2017 IEEE 11th International Conference on Semantic Computing

Semantic Framework for Electronic Health Records

Ricardo Matos Chaim
Ricardo S. Puttini André Amaro Toffanello
Software Engineering
Network Engineering Network Engineering University of Brasília,
University of Brasília University of Brasília Brasília, Brazil
Brasília, Brazil Brasília, Brazil

Érica Oliveira Carvalho

Gabriela Alves Jussara M. P. Rotzsch
Software Engineering
Network Engineering Network Engineering University of Brasília,
University of Brasília University of Brasília Brasília, Brazil
Brasília, Brazil Brasília, Brazil

Edgard Costa Oliveira

Edison Ishikawa Aletéia P. Araújo
Software Engineering
Computer Science Computer Science University of Brasília,
University of Brasília, University of Brasília, Brasília, Brazil
Brasília, Brazil Brasília, Brazil

Abstract - Many countries are in the run for implementing their in order to obtain historical information, in a safe, integrated
local electronic health strategies, on top of the existing health and opportunistic information care process [1].
systems administrative and clinical procedures. There have been
international successful initiatives in these cases, however, a The sets of information will be available via the
substantial effort is necessary in order to adapt the best practices integration of existing information systems to the recently
to national realities. In Brazil, the effort of implementing a created National EHR (Electronic Health Record). This
technological solution, inspired by these international cases, is system is available to healthcare professionals, service
underway. This paper describes the context where the solution providers as well as the patient himself and all users of
was developed: the implementation of the e-health EHR Brazilian health system – SUS. This solution provides ways
platform in Brazil. It brings the technical specification of a to support clinical decisions, in many levels, aiming at the
semantic framework developed and used in a SOA-based improvement of health services via an integrated health
infrastructure that enables the EHR solutions for the Brazilian attention view, patient-centered and focused on the
public health system. We present here the creation and use of continuity of healthcare.
an Information Model and a Knowledge Model, composed with
the use of archetypes and templates, based on the OpenEHR Thus, the Brazilian Ministry of Health has been
formats. The use of EHR archetype definition languages is also developing an interoperability clinical information systems
presented here. The semantic framework is based on the setting model, described in the technical architecture of the
up of a clinical knowledge model and a terminology server that Brazilian national electronic health record solution. This
makes use of Snomed-CT. solution provides a set of health information services via a
health service bus. They were conceived and developed
Keywords - Electronic Health Record – EHR, Semantic
with the support of semantic resources framework,
modelling, Clinical information repository, health knowledge
model. structured in a clinical knowledge models, based on the
OpenEHR. This solution assures accuracy and security of
shared clinical information even if they come from different
I. INTRODUCTION clinical systems.
The Brazilian Ministry of Health is on the way to
The ministry considers that the Brazilian EHR solutions
implement, by the end of 2016, the national Electronic
is by nature a complex process that is constantly under
Health Record infrastructure and system. This national
evolution. Its services, functionalities and resources are
SOA-based 5-year project resulted in a set of information
incorporated in a scalable and iterative manner, within each
sources and services that provides a health information
project phase. This is why the ministry is working to
distributed repository for users of the whole national health
establish the following EHR structuring system components.
system (SUS). It contains clinical information collected and
shared by various clinical information systems (CIS). • Demographic services: a set of services that provide
Particularly, this infrastructure provides information to be single users identifications in the public health system (SU).
available at the patient´s electronic records (PEPs or EHRs) This service was established based on the National Card
Identification system.

978-1-5090-4284-5/17 $31.00 © 2017 IEEE 334

DOI 10.1109/ICSC.2017.98
 • Health Service Bus: this is the service infrastructure information model. This model must be as compact (small),
created for publication, orchestration, governance and to generic and stable as possible.
make sure that EHR services usage policies are being
The OpenEHR information model classes is based in the
followed.
CIR ontology – Clinical Investigator Record. This ontology
• Security services such as authentication and access places information in relation to its categories:
control (authorization). administrative (admin) information and care information [6]
as presented on figure 1.
• Semantic Framework: consists of a set of services
created specifically to disclose and manage standardized
semantic resources such as terminologies and clinical
knowledge models, based on the OpenEHR model. This
model is mainly used when providing structure for the
shared and stored clinical information.

II.THE SEMANTIC FRAMEWORK

The semantic framework we present in this paper is
based on the production of two different models and levels:
an information and a knowledge model. In the context of the
knowledge model, we introduced the concept of archetypes
[2]. A two-level modelling is the name given to this
approach [3]. These models are directly supported by the Figure 1 - Clinical Information Ontology – CIR [5]
adoption of standardized terminologies. Thus, the CIR ontology categories influence existing
It is important to mention that besides the reference classes of the OpenEHR information model which are then
models and archetypes (including terminology binding, used to classify entrances in order to represent clinical
classifications and term lists), there are other models concepts.
expected by the OpenEHR specification. For instance, the
template model, the queries model and the extract models. IV.KNOWLEDGE MODEL
These models must be equally adopted by the national EHR The knowledge model described here broadens (and
infrastructure designed by the Brazilian Ministry of Health. restricts) the information model, similarly to what happens
Besides these models, it is possible to think of additional to an external ontology. The model specifies structures with
ones such as user interface model, aggregation models and valid concepts, its respective data types, aggregations and
relationships between the archetypes as well as the model terminologies, which requires specific formalisms and
for inference rules. These last models will be included in the structures. In the OpenEHR perspective, these clinical data
next iteration of the project, in the architecture evolution structures used to restrict the information model are the so-
perspective of the Brazilian EHR. called archetypes [3,7,8,9, 10, 11].
In this section, we will present the information model The archetypes [11] enable health specialist to formally
and the knowledge model in detail. define a clinical content. It allows higher chances of
conserving original data meaning via a well-structured and
explicit specification of clinical content, and by simplifying
III. INFORMATION MODEL
the use of clinical terminologies.
In the first stage of our semantic framework, we have a
stable information model, constructed from the An archetype instance [12] is a knowledge artifact that
identification of a small set of generic classes, generic defines how an information model hierarchy should be
enough to represent the EHR-related concepts. In this organized in order to represent the data used in a clinical
model, we avoid using attributes with explicit meanings, procedure or assistance scenario. They can be shared once
leaving any particularities of concepts to the represented in their definitions are represented in a standardized way, and
the knowledge model [4]. An information model, or are used by communities that share patient´s registers in
reference model, is composed of: a) a set of primitive types; order to define how local clinical data should consistently
b) a set of classes that define the information organization in mapped [12].
EHR systems; c) a set of classes that describe context Figure 2 illustrates the relationship between the
information; and d) classes that describe patient´s information model, the knowledge model (archetypes) and
demographic data [5]. the data (instances).
Classes are defined in a generic form, so that
development, maintenance and software evolution or
growth will cause the least impact possible to the

335
 Definition – this section is where constraints are defined
for objects and attributes considered in the archetype. All
constraints are written in cADL.
Invariant – it defines first order logic declarations used for
constraints that are created inside the definition section
structure, mostly for constraints that contain mathematical
formula or logics. Objects and properties are referred to via
paths, and must be written in FOPL.
Ontology – this is where we define the codes that represent
objects (nodes), constraints and links to terminologies. It
has the following subsections: terminologies_available,
term_definitions, constraint_ definitions, term_binding and
constraint_binding. It is written in dADL.
Figure 2 – Relationships between the information model, Revision_history – it is optional and shows the history of
archetypes and instances [7]
changes in the archetype, and it is coded in dADL.
We can observe in these relationships that the data are
The structure in a level above of an ADL is shown in figure
instances of the information model, and the archetypes are
instances of the knowledge model. The knowledge model is 3.
formally coupled to the information model. The data is
submitted to the restriction rules contained in the archetypes
in execution time. Thus, the health domain concepts are
represented in the form of archetypes. [3, 9, 7, 12].
Templates are used to construct forms that capture data that
corresponds to a particular clinic task and can match data
from different archetypes and other templates. Besides that,
archetypes and templates can be matched with clinical
terminologies [3, 9]. The archetypes are usually serialized
with the use of the ADL – Archetype Definition Language
[13, 14]. ADL is a formal language built to do archetypes
serialization, based on domain entities restriction models or
structured business rules. It describes data constraints that
are instances of a reference model and can be used in any
domain where there are formal object models that describe
data instances. We may find three different syntaxes: dADL
(Data ADL or data definition model), cADL (Constraint Figure 3 – Archetype structure in ADL 1.4 (adapted from
ADL or constraint model), and a FOPL version (First-Order [14])
Predicate Logic [15]. The archetypes make a liaison
between the information structures and the formal V.CONCLUSIONS
terminologies, and they can be created and translated into
any language. Besides that, they are path-addressable, The semantic framework contains a set of components
similar to XML files via path expressions that can be and services that are being used together with the for the
converted into Xpath expressions. The ADL language Brazilian national EHR infrastructure, particularly for the
structure is very simple and based on two other syntaxes to structure (design time) and a semantic validation (runtime)
express structured constraints (cADL) and data (dADL). In of shared information via EHR. We will present here two
summary, an ADL code presents the following basic components: the CKM and the terminology server.
sections in the code structure: The CKM - Clinical Knowledge Manager is a
framework component that collects, reviews and distributes
Header – this section describes the archetype identification
knowledge models (archetypes, templates and value sets). It
and its equivalent concept, specialization, original
operates in a distributed approach so that the national EHR
language, possible translations, authoring data, any other may define an instance of this tool and use an international
information that describes the archetype and that can be library (http://openEHR.org/ckm), by refining some of the
used for retrieval from a repository. Subsections are: models for local use and creating artifacts adopted as
archeytpe, especialize, concept, language, translation and national standards, according to local purposes. It is
description. therefore a tool that allow the organization of framework
elements (templates, archetypes and subsets) in groups,
defined from use cases. It is a tool that supports equally the

336
collaborative work of clinical specialists. This is done with [2] T. Beale, “Archetypes: Constraint-based domain models for
the help of resources to create projects that are associated to future-proof information systems ”. In: Eleven OOPSLA,
groups of semantic resources, to whom there can be 2002, New York. 17th annual ACM SIGPLAN. Baclawski,
provided some collaborative work functionalities, such as: K.; Kilov, H. (Eds). New York: ACM, 2002.
access control, versioning, change management and [3] D. Kalra, “Approaches and Research Priorities for Semantic
decision process record. The tool can also be tailored to Interoperability in Support of Clinical Care Delivery”. In:
adjust the collaborative process for production and review SEMANTICHEALTH Project IST 027328. Bruxelas:
of semantic resources. This adjustment is made to pursue Comissão Europeia, 2007.
the specific issues of the adopted standardization process. [4] D. Moner et al, “Archetype-Based Semantic Integration and
Additionally, the tool offers a web portal for publishing and Standardization of Clinical Data”. In: IEEE; EMBS 28.
querying approved semantic resources, by supporting the 2006, New York. Proceedings…New York: IEEE, 2006.
use of these resources by the software development [5] T. Beale, S.Heard, “An Ontology-based Model of Clinical
community, medical professional and clinic information Information”. In: MEDINFO, 2007, Brisbane. Proceedings…
specialists. K. Kuhn et al. (Eds). Brisbane: IOS Press, 2007a.
The terminology server performs selection, creation [6] D. Kalra, Clinical Foundations and Information Architecture
and maintenance of services for set of terms that are for the Implementation of a Federated Health Record
associated to each knowledge model (templates or Service. PhD thesis, University College London, 2002.
archetypes). It operates based on a technical specification [7] L. Michelsen, S. Pedersen, H. Tilma, S. Andersen,
CTS (OMG/HL7). It is considers a versatile and powerful Comparing different approaches to Two-level modeling of
tool approved by the IHTSDO in order to use the SNOMED Electronic Health Records. European Federation for Medical
CT [15] in the framework. This terminology server was Informatics - IOS Press, 2005. Committee for
designed to offer multiple ontologies for final users systems Standardization. Brussels, Belgium. 2003.;
in order to support the data entry and term validation. The [8] M. Eichelberg, T. Aden, J. Riesmeier, A. Dogac, and G. B.
service was created to support the use of terminologies in Laleci, A survey and analysis of electronic healthcare record
the national EHR infrastructure. Subsets of each standards. ACM Comput. Surv., 37:277–315, 2005.
terminology can be quickly defined in an interactive way
[9] S. Garde, P. Knaup, E. Hovena, S. Heard, Towards Semantic
and saved in a central server. The real subsets are stored in
Interoperability for Electronic Health Records Domain
cache to allow quick search and control over the
Knowledge Governance for openEHR Archetypes.
independent updates of new terminology releases. The TS Schattauer GmbH, 2007.
has to be configured as a national service with a waterfall
replication of the subsets definition to auxiliary services [10] V. Stroetmann et al, Semantic Interoperability for Better
nationwide, even if this access is made via a standalone Health and Safer Healthcare: Deployment and Research
computer. This allows delivery easiness, speed and also Roadmap for Europe. [S.l]: European Communities, Jan.
trust. 2009. 34 p. Semantic HEALTH Report.
[11] T. Beale, S. Heard, Archetype definitions and principles.
Rev. 0.5. The OpenEHR Foundation, 2007b.
ACKNOWLEDGEMENTS
[12] K. Dipak, Electronic health record standards. IMIA yearbook
Our thanks to the Ministry of Health of Brazil that
of medical informatics, 2006:136–144, 2006.
funded this project of designing and implementing a SOA
based infrastructure for the Brazilian National EHR solution [13] D.Wollersheim, A. Sari, W. Rahayu, Archetype-based
to be used nation wide. We would also like to thank the electronic health records: a literature review and evaluation
University of Brasília researchers who participated in the of their applicability to health data interoperability and
project and by providing requirements for the solution. access. HIM J, Austrália, v. 38, 2009,
[14] T.Beale, S. Heard, Architecture Overview. Rev. 1.1. The
OpenEHR Foundation, 2007c.
REFERENCES
[15] ISO, SNOMED CT: Clinical Terms Technical Reference
[1] Brasil, Ministério da Saúde. Portaria 2073. 2012. Available
Guide. International Health Terminology Standards
at: http://bvsms.saude.gov.br/bvs/saudelegis/gm/2011/
Development Organisation, July 2016.
prt2073_31_08_2011.html

337