Semantic Representations of Clinical Care Data Leveraging HL7 FHIR

SWAT4LS
4 Dec, 2016

http://tinyurl.com/swat4lsfhir

http://www.w3.org/2016/Talks/1205-swat4ls-egp/validation/

simple validator: https://rawgit.com/shexSpec/shex.js/master/doc/shex-simple.html
pirate pad for collaboration: http://piratepad.net/fhirshex
coded example data: http://www.w3.org/2016/Talks/1205-swat4ls-egp/example/subjects
simple example data: http://www.w3.org/2016/Talks/1205-swat4ls-egp/simple/subjects

Harold Solbrig, <@@>
- Mayo Clinic
Eric Prud'hommeaux, <eric@w3.org>
- Sanitation Engineer, W3C

FHIR Resource Schemas

Network of Schemas

simultaneous display of diagnosis-related resources

Data Integrity

Before we use data we have to trust it.

Enables consistent processing.
Reduces defensive coding.
Increases confidence in input.

Folks responsible for clinical data are especially conservative/sensitive/paranoid.

Schema Languages

How do we enforce data integrity?

	SQL	XML
missing properties	reportedBy UNSIGNED INT NOT NULL	element reportedBy { User },
missing/bad type arcs	N/A	N/A
missing referents	FOREIGN KEY (reportedBy) REFERENCES Users(ID)	<keyref refer="UserID">
inconsistent state	CHECK(assignedOn>reportedOn)	[schematron]
value set violations	ENUM('unasigned', 'assigned')	attribute state { "unassigned" \| "assigned" }

RDF Data Integrity

format is valid - every parser will interpret the same model.
- reject invalid input - parser
- all valid input has defined model - format specification
structure is valid - that model is an agreed-upon model for that information - schema validator
- relationship names
- graph shape/structure/topology
- expected datatypes and conformant values
- value sets
- entity type and labels, e.g. IRI matching http://...

Validation in RDF

Useful data needs consistent structure:

Advertise value
- What's there?
- How's it spelled?
- How's it linked?
Generate interfaces
Direct query generation
Direct query optimization

Detect and correct errors:

missing properties,
missing/bad type arcs,
missing referents,
inconsistent state (e.g. dates),
value set violations.

@prefix : <http://www.w3.org/2012/12/rdf-val/SOTA-ex#> .
@prefix foaf: <http://xmlns.com/foaf/0.1/'> .
@prefix xsd: <http://www.w3.org/2001/XMLSchema#> .

<issue7> a :Issue , :SecurityIssue ;
    :state :unassigned ;
    :reportedBy <user6> , <user2> ; # cardinality 1
    :reportedOn "2012-12-31T23:57:00"^^xsd:dateTime ;
    :assignedTo <user2>, <user1> ;
    :assignedOn "2012-11-31T23:57:00"^^xsd:dateTime ;
                       # reproduced before being reported
    :related <issue4>, <issue3>, <issue2> .
                       # referenced issues not included

<issue4> # a ???         missing type arc
    :state :unsinged ; # misspelled
    # :reportedBy ??? -  missing
    :reportedOn "2012-12-31T23:57:00"^^xsd:dateTime .

<user2> a foaf:Person ;
    foaf:givenName "Alice" ;
    foaf:familyName "Smith" ;
    foaf:phone <tel:+1.555.222.2222> ;
    foaf:mbox <mailto:alice@example.com> .

<user6> a foaf:Agent ; # should be foaf:Person
    foaf:givenName "Bob" ; # foaf:familyName "???" - missing
    foaf:phone <tel:+.555.222.2222> ; # malformed tel: URL
    foaf:mbox <mailto:alice@example.com> .

Value Sets

enumerated set
dereferenced enumerated set
too large to dereference
inclusion by lexical constraints (e.g. IRI stem)
product of OWL union/intersection

Technologies

in order of specificity:

SPARQL
path-indexed ASK
SPARQL as SPIN
OWL/ICV
Description Set Profiles
Resource Shapes
SHACL
Shape Expressions

Use Case: BP Observation

Blood pressure observation
with a systolic component
with values in mm mergury.

<Obs1>
  :component [ :code "systolic"; :value "140"^^:mmHg ] .

Resource Shapes

<BPShape> a rs:ResourceShape ;
  rs:property [
    rs:name "component" ;
    rs:propertyDefinition :component ;
    rs:valueShape [
      a rs:ResourceShape ;
      rs:property [
        rs:name "code" ;
        rs:propertyDefinition :code ;
        rs:allowedValue "systolic" ;
        rs:occurs rs:Exactly-one ;
      ] ;
      rs:property [
        rs:name "value" ;
        rs:propertyDefinition :value ;
        rs:valueType :mmHg ;
        rs:occurs rs:Exactly-one ;
      ]
    ] ;
    rs:occurs rs:Exactly-one ;
  ] .

RDF vocabulary for describing RDF vocabularies
Designed by OSLC for OSLC use cases
Focus on description rather than validation
specifications: vocabulary, use cases
no particular extensibility mechanism

SHACL

PREFIX : <http://a.example/#>
PREFIX sh:<http://www.w3.org/ns/shacl#>
<BPShape> a sh:Shape ;
  sh:property [
    sh:predicate :component ;
    sh:minCount 1 ; sh:maxCount 1 ;
    sh:shape [
      sh:property [
        sh:predicate :code ;
        sh:in ("systolic") ;
        sh:minCount 1 ; sh:maxCount 1 ;
      ] ;
      sh:property [
        sh:predicate :value ;
        sh:valueType :mmHg ;
        sh:minCount 1 ; sh:maxCount 1 ;
      ]
    ] ;
    sh:minCount 1 ; sh:maxCount 1 ;
  ] .

Vocabulary associated with user-defined functions defined in SPARQL or javascript.
Syntax like Resource Shapes/Description Set Profiles (described later).
SPARQL as extensibility mechansim.

visible diffs from Resource Shapes:

change namespace
rs:propertyDefinition → sh:predicate
rs:allowedValue X,Y → sh:in(X Y)
rs:occurs rs:exactly-once → sh:minCount 1; sh:maxCount 1

(many more diffs not seen in these examples)

What about the diastolic?

Blood pressure observation
with a systolic component
and a diastolic component
with values in mm mergury.

<Obs1>
  :component [ :code "systolic"; :value "140"^^:mmHg ]
  :component [ :code "diastolic"; :value "80"^^:mmHg ] .

Resource Shapes

<BPShape> a rs:ResourceShape ;
  rs:property [
    rs:name "component" ;
    rs:propertyDefinition :component ;
    rs:valueShape [
      a rs:ResourceShape ;
      rs:property [
        rs:name "code" ;
        rs:propertyDefinition :code ;
        rs:allowedValue "systolic" ;
        rs:occurs rs:Exactly-one ;
      ] ;
      rs:property [
        rs:name "value" ;
        rs:propertyDefinition :value ;
        rs:valueType :mmHg ;
        rs:occurs rs:Exactly-one ;
      ]
    ] ;
    rs:occurs rs:Exactly-one ;
  ] ;

…
  rs:property [
    rs:name "component" ;
    rs:propertyDefinition :component ;
    rs:valueShape [
      a rs:ResourceShape ;
      rs:property [
        rs:name "code" ;
        rs:propertyDefinition :code ;
        rs:allowedValue "diastolic" ;
        rs:occurs rs:Exactly-one ;
      ] ;
      rs:property [
        rs:name "value" ;
        rs:propertyDefinition :value ;
        rs:valueType :mmHg ;
        rs:occurs rs:Exactly-one ;
      ]
    ] ;
    rs:occurs rs:Exactly-one ;
  ] .

<Obs1>
  :component [ :code "systolic"; :value "140"^^:mmHg ]
  :component [ :code "diastolic"; :value "80"^^:mmHg ] .

SHACL

PREFIX : <http://a.example/#>
PREFIX sh:<http://www.w3.org/ns/shacl#>
<BPShape> a sh:Shape ;
  sh:property [
    sh:predicate :component ;
    sh:qualifiedMinCount 1 ; sh:qualifiedMaxCount 1 ;
    sh:qualifiedShape [
      sh:property [
        sh:predicate :code ;
        sh:in ("systolic") ;
        sh:minCount 1 ; sh:maxCount 1 ;
      ] ;
      sh:property [
        sh:predicate :value ;
        sh:valueType :mmHg ;
        sh:minCount 1 ; sh:maxCount 1 ;
      ]
    ] ;
  ] ;

…
  sh:property [
    sh:predicate :component ;
    sh:qualifiedMinCount 1 ; sh:qualifiedMaxCount 1 ;
    sh:qualifiedShape [
      sh:property [
        sh:predicate :code ;
        sh:in ("diastolic") ;
        sh:minCount 1 ; sh:maxCount 1 ;
      ] ;
      sh:property [
        sh:predicate :value ;
        sh:valueType :mmHg ;
        sh:minCount 1 ; sh:maxCount 1 ;
      ]
    ] ;
  ] .

<Obs1>
  :component [ :code "systolic"; :value "140"^^:mmHg ]
  :component [ :code "diastolic"; :value "80"^^:mmHg ] .

<Obs2>
  :component [ :code "systolic"; :value "140"^^:mmHg ]
  :component [ :code "diastolic"; :value "80"^^:mmHg ]
  :component "something invalid" .

SHACL

PREFIX : <http://a.example/#>
PREFIX sh:<http://www.w3.org/ns/shacl#>
<BPShape> a sh:Shape ;
  sh:property [
    sh:predicate :component ;
    sh:qualifiedMinCount 1 ; sh:qualifiedMaxCount 1 ;
    sh:qualifiedShape [
      …
    ] ;
    sh:minCount 1 ; sh:maxCount 1 ;
  ] ;
  sh:property [
    sh:predicate :component ;
    sh:qualifiedMinCount 1 ; sh:qualifiedMaxCount 1 ;
    sh:qualifiedShape [
      …
    ] ;
  ] .

  sh:property [
    sh:predicate :component ;
    sh:Shape [
      sh:Or (
        [ sh:property [
            sh:predicate :code ;
            sh:in ("systolic") ;
            sh:minCount 1 ; sh:maxCount 1 ;
          ] ;
          sh:property [
            sh:predicate :value ;
            sh:valueType :mmHg ;
            sh:minCount 1 ; sh:maxCount 1 ;
          ]
        ]
        [ sh:property [
            sh:predicate :code ;
            sh:in ("diastolic") ;
            sh:minCount 1 ; sh:maxCount 1 ;
          ] ;
          sh:property [
            sh:predicate :value ;
            sh:valueType :mmHg ;
            sh:minCount 1 ; sh:maxCount 1 ;
          ]
        ]
      )
    ]
  ]
.

Shape Expressions

PREFIX : <http://a.example/#>
<BPShape> {
  :component { :code ["systolic"]; :value :mmHg }
}

specs: semantics and compact syntax
provides apparent semantics for Resource Shapes (plus OR and semantic actions)
defined extensibility mechanism

Shape Expressions

PREFIX : <http://a.example/#>
<BPShape> {
  :component { :code ["systolic"]; :value :mmHg } ;
  :component { :code ["diastolic"]; :value :mmHg }
}

specs: semantics and compact syntax
provides apparent semantics for Resource Shapes (plus OR and semantic actions)
defined extensibility mechanism
purpose-built syntax and semantics

ShEx deeper dive

What is Shape Expressions?

Surface syntax for Resource Shapes
+ disjunction
+ semantic actions
grouping
+ negations
+ reverse paths
arbitrary cardinality: '?', '*', '+' and {M,N} constructs
driven by RegEx intuition

PREFIX issue: <http://ex.example/>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>
PREFIX myco: <http://myco.example/#>

start = <IssueShape>

<IssueShape> {
    issue:status (issue:unassigned issue:assigned),
    issue:reportedBy @<UserShape>,
    issue:reportedOn xsd:dateTime,
    ( issue:assignedTo @<EmployeeShape>,
      issue:assignedOn xsd:dateTime )?
}

<UserShape> {
    (foaf:name LITERAL
     | foaf:givenName LITERAL+,
       foaf:familyName LITERAL),
    foaf:mbox IRI
}

<EmployeeShape> {
    foaf:page (myco:Employee~),
    foaf:givenName LITERAL+,
    foaf:familyName LITERAL,
    foaf:phone IRI*,
    foaf:mbox IRI
}

User example

What's my user profile look like?

PREFIX foaf: <http://xmlns.com/foaf/>

<UserShape> {
    (foaf:name LITERAL
     | foaf:givenName LITERAL+,
       foaf:familyName LITERAL),
    foaf:mbox IRI
}

either a foaf:name
or one or more foaf:givenNames
and a foaf:familyName
plus an foaf:mbox

User example

What's my user profile look like?

PREFIX foaf: <http://xmlns.com/foaf/>

<UserShape> {
    (foaf:name LITERAL
     |foaf:givenName LITERAL+,
       foaf:familyName LITERAL),
    foaf:mbox IRI
}

either a foaf:name
or one or more foaf:givenNames
and a foaf:familyName
plus an foaf:mbox

User example

What's my user profile look like?

PREFIX foaf: <http://xmlns.com/foaf/>

<UserShape> {
    (foaf:name LITERAL
     |foaf:givenName LITERAL+,
       foaf:familyName LITERAL),
    foaf:mbox IRI
}

either a foaf:name
or one or more foaf:givenNames
and a foaf:familyName
plus an foaf:mbox

compare with other schema languages...

RelaxNG Compact Syntax

    (element foaf:name { xsd:string }
     | (element foaf:givenName { xsd:string }+,
        element foaf:familyName { xsd:string })),
    element foaf:mbox { xsd:anyURI }

Regex

(N|(G+F))M

NM

GFM

GGGFM

W3C XML Schema

  <xs:complexType name="UserContent">
    <xs:sequence>
      <xs:choice>
        <xs:element name="name" type="xs:string"/>
        <xs:sequence>
          <xs:element maxOccurs="unbounded" name="givenName" type="xs:string"/>
          <xs:element name="familyName" type="xs:string"/>
        </xs:sequence>
      </xs:choice>
      <xs:element name="mbox" type="xs:anyURI"/>
    </xs:sequence>
  </xs:complexType>

User example

Trying it out

PREFIX foaf: <http://xmlns.com/foaf/>

start = <UserShape>

<UserShape> {
    (foaf:name LITERAL
     | foaf:givenName LITERAL+,
       foaf:familyName LITERAL),
    foaf:mbox IRI
}

Try it at ericP's indulgent interface
or Jose's more correct interface:
<Alice>, <Bob>, <Gene>, <Pat>,
<Eve> (passes, OR is non-exclusive)

PREFIX foaf: <http://xmlns.com/foaf/>

<Alice>
  foaf:name "Alice" ;
  foaf:mbox <mailto:alice@example.org> .

<Bob>
  foaf:givenName "Robert", "Edward" ; foaf:familyName "Smith" ;
  foaf:mbox <mailto:bob@example.org> .

<Gene>
  foaf:givenname "Gene" ; foaf:familyName "Smith" ; # misspelling
  foaf:mbox <mailto:gene@example.org> .

<Pat>
  foaf:givenName "Patricia" ; # no foaf:familyName
  foaf:mbox <mailto:pat@example.org> .

<Eve>
  foaf:name "Eve" ; # both name and givenName/familyName
  foaf:givenName "Eve" ; foaf:familyName "石川" ;
  foaf:mbox <mailto:eve@example.org> .

Exercise 1

Fix Gene, Pat and Eve

PREFIX foaf: <http://xmlns.com/foaf/>

start = <UserShape>

<UserShape> {
    (foaf:name LITERAL
     | foaf:givenName LITERAL+,
       foaf:familyName LITERAL),
    foaf:mbox IRI
}

Test at ericP's or Jose's interface

PREFIX foaf: <http://xmlns.com/foaf/>

<Gene>
  foaf:givenname "Gene" ; foaf:familyName "Smith" ;
  foaf:mbox <mailto:gene@example.org> .

<Pat>
  foaf:givenName "Patricia" ;
  foaf:mbox <mailto:pat@example.org> .

<Eve>
  foaf:name "Eve" ;
  foaf:givenName "Eve" ; foaf:familyName "石川" ;
  foaf:mbox <mailto:eve@example.org> .

Now break it;
it's more fun to demonstrate what validators whine about than what they accept.

Solution 1

Fixed Gene, Pat and Eve

PREFIX foaf: <http://xmlns.com/foaf/>

start = <UserShape>

<UserShape> {
    (foaf:name LITERAL
     | foaf:givenName LITERAL+,
       foaf:familyName LITERAL),
    foaf:mbox IRI
}

Test at ericP's or Jose's interface

PREFIX foaf: <http://xmlns.com/foaf/>

<Gene>
  foaf:givenName "Gene" ; foaf:familyName "Smith" ;
  foaf:mbox <mailto:gene@example.org> .

<Pat>
  foaf:givenName "Patricia" ; foaf:familyName "Gordon"  ;
  foaf:mbox <mailto:pat@example.org> .

<Eve1>
  foaf:name "Eve" ;
  foaf:mbox <mailto:eve@example.org> .

<Eve2>
  foaf:givenName "Eve" ; foaf:familyName "石川" ;
  foaf:mbox <mailto:eve@example.org> .

ShExC is a grammar

ShExC is a BNF for RDF:

Issue: reportedOn
       reportedBy
       ( assignedOn
         assignedTo )?

reportedBy: User

assignedTo: User

User: ( name
        | givenName+
          familyName )
      mbox

example instance

<Issue1>
    issue:reportedBy <Bob> ;
    issue:reportedOn "2013-01-23T10:18:00"^^xsd:dateTime .

<Bob>
    foaf:givenName "Robert", "Edward" ; foaf:familyName "Smith" ;
    foaf:mbox <mailto:bob@example.org> .

ShExC is a grammar

ShExC is a BNF for RDF:

start = <IssueShape>

<IssueShape> {
    issue:reportedBy @<UserShape>,
    issue:reportedOn xsd:dateTime,
    ( issue:assignedTo @<UserShape>,
      issue:assignedOn xsd:dateTime )?
}

<UserShape> {
    (foaf:name LITERAL
     | foaf:givenName LITERAL+,
       foaf:familyName LITERAL),
    foaf:mbox IRI
}

example instance

<Issue1>
    issue:reportedBy <Bob> ;
    issue:reportedOn "2013-01-23T10:18:00"^^xsd:dateTime .

<Bob>
    foaf:givenName "Robert", "Edward" ; foaf:familyName "Smith" ;
    foaf:mbox <mailto:bob@example.org> .

Nested shapes

Schema:

PREFIX : <http://a.example/#>
<BPShape> {
  :component { :code ["systolic"]; :value :mmHg };
  :component { :code ["diastolic"]; :value :mmHg }
}

Nested patterns can be labeled:

<BPShape> {
  :component @<systolicComponent>;
  :component @<diastolicComponent>
}

<systolicComponent> {
  :code ["systolic"];
  :value :mmHg
}

<diastolicComponent> {
  :code ["diastolic"];
  :value :mmHg
}

Data:

<Obs1>
  :component [ :code "systolic"; :value "140"^^:mmHg ]
  :component [ :code "diastolic"; :value "80"^^:mmHg ] .

Exercise 2

Decompose this schema:

PREFIX obs: <http://hl7.org/fhir/Observations.>
<Observation> {
  obs:code @<CodeableConcept>;
  (
    obs:valueQuantity @<Quantity>  |
    obs:valueCodeableConcept @<CodeableConcept>
  )?;
  obs:component {
    obs:component.code @<CodeableConcept>;
    (
      obs:component.valueQuantity @<Quantity>  |
      obs:component.valueCodeableConcept @<CodeableConcept>
    )?;
  }*;
}

Split the obs:component object into a <Observation.component>.

requires a code.
requires either a valueQuantity or valueCodeableConcept

example instance:

<Obs1> obs:code [
    cc:coding [
      cd:system [
        fhir:value "http://a.example/sys1";
        fhir:code "code1" ] ] ];
  obs:valueCodeableConcept [
    cc:coding [
      cd:system [
        fhir:value "http://a.example/sys1";
        fhir:code "code2" ] ] ];
  obs:component [
    obs:component.code [
      cc:coding [
        cd:system [
          fhir:value "http://a.example/sys1";
          fhir:code "code3" ] ] ];
    obs:component.valueQuantity [
      fhir:Quantity.value [ fhir:value 1.23] ];
  ] .

Solution 2

Decomposed schema:

PREFIX obs: <http://hl7.org/fhir/Observations.>
<Observation> {
  obs:code @<CodeableConcept>;
  (
    obs:valueQuantity @<Quantity>  |
    obs:valueCodeableConcept @<CodeableConcept>
  )?;
  obs:component @<Observation.component>*;
}

<Observation.component> {
  obs:component.code @<CodeableConcept>;
  (
    obs:component.valueQuantity @<Quantity>  |
    obs:component.valueCodeableConcept @<CodeableConcept>
  )?;
}

Create a <EmployeeShape> distinct from an <UserShape>.

example instance:

<Obs1> obs:code [
    cc:coding [
      cd:system [
        fhir:value "http://a.example/sys1";
        fhir:code "code1" ] ] ];
  obs:valueCodeableConcept [
    cc:coding [
      cd:system [
        fhir:value "http://a.example/sys1";
        fhir:code "code2" ] ] ];
  obs:component [
    obs:component.code [
      cc:coding [
        cd:system [
          fhir:value "http://a.example/sys1";
          fhir:code "code3" ] ] ];
    obs:component.valueQuantity [
      fhir:Quantity.value [ fhir:value 1.23] ];
  ] .

ShEx Value sets

A value set is a set of possible values.

Value sets appear in []s in the value (object) position.
Values can be literals or IRIs or stems.

A stem is any URI form ending with a '~'.

issue:assigned~

matches

issue:assignedBy

,

issue:assignedTo

,

issue:assignedOn

.

A '.' followed by any number of ('-' (stem | IRI)).
A singleton value set asserts a required value.

ex:mood ["happy" "sad" "indigo"]

ex:mood [mood:happy mood:sad mood:indigo]

ex:mood [mood:~]

Exercise 3

observation type is (exactly) <fhir:Observation>
if the fhir:nodeRole property is present, it has the value fhir:treeRoot.
fhirvs:observation-status ["registered" "preliminary" "final" "amended" "cancelled" "entered-in-error" "unknown"]
fhirvs:narrative-status ["generated" "extensions" "additional" "empty"]
fhirvs:observation-relationshiptypes ["has-member" "derived-from" "sequel-to" "replaces" "qualified-by" "interfered-by"]

Schema:

PREFIX fhir: <http://hl7.org/fhir/>
PREFIX fhirvs: <http://hl7.org/fhir/ValueSet/>
<Observation> {
  a [fhir:Observation];
  fhir:nodeRole [fhir:treeRoot]?;
  fhir:DomainResource.text {
    fhir:Narrative.status @<code> AND
      {fhir:value .; # <--
    fhir:Narrative.div LITERAL;
  }?;
  fhir:Observation.status @<code> AND
    {fhir:value .; # <--
  fhir:Observation.related {
    fhir:Observation.related.type @<code> AND
      {fhir:value .?; # <--
    fhir:Observation.related.target {
      fhir:link @<Observation> OR CLOSED {a [fhir:Observation]};
    }
  }*;
}
<code>     NONLITERAL { fhir:value LITERAL?; }

Data:

PREFIX fhir: <http://hl7.org/fhir/>
<Obs1> a fhir:Observation ;
  fhir:nodeRole fhir:treeRoot ;
  fhir:DomainResource.text [
    fhir:Narrative.status [ fhir:value "generated" ];
    fhir:Narrative.div "blah blah" ;
  ];
  fhir:Observation.status [ fhir:value "registered" ] ;
  fhir:Observation.related [
    fhir:Observation.related.type [ fhir:value "replaces" ] ;
    fhir:Observation.related.target [ fhir:link <Obs2> ]
  ] .
<Obs2> a fhir:Observation .

Solution 3

Schema:

PREFIX fhir: <http://hl7.org/fhir/>
PREFIX fhirvs: <http://hl7.org/fhir/ValueSet/>
<Observation> {
  a [fhir:Observation];
  fhir:nodeRole [fhir:treeRoot]?;
  fhir:DomainResource.text {
    fhir:Narrative.status @<code> AND
      {fhir:value @fhirvs:narrative-status};
    fhir:Narrative.div LITERAL;
  }?;
  fhir:Observation.status @<code> AND
    {fhir:value @fhirvs:observation-status};
  fhir:Observation.related {
    fhir:Observation.related.type @<code> AND
      {fhir:value @fhirvs:observation-relationshiptypes}?;
    fhir:Observation.related.target {
      fhir:link @<Observation> OR CLOSED {a [fhir:Observation]};
    }
  }*;
}
<code>     NONLITERAL { fhir:value LITERAL?; }
fhirvs:narrative-status
  ["generated" "extensions" "additional" "empty"]
fhirvs:observation-status
  ["registered" "preliminary" "final" "amended"
   "cancelled" "entered-in-error" "unknown"]
fhirvs:observation-relationshiptypes
  ["has-member" "derived-from" "sequel-to"
   "replaces" "qualified-by" "interfered-by"]

Data:

PREFIX fhir: <http://hl7.org/fhir/>
<Obs1> a fhir:Observation ;
  fhir:nodeRole fhir:treeRoot ;
  fhir:DomainResource.text [
    fhir:Narrative.status [ fhir:value "generated" ];
    fhir:Narrative.div "blah blah" ;
  ];
  fhir:Observation.status [ fhir:value "registered" ] ;
  fhir:Observation.related [
    fhir:Observation.related.type [ fhir:value "replaces" ] ;
    fhir:Observation.related.target [ fhir:link <Obs2> ]
  ] .
<Obs2> a fhir:Observation .

Value sets in FHIR

codes:

fhirvs:narrative-status
  ["generated" "extensions" "additional" "empty"]

codings:

fhirvs:claim-type {
  fhir:Coding.system {
    fhir:value
      ["http://hl7.org/fhir/ex-claimtype"]};
  fhir:Coding.code {
    fhir:value
      ["oral" "institutional" "pharmacy" "professional" "vision"]
  }
}

one of many codings:

fhirvs:device-action EXTRA fhir:CodeableConcept.coding {
    fhir:CodeableConcept.coding {
        fhir:Coding.system {fhir:value ["http://hl7.org/fhir/device-action"]}?;
        fhir:Coding.code {fhir:value ["implanted" "explanted" "manipulated"]}
    }
}

EXTRA

Schema:

PREFIX fhir: <http://hl7.org/fhir/>
PREFIX fhirvs: <http://hl7.org/fhir/ValueSet/>
fhirvs:device-action EXTRA fhir:CodeableConcept.coding {
    fhir:CodeableConcept.coding {
        fhir:Coding.system {fhir:value ["http://hl7.org/fhir/device-action"]}?;
        fhir:Coding.code {fhir:value ["implanted" "explanted" "manipulated"]}
    }
}

Data:

PREFIX fhir: <http://hl7.org/fhir/>
<da1>
  fhir:CodeableConcept.coding [
    fhir:Coding.system [fhir:value "http://hl7.org/fhir/device-action"];
    fhir:Coding.code [fhir:value "implanted"]
  ];
  fhir:CodeableConcept.coding [ ] .

Reverse arcs

Test arcs coming into an object.

Useful for e.g. two Observations connected through a common Encounter.

PREFIX issue: <http://ex.example/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

start = <IssueShape>

<IssueShape> {
    issue:status (issue:assigned issue:resolved ),
    issue:reportedOn xsd:dateTime,
    issue:assignedOn xsd:dateTime,
    issue:related @<RefdIssueShape>
}

<RefdIssueShape> {
    issue:name LITERAL,
    ^issue:related <IssueShape>
}

PREFIX issue: <http://ex.example/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

<Issue1>
    issue:status     issue:assigned ;
    issue:reportedOn "2013-01-23T10:18:00"^^xsd:dateTime ;
    issue:assignedOn "2013-01-23T11:00:00"^^xsd:dateTime ;
    issue:related    <Issue3> .

<Issue3>
    issue:name "smokey" .

Closed shapes

Use case: storing application data in a

conventional (normalized) relational database.
controlled triple store
physical state (e.g. a lightswich on the wall)

"If I tell you X, will you understand it?"

Not everything is a triple store.
No service could understand everything you might send it.

Closed shapes

What about re-used nodes?

Closed shapes:

PREFIX issue: <http://ex.example/>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

start = <IssueShape>

<IssueShape> {
    issue:reportedBy @<UserShape>,
    issue:assignedTo @<EmployeeShape>?
}

<UserShape> {
    foaf:name LITERAL
    foaf:mbox IRI
}

<EmployeeShape> {
    foaf:givenName LITERAL+,
    foaf:familyName LITERAL,
    foaf:mbox IRI
}

<User2> fits multiple shapes:

PREFIX ex: <http://ex.example/>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

<Issue1>
    ex:reportedBy   <User2> ;
    ex:assignedTo   <User2> ;

<User2>
    foaf:name "Bob Smith" ;
    foaf:givenName "Bob" ;
    foaf:familyName "Smith" ;
    foaf:mbox <mailto:bob@example.org> .

Semantic actions

For extensibility:

Languages like RelaxNG, W3C XML Schema, etc.
annotations provide place to add more functionality
e.g. compare dates, test string structure

For actions:

Languages like yacc, ANTLR, etc
do something useful with the parsed tokens

ShExMap

see ShExMap docs

API-native

Provides purpose-fit expressivity with minimal implementation cost.

Can count on order.

Javascript API actions embedded in %js{ … %}.

PREFIX issue: <http://ex.example/>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

start = <IssueShape>

<IssueShape> {
    issue:state (issue:unassigned issue:assigned),
    issue:reportedBy @<UserShape>,
    issue:reportedOn xsd:dateTime
        %js{ report = _.o; return true; %},
    (issue:reproducedBy @<EmployeeShape>,
     issue:reproducedOn xsd:dateTime
        %js{ return _.o.lex > report.lex; %}
    ),
    issue:related @<IssueShape>*
}

<UserShape> {
    (foaf:name xsd:string
     | foaf:givenName xsd:string+,
       foaf:familyName xsd:string),
    foaf:mbox IRI
}

<EmployeeShape> {
    foaf:givenName xsd:string+,
    foaf:familyName xsd:string,
    foaf:phone IRI*,
    foaf:mbox IRI
}

PREFIX issue: <http://ex.example/>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

<Issue1>
    ex:state        ex:unassigned ;
    ex:reportedBy   <User2> ;
    ex:reportedOn   "2013-01-23T10:18:00"^^xsd:dateTime ;
    ex:reproducedBy <Thompson.J> ;
    ex:reproducedOn "2013-01-23T10:00:00"^^xsd:dateTime .

<User2>
    foaf:givenName "Bob" ;
    foaf:familyName "Smith" ;
    foaf:mbox <mail:bob@example.org> .

<Thompson.J>
    foaf:givenName "Joe", "Joseph" ;
    foaf:familyName "Thompson" ;
    foaf:phone <tel:+456> ;
    foaf:mbox <mail:joe@example.org> .

SPARQL

Isn't there an RDF query language?

%sparql{ … %} can complement our %js{ … %} actions.

PREFIX issue: <http://ex.example/>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

start = <IssueShape>

<IssueShape> {
    issue:state (issue:unassigned issue:assigned),
    issue:reportedBy @<UserShape>,
    issue:reportedOn xsd:dateTime
        %js{ report = _.o; return true; %},
    (issue:reproducedBy @<EmployeeShape>,
     issue:reproducedOn xsd:dateTime
        %js{ return _.o.lex > report.lex; %}
        %sparql{ ?s issue:reportedOn ?rpt . FILTER (?o > ?rpt) %}
    ),
    issue:related @<IssueShape>*
}

<UserShape> {
    (foaf:name xsd:string
     | foaf:givenName xsd:string+,
       foaf:familyName xsd:string),
    foaf:mbox IRI
}

<EmployeeShape> {
    foaf:givenName xsd:string+,
    foaf:familyName xsd:string,
    foaf:phone IRI*,
    foaf:mbox IRI
}

PREFIX issue: <http://ex.example/>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

<Issue1>
    ex:state        ex:unassigned ;
    ex:reportedBy   <User2> ;
    ex:reportedOn   "2013-01-23T10:18:00"^^xsd:dateTime ;
    ex:reproducedBy <Thompson.J> ;
    ex:reproducedOn "2013-01-23T10:00:00"^^xsd:dateTime .

<User2>
    foaf:givenName "Bob" ;
    foaf:familyName "Smith" ;
    foaf:mbox <mail:bob@example.org> .

<Thompson.J>
    foaf:givenName "Joe", "Joseph" ;
    foaf:familyName "Thompson" ;
    foaf:phone <tel:+456> ;
    foaf:mbox <mail:joe@example.org> .

Generators

Semantic actions can have side-effects.

PREFIX ex: <http://ex.example/#>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>

      %js{
           action.arguments[1].message("in init");
           doc = document.implementation.createDocument
                   ("http://ex.example/xml", "Issue", undefined);
           issue = doc.childNodes[0];
      %}

start = <IssueShape>

<IssueShape> {
    ex:state (ex:unassigned ex:assigned)
      %js{ issue.setAttribute("id", _.s.lex);
           issue.setAttribute("state", _.o.lex.substr(19));
           person = user = doc.createElement("Reported");
      %},
    ex:reportedBy @<Person>
      %js{{ issue.appendChild(user); %},
    ex:reportedOn xsd:dateTime
      %js{{ user.setAttribute("reportedOn", _.o.lex);
           person = reproducer = doc.createElement("Reproduced");
      %},
    (ex:reproducedBy @<EmployeeShape>
      %js{{ issue.appendChild(person);%},
     ex:reproducedOn xsd:dateTime
      %js{{ reproducer.setAttribute("reportedOn", _.o.lex);
      %})?,
    ex:related @<IssueShape>*
} %js{ 
window.open("data:application/xml;charset=utf-8;base64,"
  + Base64.encode((new XMLSerializer()).serializeToString(doc)),
  "toy");
 %}
VIRTUAL <Person> {
    (foaf:name xsd:string
     | foaf:givenName xsd:string+
      %js{ gn = doc.createElement("GivenName");
           gn.appendChild(doc.createTextNode(_.o.lex));
           person.appendChild(gn);
      %},
       foaf:familyName xsd:string
      %js{ fn = doc.createElement("FamilyName");
           fn.appendChild(doc.createTextNode(_.o.lex));
           person.appendChild(fn);
      %}
    ),
    foaf:mbox IRI
      %js{ fn = doc.createElement("Email-address");
           fn.appendChild(doc.createTextNode(_.o.lex));
           person.appendChild(fn);
      %}
}
<UserShape> & <Person> {
}
<EmployeeShape> {
    & <Person>,
    foaf:phone IRI+
      %js{ fn = doc.createElement("Telephone-Number");
           fn.appendChild(doc.createTextNode(_.o.lex));
           person.appendChild(fn);
      %}
}

<Issue xmlns="http://ex.example/xml" id="Issue1" state="unassigned">
  <Reported reportedOn="2013-01-23T10:18:00">
    <GivenName>Bob</GivenName>
    <FamilyName>Smith</FamilyName>
    <Email-address>mailto:bob@example.org</Email-address>
  </Reported>
  <Reproduced reportedOn="2013-01-23T11:00:00">
    <GivenName>Joe</GivenName>
    <GivenName>Joseph</GivenName>
    <FamilyName>Thompson</FamilyName>
    <Email-address>mailto:joe@example.org</Email-address>
    <Telephone-Number>tel:+456</Telephone-Number>
  </Reproduced>
</Issue>

GenX

GenX details.

=<function>?, $<namespace>, !debugger

PREFIX ex: <http://ex.example/#>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

                            %GenX{ issue $http://ex.example/xml %}

start = <IssueShape>

<IssueShape> {
    ex:state (ex:unassigned ex:assigned) %GenX{ @state =substr(19) %},
    ex:reportedBy @<Person>              %GenX{ reported = %},
    ex:reportedOn xsd:dateTime           %GenX{ [-1]@date %},
    (ex:reproducedBy @<EmployeeShape>    ,
     ex:reproducedOn xsd:dateTime        %GenX{ @date %}
    )?                                   %GenX{ reproduced = %},
    ex:related @<IssueShape>*
}                                        %GenX{ @id %}
VIRTUAL <Person> {
    (foaf:name xsd:string                %GenX{ full-name %}
     | foaf:givenName xsd:string+        %GenX{ given-name %},
       foaf:familyName xsd:string        %GenX{ family-name %}
    ),
    foaf:mbox IRI               %GenX{ email-addr %}
}
<UserShape> & <Person> {
}
<EmployeeShape> {
    & <Person>,
    foaf:phone IRI+             %GenX{ tel-number %}
}

<Issue xmlns="http://ex.example/xml" id="Issue1" state="unassigned">
  <Reported reportedOn="2013-01-23T10:18:00">
    <GivenName>Bob</GivenName>
    <FamilyName>Smith</FamilyName>
    <Email-address>mailto:bob@example.org</Email-address>
  </Reported>
  <Reproduced reportedOn="2013-01-23T11:00:00">
    <GivenName>Joe</GivenName>
    <GivenName>Joseph</GivenName>
    <FamilyName>Thompson</FamilyName>
    <Email-address>mailto:joe@example.org</Email-address>
    <Telephone-Number>tel:+456</Telephone-Number>
  </Reproduced>
</Issue>

GenJ

GenJ is GenX's JSON counterpart.

PREFIX ex: <http://ex.example/#>
PREFIX foaf: <http://xmlns.com/foaf/>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>
                                         %GenJ{ @id %}

start = <IssueShape>

<IssueShape> {
    ex:state (ex:unassigned ex:assigned) %GenJ{ state %},
    ex:reportedBy @<User>                %GenJ{ rprtBy @id %},
    ex:reportedOn xsd:dateTime           %GenJ{ rprtOn %},
    (ex:reproducedBy @<User>             %GenJ{ rpdcBy @id %},
     ex:reproducedOn xsd:dateTime        %GenJ{ rpdcOn %}
    )?,
    ex:related @<IssueShape>*
}
<User> {
    (foaf:name xsd:string                %GenJ{ name %}
     | foaf:givenName xsd:string+        %GenJ{ fname %},
       foaf:familyName xsd:string        %GenJ{ lname %}
    ),
    foaf:mbox IRI                        %GenJ{ email %},
    foaf:phone IRI?                      %GenJ{ tel %}
}

{"@context":
 {"ns0":"http://ex.example/#", "state":"ns0:state",
  "rprtBy":"ns0:reportedBy", "ns1":"http://xmlns.com/foaf/",
  "fname":"ns1:givenName", "lname":"ns1:familyName",
  "email":"ns1:mbox", "ns2":"http://www.w3.org/2001/XMLSchema#",
  "rprtOn":{
      "@id":"ns0:reportedOn", "@type":"ns2:dateTime" },
  "rpdcBy":"ns0:reproducedBy",
  "tel":"ns1:phone",
  "rpdcOn":{
      "@id":"ns0:reproducedOn", "@type":"ns2:dateTime" } },
 "@id":"Issue1",
 "state":"http://ex.example/#unassigned",
 "rprtBy":{
     "@id":"User2", "fname":"Bob", "lname":"Smith",
     "email":"mailto:bob@example.org" },
 "rprtOn":"2013-01-23T10:18:00",
 "rpdcBy":{
     "@id":"Thompson.J", "fname":["Joe","Joseph"], "lname":"Thompson",
     "email":"mailto:joe@example.org", "tel":"tel:+456"
 },
 "rpdcOn":"2013-01-23T11:00:00"
}

GenR

GenR is GenJ's RDF counterpart.

Generate RDF from RDF?!?

PREFIX ex: <http://ex.example/#>
PREFIX my: <http://my.example/#>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

                                         %GenR{ %}
start = <IssueShape>

<IssueShape> {
    ex:state (ex:unassigned ex:assigned )
    %GenR{ ISSUE = IRI(CONCAT(STR(s), "/myMark")) .
           STATE = SUBSTR(STR(o), 19) .
           ISSUE my:status STATE %},
    ex:reportedOn xsd:dateTime 
          %GenR{ ON = BNODE("on") .
                 ISSUE my:on ON .
                 DATE=CONCAT(YEAR(o),'-'MONTH(o),'-',DAY(o)) .
                 TIME=CONCAT(HOURS(o),':'MINUTES(o),':',SECONDS(o),TZ(o)) .
                 ON ex:date STRDT(DATE, xsd:date) .
                 ON ex:time STRDT(TIME, xsd:time) %}
}

PREFIX ex: <http://ex.example/#>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>

<Issue1>
    ex:state        ex:unassigned ;
    ex:reportedOn   "2013-01-23T10:18:12Z"^^xsd:dateTime .

<Issue1/myMark> my:status "unassigned" ;
  my:on [
    my:date "2013-01-23"^^xsd:date ;
    my:time "10:18:12Z"^^xsd:time
  ] .

this example

example converting between clinical data schemas

GenN

Normalize RDF.

Limited use:

Normalize for very simple parsers.
Produce a sub-langauge of RDF.
Avoid the 2nd-level application parse of an RDF graph.
Consistent examples in documentation.
Canonicalize for diffing.

%GenN{ %}

example

Resources

ShEx wiki https://www.w3.org/2001/sw/wiki/ShEx
Definition as set notation, Z-notation
Primer
Demo
algebra tests, general tests
python impl (Mayo)
http://www.slideshare.net/jelabra/semantics-2014

Semantic Representations of Clinical Care Data Leveraging HL7 FHIR

SWAT4LS4 Dec, 2016

http://tinyurl.com/swat4lsfhir

http://www.w3.org/2016/Talks/1205-swat4ls-egp/validation/

FHIR Resource Schemas

Network of Schemas

Data Integrity

Schema Languages

RDF Data Integrity

Validation in RDF

Value Sets

Technologies

Use Case: BP Observation

Resource Shapes

SHACL

What about the diastolic?

Resource Shapes

SHACL

SHACL

Shape Expressions

Shape Expressions

ShEx deeper dive

User example

User example

User example

RelaxNG Compact Syntax

Regex

W3C XML Schema

User example

Exercise 1

Solution 1

ShExC is a grammar

ShExC is a grammar

Nested shapes

Exercise 2

Solution 2

ShEx Value sets

Exercise 3

Solution 3

Value sets in FHIR

EXTRA

Reverse arcs

Closed shapes

Closed shapes

Semantic actions

ShExMap

API-native

SPARQL

Generators

GenX

GenJ

GenR

GenN

Resources

SWAT4LS
4 Dec, 2016