CARE MAPS: DISTRIBUTED SEMANTIC HEALTHCARE WORKFLOWS
Daniel Aronne ~ George Mason University ~ 2011
Objective
How to provide patients with a seamless healthcare treatment environment?
A holistic approach.
Healthcare Industry
A highly dynamic industry in a very complex environment.
Multiple stakeholders spanning multiple organizations and geographical locations: Laboratories Physicians Clinics Pharmacies Insurance Companies Federal Agencies The Patient Himself!
PatientsProviders
Employer Employer
Payer Payer
Pharmaceutical Manufacturers
Medical Devices
Integrated
Networks
Hospitals
LTCFacilities
Outpatient
Other
Biotech
Distributor/Wholesaler
RegulatoryAgency
Physicians
Payers/ Regulators
Healthcare Industry Stakeholders
AHRQ 2007 Annual Conference Presentation: http://www.ahrq.gov/about/annualmtg07/0927slides/juhn/Juhn-contents.html
Current Challenges and Issues
Interaction Intra-organizational and Inter-organizational Orchestration vs. choreography
Integration Service discovery and matching
Adaptability Adaptive medical workflows
Quality of Service (QoS) Reliability, availability, scalability, error
handling
Current Challenges and Issues
Localization Local jurisdiction requirements
Usability How much user interaction?
Behavioral A human driven industry How much should we automate without
human intervention?
Care Maps
A roadmap of a patient’s journey. Consists of a series of steps and
decisions points in the management of a condition.
Is usually based on medical guidelines, recent evidence and expert consensus.
Patient centered.
Agent-based system approach
Multi-Agent Systems
Have been recognized as a technology to efficiently build complex systems.
Suitable for describing the coordinating and negotiating nature of healthcare service providers and consumers.
Multi-Agent Systems
Previous works have demonstrated the added values of agent-based systems in healthcare, and specifically in healthcare workflows [9]: Reusability Reliability Flexibility Robustness Maintainability Adaptability
Multi-Agent Systems
Added values (continued): Support the integration of legacy systems Tackle the shortcomings of centralized
systems such as: performance bottlenecks resource limitations other kinds of failures
Multi-Agent Systems
Shortcomings: Most of the systems are only prototypes. Most are not widely deployed in real
environments. Further study is required.
Decentralized workflow execution
Decentralized Workflow Execution
Supports the dynamic nature of the healthcare industry
Ad-hoc adaptation to changing conditions at runtime.
Run-time process fragmentation and process migration.
Decentralized Workflow Execution
• Process fragmentation vs process migration.
Zaplata, Sonja, Kristof Hamann, Kristian Kottke, and Winfried Lamersdorf. "Flexible Execution of Distributed Business Processes Based on Process Instance Migration." Journal of Systems Integration 1.3 (2010): 3-16.
Decentralized Workflow Execution
Enhance existing processes with non-intrusive migration data.
Non-modifying annotation of process descriptions: migration meta-model.
All potential participants have to provide a compliant interface in order to receive process descriptions from preceding process engines (e.g. XPDL, WS-BPEL)
Support encryption and decryption of process fragments and/or migration data for security and privacy purposes.
Distributed directory service (DDS)
Distributed Directory Service Inspiration from:
Domain Name System (DNS) Namespace hierarchy Authoritative servers Replication
P2P protocols (i.e. Bit Torrent) Query routing Network overlays
No single point of failure, better reliability Scalable
Semantic Matchmaking
Match service providers and service consumers.
Compute syntactical and semantic similarity among service capability descriptions.
Requires use of a semantic model (e.g. ontology) to describe service descriptions.
A distributed semantic workflow management, multi-agent system approach
Proposed framework
Decentralized Directory Service (DDS):
• Resource and service discovery.
• Provides support for:• Semantic
querying.• Federated query.• Security.
• Solves JADE centralized DF.
Healthcare Entity Agent (HEA):• Storefront representative of any healthcare service provider.• Initiates execution of process instances.• Allocates process fragments to other healthcare entities .• Executes process fragments.• Can migrate process instances to other entities.• Interacts with any BPM engine that supports a standardized workflow definition format (i.e. XPDL).
Broker Agent (BA):• Semantic matchmaker: Matches service requests with service providers.• Queries local Directory Service Ontology which contains semantic service descriptions.• If no suitable match, requests the Directory Service Agent to route his query to other Directory Services.
Directory Service Agent (DSA) • Storefront for the Directory Service.• Handles new Service Providers registration.• Propagates newly registered SPs to other Directory Service nodes.• Routes queries to other BA in the Distributed Directory Service.
User Agent (UA) • Acts on behalf of a human person• May be delegated atomic tasks. • May reside in a desktop computer or in a mobile device.
Future Work
A prototype to test these concepts. Address security and privacy concerns.
References
[1] D. Alexandrou and G. Mentzas. “Research Challenges for Achieving Healthcare Business Process Interoperability”, in Proceedings of the 2009 International Conference on eHealth, Telemedicine, and Social Medicine, ETELEMED '09, IEEE Computer Society.
[2] J. Emanuele and L. Koetter, "Workflow Opportunities and Challenges in Healthcare", in 2007 BPM & Workflow Handbook, 2007.
[3] Song, X., Hwong, B., Matos, G., Rudorfer, A., Nelson, C., Han, M., Girenkov, A., “Understanding Requirements for Computer-aided Healthcare Workflows: Experience and Challenges”, in Proceeding of the 28th international conference on Software engineering, ICSE’06, ACM Press.
[4] Wei Tan, Yushun Fan, "Decentralized Workflow Execution for Virtual Enterprises in Grid Environment," Grid and Cooperative Computing Workshops, International Conference on, pp. 308-314, Fifth International Conference on Grid and Cooperative Computing Workshops, 2006
[5] J. Dang, A. Hedayati, K. Hampel, and C. Toklu. “An ontological knowledge framework for adaptive medical workflow”. Journal of Biomedical Informatics, 41(5):829–836, October 2008.
[6] Z. Maraikar. “Resource and service discovery for mobile agent platforms”. Master’s thesis, Department of Computer Science, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands, August 2006
[7] Zaplata, Sonja, Kristof Hamann, Kristian Kottke, and Winfried Lamersdorf. "Flexible Execution of Distributed Business Processes Based on Process Instance Migration." Journal of Systems Integration 1.3 (2010): 3-16. Print.
[8] Huser, Vojtech, Luke Rasmussen, and Justin Starren. "Representing Clinical Processes in XML Process Definition Language (XPDL)." Web.
[9] Isern, David, David Sanchez, and Antonio Moreno. "Agents Applied in Health Care: A Review." International Journal of Medical Informatics 79.3 (2010): 145-66.