detailed clinical models and their relation with electronic health records

26 Enero 2016DETAILED CLINICAL MODELS AND THEIR RELATIONWITH ELECTRONIC HEALTH RECORDSDiego Boscá Tomás

PRESENTED BY: DIEGO BOSCÁ TOMÁSSUPERVISED BY: PhD. MONTSERRAT ROBLES VIEJO and PhD. ALBERTO MALDONADO SEGURA

DETAILED CLINICAL MODELS AND THEIR RELATION WITH ELECTRONIC HEALTH RECORDS

26 Enero 2016DETAILED CLINICAL MODELS AND THEIR RELATIONWITH ELECTRONIC HEALTH RECORDSDiego Boscá Tomás 2

CONTENTS

• Introduction and Objectives• Outline• Research work

I. Archetype representation of non-dual model architectures

II. Archetype-based mapping between clinical information models

III. Automatic generation of implementation guides from clinical information models

• Conclusions, Contributions, and Future Work• Projects, Publications, and Research Stays


INTRODUCTION AND GOALS


INTRODUCTION

• Accessing the full Electronic Health Record (EHR) of a patient is still difficult as this information tends to be distributed into different systems with different semantics, and needs semantic interoperability for their meaningful communication

• Semantic interoperability is still far from being achieved in the clinical domain, mostly due to the complexity, variability, and evolving knowledge


INTRODUCTION

• One of the prerequisites to achieve semantic interoperability is the standardization of both the data and clinical information models present in the information system (SemanticHealth)

• Standards need to be brought to live. There is a lack of adequate methodologies and tools for they widespread use


INTRODUCTION

HYPOTHESISArchetypes can be applied to describe the structure, content, and meaning of existing EHRs as well as to facilitate the development and deployment of new EHR systems that require semantic interoperability


OBJECTIVE

GENERAL OBJECTIVEProvide methodologies and advanced tools based on archetypes to ease the achievement of higher levels of EHR semantic interoperability


OBJECTIVE

SPECIFIC OBJECTIVESTo provide means of applying dual model methodology to non-dual information models To provide tools and methodologies for the transformation of existing data into archetype data instancesTo generate reference materials automatically from archetypes to achieve suitable views for each type of user

O1.

O2.

O3.


OUTLINE

R I

R III

R II


ARCHETYPE REPRESENTATION OF NON-DUAL MODEL ARCHITECTURES

10


MOTIVATION

• Current EHR systems (typically based on a single model EHR architecture) tend to have clinical knowledge mixed into the information model

• Dual model methodology allows the formal description of clinical models of a given EHR information model

• Even single model EHR architectures may be suitable to be used as a reference model for archetype definition

• Dual model methodology provides a set of advantages, such as formalization, reuse, terminology bindings, or multilinguality


OBJECTIVE

Create a set of methodologies and tools to define archetypes for non-dual information models, ease their definition, and support their advanced use

O1.


REFERENCE MODEL ARCHETYPES

• Only a subset of the classes contained in reference models define logical building blocks of EHRs and can be used to define archetypes (business class)

• The representation of a business class as an archetype is what we call a Reference Model Archetype (RMA) †

• Any other archetype must be a specialization of one of them– Archetype editing becomes a process of subtyping by

constraints

• This approach opens the door to support multiple reference models

† Maldonado et al, LinkEHR-Ed: a multi-reference model archetype editor based on formal semantics. IJMI 2009


EXAMPLE: ISO13606 BUSINESS CLASSES


BUILDING REFERENCE MODEL ARCHETYPES

• In order to define archetypes in a given reference model, we must build the set of RMAs first– We achieve this by analyzing the formal schema

definitions (typically XML Schema)– This is not always possible, as the schema may not

be public, is defined in an unsupported format, or is underspecified (e.g. XML Schema any type)

• Two alternatives: Creating them from other metamodel representations or defining them manually


CREATING RMA FROM METAMODELS

• For this process, a suitable metamodel that faithfully represents a given standard is chosen

• The metamodel, e.g. ecore definitions or Basic Meta Model (BMM) definitions, is analyzed


CREATING RMA FROM METAMODELS

Base of

FHIR Ecore model

Transformed into

FHIR archetypes / extended resources

FHIR Reference Model Archetype

Defined in

FHIR Resource

Base of

openEHR BMM

Transformed into

openEHR archetypes

Defined in

openEHR RM openEHR RMAs

• This approach was used to support archetype creation for FHIR DSTU and BMM-based models such as openEHR and CIMI


CREATING RMA MANUALLY

• As RMAs are archetypes it is feasible to build them by defining the object, attribute, and primitive constraints

• An Archetype Object Model (AOM) editor was created• Japanese standard MedXML MML RMAs were built

using this approach


ADVANCED ARCHETYPE EDITING

• Target users of archetype editors come from the clinical domain. Archetype editors must hide the complexity of the underlying reference model– The easiest solutions is to hard-code the reference model into the editors but…– It is harder to cope with multiple reference models and their evolution.

• In order to support multiple reference models and at the same time keeping the editing process simple for clinical users, different approaches and methodologies have been investigated, namely:

– Use of specific editors– Semantic patterns– Archetype creation from data instances– Syntactic model transformation between standards– Mapping to other standards


USE OF SPECIFIC EDITORS

• Specific editors for each reference model can be defined to hide their complexity when creating archetypes.

• These editors are based on a documentation file and can be dynamically included in the tool.

• Specific editors for ISO13606, openEHR, and HL7 CDA have been defined.


USE OF SPECIFIC EDITORS


SEMANTIC PATTERNS

• Archetype reuse is not limited to archetype slots and internal references• Semantic patterns are reusable solutions to recurring modeling problems

– They combine structural, terminological, and ontological representations• We implemented the semantic patterns as archetype fragments with known

semantics• These semantic patterns are useful for creating equivalences between standards

– E.g. How an openEHR Observation is represented in ISO13606 or HL7 CDA reference model

• Examples of semantic patterns are– General use structures such as table, tree, or panels– Generic clinical models such as exam of XYZ– Representation of archetypable classes of different reference models such as observation,

event, or history in ISO13606• In addition to the support for using semantic patterns in archetype edition, a

semantic pattern manager was implemented to allow their versioning and management


ARCHETYPE CREATION FROM DATA INSTANCES

• One of the biggest challenges when applying archetypes to a non-dual reference model is that defining a minimum set of archetypes is required in order to take advantage of archetype-based methodologies

• As the systems are already deployed, it is feasible to obtain sample data instances

• A methodology for the creation of archetypes from data instances was developed

• This process considers each data instance as a “constant archetype” (an archetype with every constraint fixed to a constant) and reduces the problem to merging two archetypes into a single one


ARCHETYPE CREATION FROM DATA INSTANCES

• The automatic merging process traverses each instance and loosens the resulting archetype constraints according to the new values

Instance 1 Instance 2 Resulting archetype

|102..115|

|60..70|


SYNTATIC MODEL TRANSFORMATION BETWEEN STANDARDS

• Transformation of clinical models between different standards is a difficult problem– Ontology-based model and data transformations is one of the most

promising approaches• However, most of the time equivalences are defined in the standards

themselves or are clear enough so complex ontology reasoning is not needed

• These equivalences can be declared as rules to be executed• Three different kinds of rules were detected depending on the trigger:

– Structure-based rules: triggered by the structure and values– Terminology-based rules: triggered by the terminological binding– Generic rules: transformation to generic classes

• Defined rules transform the archetypes and create the mapping equivalences for the automatic data transformation


SYNTATIC MODEL TRANSFORMATION BETWEEN STANDARDS

• This methodology is demonstrated by a proposed transformation between openEHR and ISO13606 (and vice versa). Rules were implemented in drools– All archetypes from openEHR CKM (415) were translated to ISO13606 and the

Spanish Ministry of Health archetypes (23) were translated to openEHR and tested as valid for their respective reference model

Target class in openEHR Source term binding contained in SNOMED CT subsetObservation << 363787002 | Observable entity (observable entity) |

OR << 284365007 | Examination of body site (procedure) | OR << 122869004 | Measurement procedure (procedure) |

Evaluation << 243814003 | Interpretation of findings (observable entity) |Instruction << 243120004 | Regimes and therapies (regime/therapy) |

OR << 400999005 | Procedure requested (situation) |Action << 129264002 | Action (qualifier value) |

OR << 416118004 | Administration (procedure) | OR << 443938003 | Procedure carried out on subject (situation) | OR << 71388002 | Procedure (procedure) |

Admin_entry << 14734007 | Administrative procedure (procedure) | OR << 304784009 | Administrative form (record artifact) |

Event << 272379006 | Event (event) |


ARCHETYPE-BASED MAPPING BETWEEN CLINICAL INFORMATION MODELS

27


MOTIVATION

• Patient health data is distributed among different healthcare facilities, which makes crucial their integration and communication

• Health data communication must be done in a meaningful way, avoiding all possibility of misunderstanding or misinterpretation

• This crucially depends on the standardization of the EHR architecture used to represent the data

• Standardizing existing data is one of the main problems in adopting EHR standards


OBJECTIVE

To define and implement a set of methodologies and tools for the normalization of legacy data and the transformation between EHR standards

O2.


PROBLEM

• The problem of data transformation is a difficult one, as deals with differences and mismatches between heterogeneous data formats and models

• This is even harder in healthcare domain, due to the intrinsic complexity of EHR data

• Standard EHR architectures and archetypes are defined without any consideration to the internal database design

• The definition of these transformation programs is a complex process, usually involving the creation of custom non-reusable software


MAPPING APPROACH

• Existing data transformation formalisms have been studied and adapted to cope with the special requirements of the archetypes

• We use a “specify-generate” approach– High-level declarative assertions are used to

specify relationships between the source and target schemas

– These assertions are compiled automatically into executable scripts


FORMAL DEFINITION

• Source schemas can be either XML Schema or an archetype

• Target schema is an archetype• We use the Nested Relational (NR) model for the

specification of both source and target schemas, as archetypes impose hierarchical structures

• NR model generalizes relational model (tuples and relations) by modeling them as records and sets of records respectively

• This model allows the representation of hierarchical structures as the archetype


FORMAL DEFINITION

• The set of atomic data types of our model are the ones supported by archetypes:– string, integer, real, date, time, date and time, duration, and boolean

• Non-atomic types are – Record types of the form – Set types of the form – Choice types of the form

• where:– represents either an atomic, set, or record type

– li is a natural number, ui is a natural number or ∞ and li ui

– and


FORMAL DEFINITION

•


MAPPING DEFINITION

• Only constrained entities (classes and attributes) of the reference model appear in the archetypes– The underlying reference model constraints are

implicit– This is also the case of the parent archetype in

specialized archetypes• In order to generate correct data instances,

archetypes must be merged with the parents archetypes and the underlying reference model


MAPPING DEFINITION


MAPPING DEFINITION

• Mapping language is based on tgds † (tuple-generating-dependencies) • To describe how to compute a value for an atomic attribute of the

target schema (archetype) a list of filter/function pairs can be defined

$gender={ If (/patient/gender=’M’ OR /patient/gender=’m’) then 0Else if (/patient/gender=’W’ OR /patient/gender=’w’) then 1Else if (/patient/gender=0 OR /patient/gender=1) then /patient/genderElse 9 }

† Fagin et al, Data exchange: semantics and query answering. TCS 2005


• Value correspondences lack expressive power to completely describe grouping semantics

• Default grouping semantics is based on the Partition Normal Form (PNF)– In PNF, two distinct records that coincide in all the

atomic (non-multivaluated) attributes cannot exist• We use Skolem functions † to achieve PNF

GROUPING

† Hazewinkel et al, Encyclopædia of Mathematics


GROUPING

• Grouping semantics can be customized by modifying the Skolem functions or by defining Object Builders

• Skolem parameters can be removed from the Skolem functions

• Object builders control the generation of target instances taking into account the data in the source– Object Builders represent iterators on the source nodes

they are connected to– In each iteration, a new element is constructed, of the

kind of the target node reached by the builder


MAPPING DEFINITION EXAMPLE

Object builder

Attribute mappingObject builder nesting

$prof/cargo=“cirujano”$prof

$ingr

$hosp


XQUERY GENERATION

• Archetype constants are assumed to be transformation functions whose filter has ‘true’ value and the function is fixed to the constant value

• Constraints on primitive types (valid ranges, lists of values) are enforced in the transformation script

• When the mapping correspondences have been defined, they are traversed in order to generate the transformation script

Archetype constraint Value correspondence XQuery extractnumber matches {|0..10|} If true

then /data/measurements/value(specified by user)

For $val in /data/measurements/valueWhere $val≥0 and $val≤10Return<number> {data($val)}</number>

Archetype constraint Value correspondence XQuery extractnumber matches {1} If true then 1

(automatically generated)Return <number> 1 </number>


XQUERY GENERATION

• More than one mapping can be applied to a single attribute. In this case, each one of these mappings produces a different transformation program

• For each object in the archetype a nested XQuery FLWOR is produced


XQUERY GENERATION


MAPPING SUMMARY

Archetype

ReferenceModel

XQuery scriptSource

XML instance

OutputXML

instance

Mapping source (XML Schema or

archetype)MappingComprehensive

ArchetypeMerge

Defines

Defines

Autogenerates Instance of

Generates

Instance of

Instance of


TECHNICAL EVALUATION

• For the technical validation of the mapping capabilities, the STBenchmark was used– This benchmark describes 17 mapping scenarios

that should be supported by mapping systems• Each scenario contains a source and target

schema expressed as an XML Schema, an instance of the source schema and a visual and textual description of the scenario– In our case, target schema is modeled as an

archetype


TECHNICAL EVALUATION

• From the 17 scenarios, 15 could be successfully tested as the expected target instance could be generated

• Two scenarios could not be tested– One scenario could not be tested due to

archetypes not supporting the definition foreign keys

– One scenario could not be tested due to limitations in the expressive power of our mapping language


EVALUATION IN REAL SETTINGS:MEDICINES RECONCILIATION PROJECT

• The objective was to obtain and evaluate a complete medication list of patients regardless where they came from

• Archetypes are based on epSOS specifications

• Hospital EHR system was upgraded to include the patient summary

• The project was awarded with the Spanish Ministry of Health quality award

LinkEHR Platform

DATA NORMALIZATION

DATA INTEGRATION

NOMENCLATORDIGITALIS

INDEPENDENTWEB VIEWER

HOSPITAL EHR (SELENE)

Web Service

Primary care

OMI-AP

Hospital de Fuenlabrada

FARMATOOLS SELENE


EVALUATION IN REAL SETTINGS:PATIENT ELIGIBILITY

• Use of archetypes and mappings in different abstraction levels to determine if a patient is eligible for a cancer clinical trial

• Patient is not eligible if it has severe comorbidity, which is derived from a series of cascade mappings


EVALUATION IN REAL SETTINGSPATIENT ELIGIBILITY

• Presence/absence of metastatic tumor (source EHR)

• Presence/absence of metastatic solid tumor (source archetype)

Condition Mapping@count(summary/problems/problem, @in(summary/problems/problem/code, @descendents(“128462008”))>0 TRUETRUE FALSE

Condition Mapping(Evaluation_problem_DS_metastatic_tumor_v1/structure/present = TRUE) AND (Evaluation_problem_DS_solid_tumor_v1/structure/present = TRUE)

TRUE

TRUE FALSE


ARCHETYPES FOR THE GENERATION, VALIDATION, AND USE OF EHR SYSTEMS

50


MOTIVATION

• Archetypes ease the communication between clinicians and IT staff

• Those clinicians and IT staff who are not experts in dual model architectures need to understand the archetypes– Clinicians are able to provide valuable inputs for archetype

validation and evolution– IT staff can use the archetypes as formal specifications of

the system• Clinicians and IT staff make use of different sets of

artifacts to understand the archetypes and bring them to life in their systems


OBJECTIVE

To enable the automatic generation of artifacts from clinical archetypes for the validation of current systems and the creation of new ones from both clinical and technical perspectives

O3.


GENERATION OF ARTIFACTS

• Archetypes can be used as guidance to automatically generate a wide range of artifacts– Validation rules– Sample data instances– Mindmaps and Sample forms– Implementation guides


VALIDATION: NATURAL RULE LANGUAGE

• Natural Rule Language (NRL) provides a formal and executable way of defining English-like rules

• Archetype constraints are interpreted and translated to NRL

• These rules can be translated to OCL or Schematron


VALIDATION: HL7 STYLE RULES

• In the same way, rules from archetypes can be translated into HL7 conformance statements

• These statements can be generated in the desired language

VALIDATION: SCHEMATRON RULES

• Different kinds of Schematron rules are derived for each kind of archetype constraint

• The definition of two different Schematron phases allows to validate both the explicit and the implicit archetype constraints (only archetype or archetype + reference model)

<rule context="ENTRY/items/parts[archetype_id='at0001']/value">

<report test="count(units)=0">Attribute 'units' is required and does not appear in data</report><assert test="(count(units)=1)">The attribute 'units' has more than one alternative or empty alternatives</assert><report test="count(value)=0">Attribute 'value' is required and does not appear in data</report><assert test="value>=0.0 and value<1000.0">Constraint for 'value' does not follow the archetype: 'value' <value-of select="value" /> is not in [0.0,1000.0[</assert>

</rule>


SAMPLE DATA INSTANCES

• Sample data instances allow to perform stress tests on systems• Data instances are created by using the mapping mechanism explained before

– Constant values are assigned to each archetype constraint– These constant values are used to generate a transformation script to produce the

sample data instance– Generated instances follow the archetype and the reference model, but do not

describe real clinical cases


ADDITIONAL ARTIFACTS

• A mindmap representation from the archetype is created in order to give an overview of the clinical model

• Archetype constraints can also be interpreted as fields of an sample html form

• Archetype constraints bound to terminologies can be displayed in summary value set tables


IMPLEMENTATION GUIDE GENERATION

• Presented artifacts can be combined according to user’s needs to provide helpful implementation guides

• Each object in the archetype generates a set of NRL, HL7, and Schematron rules, as well as a sample data instance for that object

• A general mindmap, archetype metadata, value set tables, sample form, reference model documentation, and a table of contents can also be generated


CONCLUSIONS, CONTRIBUTIONS, AND FUTURE WORK


CONCLUSIONS

• The specifications of eHealth interoperability standards must be brought to life by system designers and implementers

• The challenge is not the lack of standards, but its implementation at a reasonable cost

• Presented methodologies and tools ease the joint use of the three layers of artefacts (Clinical Models, Reference Models, and Clinical Terminologies) needed to achieve high levels of semantic interoperability


CONCLUSIONS

• The advantages of a joint use of archetypes with non-dual model standards outweigh the disadvantages– Using archetypes for currently used reference models such as HL7

CDA, HL7 FHIR, or MedXML MML will help to the rapid adoption of both the original standard and the dual model approach

• The presented set of integrated innovative tools help current systems to achieve semantic interoperability by normalizing data based on clear semantically-rich clinical models

• Proposed methodology promotes the involvement of clinical staff in the modeling and validation process

• All presented developments have been included in LinkEHR platform


CONTRIBUTIONS

• Design and implementation of a set of methodologies and tools to support the definition of archetypes from multiple reference models

• Design and implementation of a new methodology for the transformation of legacy data into archetype-based data

• Reuse of this methodology for the data transformation between archetype-based standards

• Design and implementation of a set of methodologies for the automatic generation of implementation guides and reference materials from archetypes


FUTURE WORK

• From data perspective, variability in original sources continuously provides new challenges– Support to other source schemas such as RDF or OWL– Automatic schema matching techniques

• Based both in clinical terminologies and tree-matching algorithms

• Use of clinical vocabularies and ontologies provides challenges on their own. The integration of ontologies in queries and reasoning will benefit both patient and public health– Mapping specification must be improved to support this advanced use

• Automatic artifact creation can benefit from terminology bindings• Having available big quantities of normalized archetype-based data

provides the opportunity to generate semantically rich research and public health repositories


PROJECTS, PUBLICATIONS, AND RESEARCH STAYS


RESEARCH PROJECTS

• The results of this PhD Thesis have contributed directly to the results of the following research projects:– ResearchEHR: Platform for the acquisition and sharing

of information and knowledge for network-based clinical research communities (TSI2007-66575-C02-01)

– Trial-Me: Intelligent tools for linking EHR to clinical trial systems (TIN2010-21388-C02-01)

– Clin-IK-Links: Clinical information and knowledge models for linking EHR and Clinical Decision Support Systems (TIN2014-53749-C2-1-R)


JOURNAL CONTRIBUTIONS AND BOOK CHAPTERS

• J. A. Maldonado, D. Moner, D. Boscá, J. T. Fernández-Breis, C. Angulo, and M. Robles, “LinkEHR-Ed: A multi-reference model archetype editor based on formal semantics,” International journal of medical informatics, vol. 78, no. 8, pp. 559–570, 2009

• M. Marcos, J. A. Maldonado, B. Martínez-Salvador, D. Boscá, and M. Robles, “Interoperability of clinical decision-support systems and electronic health records using archetypes: a case study in clinical trial eligibility,” Journal of biomedical informatics, vol. 46, no. 4, pp. 676–689, 2013

• J. A. Maldonado, C. M. Costa, D. Moner, M. Menárguez-Tortosa, D. Boscá, J. A. M. Giménez, J. T. Fernández-Breis, and M. Robles, “Using the ResearchEHR platform to facilitate the practical application of the EHR standards,” Journal of biomedical informatics, vol. 45, no. 4, pp. 746–762, 2012

• D. Boscá, J. A. Maldonado, D. Moner, and M. Robles, “Automatic generation of computable implementation guides from clinical information models,” Journal of biomedical informatics, vol. 55, pp. 143–152, Jun. 2015

Publications in Q1 Journals

RI

RII

RIII


JOURNAL CONTRIBUTIONS AND BOOK CHAPTERS

• S. Kobayashi, D. Boscá, N. Kume, and H. Yoshihara, “Reforming MML (Medical Markup Language) Standard with Archetype Technology,” Indian Association for Medical Informatics, vol. 91, p. 57, 2014.

• J. A. Maldonado, D. Boscá, D. Moner, and M. Robles, “LinkEHR: A Platform for the Normalization of Legacy Clinical Data Based on Archetypes,” Interoperability in Healthcare Information Systems: Standards, Management, and Technology: Standards, Management, and Technology, p. 45, 2013.

• M. Marcos, J. A. Maldonado, B. Martínez-Salvador, D. Moner, D. Boscá, and M. Robles, “An archetype-based solution for the interoperability of computerised guidelines and electronic health records,” in Artificial Intelligence in Medicine, Springer Berlin Heidelberg, 2011, pp. 276–285.

RI

RII


CONGRESS PAPERS

• CONGRESS PAPERS RELATED WITH THE THESIS

– D. Boscá, D. Moner, J. A. Maldonado, and M. Robles, “Combining Archetypes with Fast Health Interoperability Resources in Future-proof Health Information Systems.,” Studies in health technology and informatics, vol. 210, pp. 180–184, 2014.

– D. Boscá, L. Marco, D. Moner, J. A. Maldonado, L. Insa, and M. Robles, “Detailed Clinical Models Governance System in a Regional EHR Project,” in XIII Mediterranean Conference on Medical and Biological Engineering and Computing 2013, Springer, 2014, pp. 1266–1269.

– D. Moner, J. A. Maldonado, D. Boscá, A. Mañas, and M. Robles, “Development of a Visual Editor for the Definition of HL7 CDA Archetypes,” in XIII Mediterranean Conference on Medical and Biological Engineering and Computing 2013, Springer, 2014, pp. 1258–1261.


CONGRESS PAPERS


– D. Boscá, L. Marco, V. Burriel, T. Jaijo, J. M. Millán, A. M. Levin, O. Pastor, M. Robles, and J. A. Maldonado, “Genetic testing information standardization in HL7 CDA and ISO13606.,” in MedInfo, 2013, pp. 338–342.

– C. Martínez-Costa, D. Boscá, M. C. Legaz-García, C. Tao, B. J. Fernández, S. Schulz, and C. G. Chute, “Isosemantic rendering of clinical information using formal ontologies and RDF.,” Studies in health technology and informatics, vol. 192, pp. 1085–1085, 2012.

– D. Boscá, J. A. Maldonado, D. Moner, and M. Robles, “Detailed clinical models to facilitate interstandard interoperability of data types,” 23rd International Conference of the European Federation for Medical Informatics. Oslo, Norway: European Federation for Medical Informatics, 2011


CONGRESS PAPERS


– J. Maldonado, D. Moner, D. Boscá, C. Angulo, L. Marco, E. Reig, M. Robles, and others, “Concept-based exchange of healthcare information: The LinkEHR approach,” in Healthcare Informatics, Imaging and Systems Biology (HISB), 2011 First IEEE International Conference on, 2011, pp. 150–157.

– D. Moner, J. A. Maldonado, D. Boscá, C. Angulo, M. Robles, D. Pérez, and P. Serrano, “CEN EN13606 normalisation framework implementation experiences,” Seamless Care, Safe Care: The Challenges of Interoperability and Patient Safety in Health Care: Proceedings of the EFMI Special Topic Conference, June 2-4, 2010, Reykjavik, Iceland, vol. 155, p. 136, 2010.

– M. Robles, J. T. Fernández-Breis, J. A. Maldonado, D. Moner, C. Martínez-Costa, D. Boscá, and M. Menárguez-Tortosa, “ResearchEHR: Use of semantic web technologies and archetypes for the description of EHRs,” Studies in health technology and informatics, vol. 155, p. 129, 2010


RESEARCH STAY

• RESEARCH STAY– Main Subject: Detailed clinical model representation of MML Japanese

standard– Place: Division of Medical Information Technology and Administration

Planning, Kyoto University (Japan)– Length: 3 Months (2011)


Thank for your attention