An Architecture-Centric and Ontology-Based Approach to Cross-Domain Interoperability of Health Information Systems for Diabetes Care

PhD(c). Gustavo Andrés Uribe Gómez
PhD. Diego Mauricio López
PhD. Bernd Blobel

University of Regensburg

Introduction

Interoperability Challenge

Related Work

EHR, PHR and DSS

interoperability

(Booker and Trabulsi, 2009; Fahey, 2012; Osborn et al., 2010; Quinn et al., 2011; Schnipper et al., 2012; Wake and Cunningham, 2013; Santana, 2013)

Related Work

Semantic Interoperability

(Chungoora et al., 2013; Heywood et al., 2011; Sonsilphong and Arch-int, 2013; Tessier, 2011)

Related Work

Ontology-based systems

(Tessier, 2013; Archer et al. 2011; Sonsilphong et al., 2013, Snyder et al. 2013)

Related Work

Architecture-centric approach

(Oemig F and Blobel B., 2011; Health Level 7 International, 2013; Blobel B., Goossen W, Brochhausen M., 2012; Lopez DM and Blobel B, 2009)

Related Work

No single approach supports the following issues:

Consideration of the cross-domain interoperability
Consideration of the system architecture
Adherence to policies and guidelines
Decision support
Knowledge management
- Knowledge formalization
- Inference
Flexible and adaptable
- To add new actors
- To adapt to new rules or knowledge

Research Question

How to achieve cross-domain interoperability in health informatics systems for supporting Type 2 Diabetes Mellitus care?

Hypothesis

By using an architectural-centric approach to analyze, design and implement health information systems based on the Generic Component Model (GCM) and representing the components through ontologies it is possible to achieve cross-domain interoperability of health information systems supporting the diabetes care.

General Objective

Propose an approach to achieve interoperability of health information systems in the diabetes care

Specific Objectives

Define formally and architecturally the system of diabetes care, its components and relationships.
Define use case specific architectures for the relevant use cases in the diabetes care including the related actors.
Develop a pilot software solution to support the relevant diabetes care use cases enabling interoperability.
Evaluate the interoperability functionalities of the software solution developed

Methodology

Generic Component Model - GCM

(B.Blobel, 2012)

Generic Component Model - GCM

(B.Blobel, 2012)

Generic Component Model - GCM

(B.Blobel, 2012)

Generic Component Model - GCM

(B.Blobel, 2012)

RM - ODP

Business Process Modeling

Business Process Modeling Notation (BPMN)

Graphical Notation
Representation of computer independent elements
Standardized

Formal Language for Describing Rules and Policies

SPARQL Inference Notation (SPIN)

Compatible with OWL and RDF
Standardized (Based on RDF)
Provides an integrated development environment

Contributions

A Generic Architecture for an Adaptive, Interoperable and Intelligent Type 2 Diabetes Mellitus Care System

Uribe, G. A., B. Blobel, D. M. López, and S. Schulz. "A generic architecture for an adaptive, interoperable and intelligent type 2 diabetes mellitus care system." Studies in health technology and informatics 211 (2015): 121. Categoria Colciencias A2

Generic Model of the T2DM Care System

GCM Representation of the Medical Domain

14-1

GCM Representation of the Policy Domain

14-2

GCM Representation of the Resource Domain

14-3

UML Representation of Cross-domain Relations

15-1

UML Representation of Medical Domain

15-2

UML Representation of Policy Domain

15-3

UML Representation of Resources Domain

Ontological Representation

Description of Business Process

Description of Processes at the Health Service Level

18-1

Description of Processes at the Disciplines Level

18-2

Description of Processes at the Tasks Level

18-3

Contributions

Specializing Architectures for the Type 2 Diabetes Mellitus Care Use Cases with a Focus on Process Management

Uribe, G. A., B. Blobel, D. M. López, and A. A. Ruiz. "Specializing architectures for the type 2 diabetes mellitus care use cases with a focus on process management." Studies in health technology and informatics 211 (2015): 132. Categoría Colciencias A2.

Main Interventions

Lifestyle intervention

Pharmacological intervention

Medical Domain in GCM

21-1

Policy Domain in GCM

21-2

Resource Domain in GCM

21-3

UML Cross-domain Model

22-1

UML Medical Model

22-2

UML Policy Model

22-3

UML Resource Model

Disciplines in the Glycemic Control

23-1

Observation Health Care Serive Process

23-2

Physical Examination Process

23-3

Hyperglycemia Alert Rule

CONSTRUCT {
    ?id btl2:isPartOf ?patientLife .
    ?id btl2:hasCondition ?id .
    ?id a dm2co:Hyperglycemia .
    ?id a dm2co:MedicalAlert .
    ?id rdfs:label ?cause_type_en .
    ?id rdfs:label ?cause_type_es .
    ?this btl2:represents ?id .
}
WHERE {
    ?patient btl2:isBearerOf ?blood_glucose .
    ?patient btl2:hasLife ?patientLife .
    ?this btl2:represents ?blood_glucose .
    ?blood_glucose a dm2co:BloodGlucoseConcentration .
    ?this dm2co:hasValueIn_mg_dL ?value .
    FILTER ((?value >= 200.0) && (?value < 300.0)) .
    OPTIONAL {
        ?clonAlert a dm2co:MedicalAlert .
        ?this btl2:represents ?clonAlert .
    } .
    FILTER (!bound(?clonAlert)) .
    BIND (STRLANG("hyperglycemia medical alert", "en") AS ?cause_type_en) .
    BIND (STRLANG("alerta médica por hiperglucemia", "es") AS ?cause_type_es) .
    BIND (IRI(fn:concat("http://purl.org/unicauca/dm2co#", STRUUID())) AS ?id) .
}

24-1

Patient Safety Policy

# if physical examination is planned then handwashing is planned before
CONSTRUCT {
    ?this btl2:hasPart _:b0 .
    _:b0 a bpmn:SequenceFlow .
    _:b0 btl2:hasComponentPart _:b1 .
    _:b1 a bpmn:SequenceFlow_Target .
    _:b1 btl2:represents ?physical_examination_plan .
    _:b0 btl2:hasComponentPart _:b2 .
    _:b2 a bpmn:SequenceFlow_Source .
    _:b2 btl2:represents _:b3 .
    _:b3 a dm2co:HandwashingPlan .
}
WHERE {
    ?this btl2:hasPart ?physical_examination_plan .
    ?physical_examination_plan a  dm2co:PhysicalExaminationPlan .
}

24-2

Contributions

Ontology-based and Model-driven Implementation of Health Information Systems. A Diabetes Mellitus Care System Use Case

Uribe, G. A., D. M. López and B. Blobel. "Ontology-based and Model-driven Implementation of Health Information Systems. A Diabetes Mellitus Care System Use Case ". To be submitted.

Uribe, G. A., D. M. Lopez, and B. Blobel. "Towards automated biomedical ontology harmonization." Studies in health technology and informatics 200 (2014): 62-68. Categoría Colciencias A2.

Implementation Methods

MDA

- Hard-coded logics

- Weak semantics

- Inflexible transformations

+ Computer Independent Models

+ Modeling of behavior

25-1

Implementation Methods

Semantic Web

- Difficulties for modeling of behavior

+ Computer Independent Models

+ Inference capabilities

26-2

Implementation Methods

Hybrid

+ Inference capabilities

+ Computer Independent Models

+ Modeling of behaviour

+ Architectural-centric approach

26-3

Implementation Process

Enterprise View

Informational View

How to Achieve Interoperability?

Technical, Structural and Syntactical
Semantic
Organizational/Service

(Blobel [changed])

Computational View

Engineering View

Technology View

Test Scenario

Screenshots

Contributions

Effectiveness of an Ontology-based Decision Support System in the Glycemic Control Use Case

Uribe, G. A., D. M. López, B. Blobel and A. A. Ruiz. "Effectiveness of an Ontology-based Decision Support System in the Glycemic Control Use Case ". In preparation.

Pilot Experiment

Controls the execution of the healthcare process according to policies and national medical guidelines and organizational protocols.
Supports the actors in the decision making process.
Maps the information considering the heterogeneous qualities of the actors.

The software provides interoperability at least in the following three ways:

Pilot Experiment Design

Pilot Experiment - Hypothesis

The efficiency of the system's recommendation, measured through the F-measure, is higher than 0.71 using as gold standard the suggestions provided by an internist.

The threshold of 0.71 corresponds with the F-measure average of the algorithms C4.5 and CART evaluated for the diagnosis of diabetes

D. Senthil Kumar, G. Sathyadevi, and S. Sivanesh, “Decision Support System for Medical Diagnosis Using Data Mining,” International Journal of Computer Science Issues 8, no. 3 (2011): 147–53.

Results

Results - Descriptive Stadistics

Results - F-Measure

Results - One-Sample Test Significance

Results - F-Measure Improved

Results - One-Sample Test Significance

Results Discussion

In order to improve even more the F-measure is possible to add the entities and rules corresponding with the diabetic complications and all the related findings.
Two special diagnoses found by the doctor are “No chronic complications” and “Uncomplicated diverticular disease colon” because correspond with the absence of one medical condition.

Conclusions

Hypothesis

By using an architectural-centric approach to analyze, design and implement health information systems based on the Generic Component Model (GCM) and representing the components through ontologies it is possible to achieve cross-domain interoperability of health information systems supporting the diabetes care.

General Conclusion

A health information system was developed using the GCM demonstrating, in the glycemic control use case, the cross-domain interoperability of the medical, policy and resource domains. Also supports the interoperability by: policies and guidelines, decision support and knowledge mapping.

Other Conclusions

Using the GCM principles enables comprehensive interoperability
The consideration of the top-domain and standardized ontologies facilitates the harmonization between the different domains
The level of generality used in the generic description facilitates the adaptive nature of the system and the components re-usability
The methodology and models used facilitate the inter-disciplinary communication and allows the development of systems taking into account the experts’ knowledge and relevant policies.

Future Work

Improve the system evaluation
Implement the data models mapping
Implement the automatic language transformation
Automate the planner composer and service discovery
Development of a framework for the proposed development process

¡¡Thank you!!

¿¿Questions??