239 Not Just Your Typical Building Blocks

1
239 Not Just Your Typical Building Blocks

• Bo Dagnall, HDR Architect, CDS Team Lead Matt
  Greene, RHIA, CCS

2
HealtheVet Principles

• Distributed, networked collection of
  capabilities, aligned with business needs and
  encapsulated in services
• Consistency and governance of service interfaces
  (-gt VDIM) and service payload/parameters (-gt
  VHIM)
• Data is an Enterprise asset not owned by any one
  business unit application (-gt HDR ADR)

3
Data in Modern Software

• The strategy used to be to define data at the
  database level only and then use loosely at the
  application level as parameters.
• Object-Oriented (OO) development changed the way
  data is used programmatically
• Procedural programming data is used as
  parameters passed through business logic
• OO programming everything is an object.
• Data is materialized as data objects
• Every data structure becomes its own object with
  defined attributes and behavior

4
Data in Modern Software

• Domain models
• OO development necessitates that data be
  stringently modeled so that the data objects
  created accurately reflected the definitions,
  semantics and constraints that meet the business
  needs
• Modern OO applications have a domain model, which
  is a design pattern that creates a package of
  classes/objects that are the physical
  manifestation of the data model
• Domain models describe a system in terms of real
  world business entities such as "Patient",
  "Hospital", "Allergy"
• Domain models provide the benefit of separating
  business logic from data and by providing data
  with an object that can independently control
  behavior and provide state

5
Data in Modern Software

• Enterprise development principles
• Reduce inter-service dependencies and network
  demand by moving from fine-grained data objects
  to coarse-grained data objects
• This requires defined payload structures
• Data is now used objectively and implemented in
  multiple layers
• Physical persistence database level
• Run-time processing the domain model level
• Parameter/payload structure and governance the
  message level
• This approach requires an Enterprise information
  model that provides an enterprise standard for
  data definitions, semantics and constraints that
  can be used in all three layers

6
Interoperability
Courtesy Dr. Charlie Mead

• interoperability
• ability of a system...to use the parts or
  equipment of another system Source
  Merriam-Webster web site
• interoperability
• ability of two or more systems or components to
  exchange information and to use the information
  that has been exchanged. Source IEEE Standard
  Computer Dictionary A Compilation of IEEE
  Standard Computer Glossaries, IEEE, 1990

Semanticinteroperability
Syntacticinteroperability
7
Information is an Enterprise Asset

• Accessibility of quality computable information
  enables an organization not only to function, but
  to manage its processes, rapidly adapt to change,
  and provide superior quality of service to its
  customers
• In the past organizations tried to enforce
  semantic interoperability by forcing all systems
  to use the same database structures, but this
  proved unworkable.
• Today, semantic interoperability is enforced by
  controlling the system interfaces and the payload
  from those interfaces
• Underlying database dont need to be consistent
  to ensure interoperability, although there may be
  other benefits related to database maintenance,
  database splitting/merging and governance that
  still require consistency across databases.

8
VHIM Basics

• The VHIM is the enterprise-wide information model
• The VHIM only models data that is important
  across multiple business applications within the
  enterprise system HealtheVet shareable or
  viewable data
• The VHIM follows the information domain principle
• An information domain is grouping of data that
  represents a logical data division aligned with
  business uses of the data (e.g., lab, pharmacy,
  etc.)
• An information domain is pseudo-independent unit
  of data with inclusive semantics and integrity
  constraints
• A single clinical transaction is typically a
  subset of an information domain and follows the
  semantics and constraints defined within the
  information domain

9
VHIM Basics

• The VHIM uses the Unified Modeling Language to
  model data
• UML class diagrams include constructs such as
  packages (domains), classes (data objects),
  attributes (data elements), data relationships
  (semantics constraints), data types (definition
  constraints), cardinality (required/optional
  constraints) as well as custom properties that
  can include other types of constraints (e.g.,
  formats, valid ranges, permissible values, etc.)
• The VHIM uses class diagrams to provide a static
  view of the data without defining system or data
  behavior

10
VHIM Basics
11
VHIM Basics
Note that the domain model contains all elements
for a subject area. But not all are needed for
each payload type. For instance, the home
Appliance is only used by Home Telehealth
Note the cardinality A vital sign must have one
Hospital Location, and may have one Division
12
VHIM Details

• How can one static data model meet all the
  implementation needs (database, domain model,
  payload structure)?
• Answer Model transformations
• Computationally Independent Model (CIM)
  conceptual model
• Platform Specific Model (PSM) logic model
• Model implementation physical model
• Model Driven Architecture (MDA) and UML software
  provide tools for automating model transformations

13
VHIM Details
VHIM CIM
Conceptual to Logical Model to Model
Transformations
Payload PSM
Database PSM
Java Object PSM
Logical to Physical Model to Implementation
Transformations
Java Application
Database
Payload
14
VHIM Details CIM to PSM Example
The VHIM model is a Computationally Independent
Model, which does not have built-in bias towards
a particular implementation platform Note that
the model contains four subclasses to distinguish
the type of the allergy. But databases dont
understand subclasses. They use type codes to
determine what kind of a thing is in the
record. So, we transform the model into another
model, which is more conducive to databases, from
which we can then generate a database. The same
can be done for XML, Java, etc. The advantage is
that from a single model, other models can be
automatically produced for suit particular
purposes importantly, the derived models all
have the same semantic meaning, thus preserving
the integrity of the data regardless of how or
where its used.
15
VHIM Details PSM to Implementation Example
VHIM Class Diagram
Java
XML Schema Definition (XSD)
ltxselement name"PatientIdentity"
type"VaIdentity" substitutionGroup"personIdentit
y" /gt - ltxscomplexType name"PatientIdentity"gt -
ltxscomplexContentgt - ltxsextension
base"PersonIdentity"gt - ltxssequencegt -
ltxselement name"administrativeGender"
type"AdministrativeGenderCode"gt -
ltxsannotationgt ltxsdocumentationgtA value
representing the gender (sex) of a person. The
for allowable values for this field as specified
by the DS DAT for Demographics are F (Female), M
(Male) and UN (unspecified).lt/xsdocumentationgt
lt/xsannotationgt ...
public interface PatientIdentity extends
personSRDTs.PersonIdentity
livingSubject.AdministrativeGenderCode
getAdministrativeGender() void
setAdministrativeGender(livingSubject.Administrati
veGenderCode administrativeGender)
livingSubject.AdministrativeGenderCode
addNewAdministrativeGender()
livingSubject.BirthTime getDateOfBirth()
void setDateOfBirth(livingSubject.BirthTime
dateOfBirth) livingSubject.BirthTime
addNewDateOfBirth() ...
16
VHIM Details

• Challenge model the future without full
  requirements, define constraints to enhance data
  quality, provide backwards compatibility to
  recent VistA data and really old VistA data
• Decisions and Gold Standards
• Requirements from reengineering teams
• SME reviews of models
• Consistency with current data standards
• HL7 Message Profile Specifications (which is not
  always used to validate message content)
• VistA Data Dictionary (which is not always
  enforced at the VistA data file level)

17
VHIM Details

• Previous approach (versions 3.0-3.4)
• Take best guess at future model
• Get SME input
• Current approach (versions 3.5-3.6)
• Focus on primary customer HDR
• Align model with HDR IMS HL7 Message Profiles
• HDR IMS data will seed HDR II

18
VHIM Details Alignment withHL7 Specifications
19
VHIM Details Alignment withHL7 Specifications
20
VHIM Details Evolution Strategy
HDR Hx
value 136/74 units
HDR IMS / HDR II (today)
value 136/74 units mmHg
21
VHIM Details Evolution Strategy
value 136/74 units
HDR Hx
value 136/74 units mmHg
HDR IMS / HDR II (today)
diastolic value (float) 74 units (code)
code 481298 ? VUID displayText
mmHg systolic value (float) 136 units
(code) code 481298 ? VUID displayText
mmHg
HDR II (after re-eng)
22
VHIM Details

• Role in data quality
• Incorporate data constraints into the model that
  can be used to validate the data as it is passed
  between applications
• Incorporate data constraints that can be
  reflected in the physical models derived from the
  VHIM
• Constraints
• Data types definitional constraints
• Semantics relationship constraints
• Cardinality required/optional constraints
• Permissible Values define that the concept
  strawberry belongs to the value set fruits
  this provides the STS team requirements for what
  value sets are needed for coded data
• Regular expressions can be used to define
  formats, ranges, etc.

23
More HealtheVet Principles

• HealtheVet will use the Service Oriented
  Architecture (SOA) strategy
• Services are used to provide capabilities to meet
  business needs
• Data services meet the business need of providing
  access to data
• Data services abstract the details of data
  location, data persistence, data retrieval, data
  transformations, data validation and data
  packaging

24
HDR Basics

• Data is a common asset that needs a common home
• HDR is the common, shared repository for clinical
  data
• The HDR data set is divided into the HDR IMS (HDR
  II in the future) and HDR Hx
• In the future, the HDR II databases will be
  regionalized (HDR Local) to increase performance,
  reduce scalability demands and partially meet
  failover and availability requirements

25
HDR Basics

• Clinical Data Service (CDS) provides the service
  layer to abstract applications that use clinical
  data from where and how data is persisted,
  retrieved and manipulated.
• CDS will hide the details of interfacing with HDR
  II National, HDR II Local and HDR Hx
• In the future, CDS will also provide data access
  to VistA for clinical data not yet moved into the
  HDR
• CDS will adhere to enterprise requirements
  stating that the message payload coming into and
  out of CDS will be VHIM compliant

26
HDR Basics

• In addition to providing data access to multiple,
  disparate data sources, CDS will also support
  payloads based on multiple VHIM versions

Data Source
VHIM version A payload
CDS
Data Source
VHIM version B payload
VHIM version C payload
Data Source
27
HDR Details

• To support multiple data sources and to apply
  data aggregation and other processing logic, CDS
  needs one internal domain model to execute logic
  against

Data Source
VHIM version A payload
CDS
Data Source
VHIM version B payload
CDS Internal Domain Model
VHIM version C payload
Data Source
28
HDR Details

• To support multiple VHIM version specific
  payloads, CDS needs a VHIM derived domain model
  for each VHIM version supported

Data Source
VHIM version A payload
CDS
VHIM vers. A Domain Model
Data Source
VHIM version B payload
CDS Internal Domain Model
VHIM vers. B Domain Model
VHIM vers. C Domain Model
VHIM version C payload
Data Source
29
HDR Details

• For HDR National Version 2, data models are
  implemented in the following areas
• HDR II National database schema
• HDR Hx database schema
• CDS internal domain model
• CDS VHIM version specific domain models
• VHIM version specific payloads
• MDA is used to auto-generate the VHIM version
  specific domain models and payloads
• CDS VHIM version specific domain models
• VHIM version specific payloads
• The VHIM is used as a blueprint for custom
  creation of the HDR II schema and CDS internal
  domain model
• HDR Hx is based on VistA and has no correlation
  to the VHIM

30
HDR Details
Significant portions of the HDR II National and
CDS National architecture are derived from the
VHIM
CDS National
HDR II National
VHIM version A payload
VHIM vers. A Domain Model
CDS Internal Domain Model
VHIM version B payload
HDR Hx
VHIM vers. B Domain Model
Created directly from the VHIM (MDA)
Created manually using the VHIM as a blueprint
No correlation to the VHIM
31
HDR Details
Significant portions of the HDR II Local and CDS
Local architecture are derived from the VHIM
CDS Local
HDR II Local
VHIM version A payload
VHIM vers. A Domain Model
CDS Internal Domain Model
VHIM version B payload
VHIM vers. B Domain Model
VistA DBs
VistA DB
VistA DB
VistA DB
VistA DB
VistA DB
VistA DB
VistA DB
Created directly from the VHIM (MDA)
Created manually using the VHIM as a blueprint
No correlation to the VHIM
32
Why should you care?

• The VHIM defines data content and structure
• The VHIM defines data quality constraints
• The VHIM is transformed into platform specific
  implementations ensuring that what is in the VHIM
  is reflected in the implementation
• The VHIM governs payload structure enforcing data
  quality constraints at the payload level

33
Why should you care?

• The HDR holds all the clinical content you want
• The HDR / CDS system uses several different data
  models
• Most of these data models are derived from the
  VHIM
• As a result, if data is not modeled in the VHIM,
  that data will not be in the HDR!
• As a result, if payload sent to the HDR fails
  VHIM defined data quality constraint validation,
  it cannot be persisted in the HDR!

34
Some Requests

• Continue to support the HDR
• Appreciate the architectural benefits creating
  the reliance of the HDR on the VHIM
• Embrace and understand the VHIM
• Review and approve the VHIM for
• Content (inclusion/exclusion)
• Content structures and semantics
• Data quality constraints
• Use the VHIM as the vehicle for changing the
  content, content structure and/or data quality
  constraints in the HDR
• Provide guidance on error resolution what
  should the HDR do when it receives data that
  violates the validation constraints?