIM processes ensuring greater governance and overall consistency of the RIM. Each successive time-
stamping veritably captures a successor phase in the overall RIM processing workflow. This can even be
transcribed onto a timeline of phased activity. Presently, there is no explicit time-relatedness in RIM. All
current time-related attributes are deeply embedded in relevant class structures presenting little or no value;
only the final snapshot of any domain-related RIM being used for further processing.
Figure 2 Proposed RIM Ontology
Act, Act Relationship, Role, Role Relationship are as present.
Participation component was found to be superfluous, often under-used and sparse (few entries) and thus was
dropped for brevity’s sake. This is significant in order to maintain RIM conciseness because of the two new
components CORE and ANCILLARY (convergent on ENTITY) inserted into the arrangement.
Look ahead on required/pending ACT/s is needed before the ENTITY component can spawn the relevant
ROLE subset (Entities-in-Roles) which would perform these ACT/s. Apparent disassociation between the
ROLE and ACT components until ultimately bound by the PARTICIPATION component has been rectified.
This present process is semantically flawed; it is not possible for the ENTITY component to spawn a
pertinent ROLE subset without knowledge of the pending ACT/s. otherwise an un scoped, infinitely
expansive ROLE set would result. Further, sequencing and temporal refinement of the current flat RIM
structure has been proposed (as shown in the diagram below), in order to infuse future automation to native
RIM processes; indeed validation and traceability of inherent sub processes is another derived bonus.

IV. SOLUTION METHODOLOGY
1. NULL/NULL flavour: handled through NULL value entry in attribute or NULL demarcator (2 distinct
options - Induced Semantics). The NULL demarcator supplants the previous “absence of attribute” option
infusing clearly defined semantics to the RIM structure.
2. Object-Oriented principles not relevant anymore (most logical/ontological flaws due to OO representation).
The proposed UDA representation is clearly a departure, indeed an significant enhancement on present OO
representation, and avoids and eliminates known, inherent demerits.
Figure 3 Temporal Phase-wise RIM Refinement
3. Proposed UDA vocabulary models Ontologies (the world) and Information (about the world) with equal
adeptness. RO and IM handled individually and distinctly. No concept of record objects or classes which
cause the ambiguity. Indeed, the proposed UDA representation models RIM objects more closely with greater
refinement and precision.
4. ENTITY component of RIM now precisionly defined in UDA representation. ENTITY versus information
recorded about an ENTITY now distinctly identified. eg., the RIM classes such as LivingSubject modelled by
Demarcators, whilst information about such entities actualized by regular DataAtoms. Similar modelling
strategy used for ROLES.
5. No confusion in representing ACTs. Report/record of performed ACT (DataAtomised) separately identified
from ACT itself (Demarcated). Actual ACT is not a propositional comment anymore, Eg., Actual ACT (not
its record) versus patientEncounter (instance of ACT), now stable, distinguishable. With proposed solution
representation, RIM stays consistent and stable as defined, without switching back and forth (Issue No. 5.)
6. Object-Orientation of RIM objects causes consistency issues, eg., person instantiations such as John Silva,
Mary Perera, Justin Alwis, and so on. Currently, object record of John Silva holds his demographic and
medical information. The obvious distinction afforded between instantiated person (Reference Ontology) and
record of his/her demographic information (Information System) by the proposed UDA language, in addition
to the inherent uniqueness of its data storage eliminates all consistency issues. Further, unique record
identifiers like NIC are used in dealing with Syntactic Synonyms, eg., identically-named person instantiations.
7. Indeed, commensurate with current technology, web-centric, service-based technology is recommended, and
the RIM should be recast accordingly. The current HL7 v3 message-driven methodology is antiquated and
decrepit and it is advisable that it should be discontinued.
8. Presently, the RIM scope covers only healthcare-related vocabulary; however, it is proposed that ancillary
terminology affined to allied processes and objects (outside the core healthcare periphery) be also available,
if not directly incorporated in the RIM. An open-source approach can be used for this ANCILLARY
component, and the client/implementor allowed discretion on its inclusion. This valued insertion to the RIM
ontology would afford enhanced international interoperability and interphase cohesion in the HL7
specifications development continuum.
9. Presently the RIM structure possesses components which vary from statically defined and statically related,
to fully-fledged dynamically defined and dynamically related, in the same diagram and space. Thus,

ambiguity abounds; part processing and instantiations lead to incoherent, incongruous, and inconsistent
RIM-based domain structures. The proposed, recast RIM ontology annotates dynamic, temporal, lookahead,
and sequencing features thereby accurately and closely modelling the actual RIM an its many evolutions.
This enhanced, fine-grained RIM actualization exudes the following merits:
 The chance of in-process verification, validation, and traceability, instead of the terminal, static snapshot of
the RIM; presently the mix of passive and active domain features on the same RIM diagram results in
ambiguity and inconsistency rippling to subsequent phases of the specifications development continuum. The
new RIM ontology provides a distinctive demarcation of the static and dynamics related facets of the RIM,
which eliminates ambiguity.
 The lookahead used on the ACT component enables a convergent, concise, and pertinent ROLE set to be
spawned. Otherwise a divergent, inordinate ROLE set would result.
 Temporal and sequencing injection into inherent RIM processes would enable scoreboarding of valuable sub-
process data which occur ahead of the terminal snapshot, affording much desired validation and traceability.
10. Interoperability, a significant enhancement and principle goal of this research and study, is infused into the
RIM ontology via the ANCILLARY component.
11. The proposed RIM-related structural enhancements and representational vocabulary promotes, enhances, and
re-articulates the original RIM description in [2] as credible, clear, comprehensive, concise, consistent,
universally applicable, and extremely stable.
V. SOLUTION FORMALISM
The UDA vocabulary formalism is presented below. The Necessary condition for transformation TR
X-U
(meaning
RIM transformation from XML to UDA) is stated as follows :
If U signifies the set of transliterated, target DataAtoms {u1, u2, u3, u4, ……………., uk} as a result of the
Complete Transformation TR
X-U
acting on the source XML-based RIM schema X where {d1, d2, d3, ………., dl}
ЄU represent source RIM elements, then
TR
X-U
: X U X ⊑ X and U ⊑ Ṹ where
X - Problem domain RIM-related XML super schema, and
Ṹ - Problem-related target UDA super schema
U ⊂ U(ui uj) where {i, j, = 1, 2, 3, , …, k}
where U : set of spawned target DataAtoms
U : union of bidirectionally inter-connected, target DataAtom pairings (all-over, complete
connectivity, all pairings) as defined by the UDA structure definition. (i : index of ui, 1 ≤ i ≤ k,
j : index of uj, 1 ≤ j ≤ k)
It was proved that the mapping TR
X-U
: X⟶ U denotes a Complete Transformation, meaning the result of the
transformation TR
X-U
is a necessary and sufficient target set U in relation to the source set X. This would also
satisfy the necessary condition for the X ⟶U mapping. By definition,
Injection of TR
X-U
: ∃ 1:1 mapping TR
X-U
of elements from domain X to codomain U
∀diЄX Ʌ ∀ujЄU, iЄN Ʌ jЄN (di are RIM elements)
Surjection of TR
X-U
: ∀ujЄU there exists diЄX such that TR
X-U
(di) =uj
In other words, TR
X-U
: X⟶ U is an “onto” relationship.
Thus, every member uj of target co-domain U is mapped onto by at exactly one diЄX. There are no unmapped
elements in either X or U.
Hence, TR
X-U
: X ⟶U represents a Complete and Sound Transformation.
The following were also proved in this study for the XML to UDA transformation or mapping (TR
X-U
in this
case).
1. Equivalence.
2. Completeness and Exhaustiveness.

3. Syntactic Completeness and Validity.
4. Semantic Validity.
5. Preservation of XML source precedence and nested ordering.
VI. RESULTS
The table below illustrates the modality of the TR
X-U
transformation. Indeed this exercise is extrapolatable to all
RIM structures.
Table 1: RIMXML to UDA Transformation and Syntax Mapping [6]
VII. DISCUSSION
Thus, the proposed structural enhancements of the RIM, including the much-sought International
Interoperability and the optional ancillary information related enhancements, look ahead, temporal, and
sequencing injections at the head ENTITY component, all blended with the fine-grained, enhancing, more
appropriate UDA-modelled internal representation, derive optimum benefits and facilitate high-calibre
specifications development. Further, ENTITY and ROLE components were often found to have identical tags,
e.g., patient, surgeon etc., although used in the passive (ENTITY value is static) and active (ENTITY-IN-A-
ROLE is dynamic), thus creating unwarranted ambiguity. This has been resolved in the new, proposed RIM
Ontology. In addition, the PARTICIPATION class was found to be relatively sparse, often consisting of
repetitive entries. Since it was determined that no undesirable effects would accrue, this component was
removed in the new proposed RIM ontology. The semantics of the ROLE–ACT relationship can still be
maintained with the UDA link component, thus avoiding any loss of information and ensuring the desired
structural brevity.
Hence, the revamped, recast RIM ontology affords and facilitates high-calibre specifications generation,
promoting true, network-wide International Interoperability and inclusive efficiency (our principle goals) during
system implementation and operation.
REFERENCES
[1]. Dillert, J., “HL7 RIM – An Introduction for Non-Technical Professionals”, PhUSE Paper CD04, 2013.
[2]. Smith, B., Ceusters, W., “HL7 RIM : An Incoherent Standard”, Studies in Health Technology and Informatics, Vol. 124, 2006.
[3]. “RIM-Based Templates (TRIM)”, TolvenHealth, 2009.
[4]. Schadow, G., Mead, C., N., Walker, D., M., “The HL7 Reference Information Model Under Scrutiny”.
[5]. Sabar, M. I., Jayaweera, P. M., Edirisuriya, E. A. T. A., “Requirements Traceability Interpolation For HL7 Specification Generation”,
Postgraduate Institute of Science (PGIS) Research Conference, University of Peradeniya, Sri Lanka, October 2015.
[6]. Sabar, M. I., Jayaweera, P. M., Edirisuriya, E. A. T. A.,”Techno-Platform Independance Interpolation in HL7 Ontologies”, eHealth-
Asia Medical Conference, Colombo, Sri Lanka, October 2015.
[7]. “HL7 International New”, Health Level 7 International, May 2011.
[8]. Spronk, R., “History of the HL7 RIM”, Ringholm Whitepaper v1.0, September 2011.
RIM Element XML
Representation
Example Target UDA
(Syntax Mapping-Satisfied by TR
X-U
)
Generic Class
Helper Class
Sub-Class
Attribute
Inter-Class
Multiplicity/
Cardinality
Inter-Class
Association Label
General Constraint -
DataType
Vocabulary
Constraint
Relationship
Class
Class
Sub-Class
Attribute
Inter-Class
Multiplicity/
Cardinality
Inter-Class Association
Label
DataType
Vocabulary Constraint
Relationship
CORE, ANCILLARY,
ENTITY, ACT, ROLE
ACT RELATIONSHIP,
ROLE RELATIONSHIP
Patient, Country, UID
Attribute
Inter-Class
Multiplicity/Cardinality
Inter-Class Association Label
DataType
Vocabulary Constraint
client-of,peer-of, practitioner,
instance
RIM Structure Demarcator (yes)
RIM Structure Demarcator (yes)
Dataset Demarcator
LabelDataAtom, (yes) ValueDataAtom
(yes)
Property Demarcator
(Value and Name)(yes)
Intrinsic DataAtom DataType (yes)
Restriction Demarcator (yes)
Property Demarcator
(Value field indicates direction of
relationship) (yes)

RIM-Recasted, Value-Added Efficiency Interpolation in the HL7 Development Paradigm

More Related Content

Viewers also liked (11)

Similar to RIM-Recasted, Value-Added Efficiency Interpolation in the HL7 Development Paradigm (20)

Recently uploaded (20)

RIM-Recasted, Value-Added Efficiency Interpolation in the HL7 Development Paradigm