Janos L. Mathe

VOIS

Edited on 2018-08-28

Vanderbilt Oncology Information System (VOIS)

Project Summary

The VOIS program was launched with the goal of creating a robust and integrated clinical information system (CIS) to support the cancer care continuum. The program included informatics methods to support longitudinal cancer treatment plan management, genome-directed cancer treatment prioritization, and team-based approaches to cancer diagnosis and treatment response assessment.

Motivation

Challenges of cancer management

  • Treatment consists of a complex care plans that span from months to years
  • Care plans are managed by multiple members of the clinical care team
  • Safe and effective plan management requires team knowledge of the past, current and future states of the events in the care plan presented at multiple levels of abstraction
  • Existing computerized physician order entry (CPOE) systems are not optimized for management of longitudinal protocolized team-based care

Goals

  • Standardize Care: Increase compliance with standard of care evidence-based guidelines (1200+) and clinical trial protocols
  • Improve Safety and Efficiency: Coordinate actions to reduce risks and costs
  • Provide Agility and Scaling: Enable rapid prototyping and vendor-independent curation of protocols by subject matter experts (SMEs)

Approach

We implemented the VOIS tool suite to enable longitudinal cancer treatment plan management. VOIS adopted a model-driven approach to represent, analyze, instantiate and manage patient chemotherapy plans. As part of the tool chain, domain-specific models are used to capture clinical protocol knowledge that is translated by interpreters to the analysis and execution environments.

The execution environment is composed of a model-driven execution engine, a graphical user interface, and system integration interfaces. These interfaces allowed for the integration with a broad range of CISs at VUMC, including the computerized provider order entry system (CPOE), the pharmacy information system (PIS), the nursing documentation system, the electronic medical record system (EMR), and the outpatient whiteboard.

VOIS utilizes a custom-built, visual, domain-specific language to express the executable treatment protocols and related knowledge, all of which has been curated by SMEs. Such formal approach allowed for a tremendous reduction in the ambiguity of the treatment protocols, and through the promotion of component reuse, conciseness and an extraordinary speed in implementation.

VOIS Architecture

Results

VOIS in numbers

Project

  • Time Frame: September of 2010 – November of 2017
  • Builder FTEs: 2 builders (4-10 FTEs total)

Models

  • Number of Protocols: 539 modeled / 421 used
  • Number of Concepts: 5M instances + 6M relationships

Utility

  • Time Frame: September of 2013 – November of 2017
  • Users: 162 planners / 100 signers

    with gradual rollout to all outpatient chemotherapy prescribing providers

  • Number of Patients: 5828
  • Number of Instances: 9,891 plans / 292,345 med orders

Implementation Efficiency

The model based approach yielded an efficient method for dealing with the non-trivial complexity of the cancer management domain. The example below is a demonstration on how the VOIS models could provide analytical views of the content structure, thus reveal insights on how to optimize the represented content.

Knowledge Driven Views

Furthermore, the VOIS modeling approach proved one of the key benefits of domain-driven design, according to which complexity can be adequately managed through the use of appropriate abstractions.

Cost of Approaches

Contribution

  • Domain analysis, documentation
  • Solution architecture design
  • Domain model development
  • Domain modeling tool configuration and user education
  • Model translator creation for model evolution, analysis and visualization
  • Contributed to UX (user experience) design

Publications

Mathe, J.L.; Levy, M. The ephemeral life of the computer interpretable, guideline-driven Vanderbilt Oncology Information System. Presented at the Mobilizing Computable Biomedical Knowledge (MCBK) Summer Meeting, National Library of Medicine (NLM), Bethesda, MD, p. 1 2018.

Mathe, J.L. Practical Knowledge Engineering for Clinical Decision Support. Presented at the Biomedical Informatics Seminar, Vanderbilt University 2018.

Mathe, J.L. The Precise Construction of Patient Protocols: Modeling, Simulation and Analysis of Computer Interpretable Guidelines. Vanderbilt University, Nashville, TN 2012.

Mathe, J.L.; Sztipanovits, J.; Levy, M.; Jackson, E.K.; Schulte, W. Cancer Treatment Planning: Formal Methods to the Rescue, in: 4rd International Workshop on Software Engineering in Health Care (SEHC 2012). Presented at the SEHC 2012 Workshop @ ICSE, Zurich, Switzerland 2012.

Back Home   

Contact Information

e | janos.l.mathe@vumc.org
p | +1 (615) 875-8778
a | Rm 515B, Ste 500, Bld 3401WE