DC2

Design conventions and principles

How to use terminologies (preferred binding)

As stated in section 1.5, to be universally exchangeable the International Patient Summary must rely on international multilingual reference terminologies. To that effect, each codeable element of the international patient summary template is bound to a Value Set built upon an international reference terminology, such as SNOMED CT, LOINC, UCUM or EDQM Standard Terms. In some selected cases, in consideration of the availability of other globally usable reference terminologies and for alignment with a future FHIR version of the IPS, FHIR-defined terminologies have been specified. These terminologies have been selected to provide the preferred bindings for the codeable elements of the patient summary. They are the primary terminologies of this specification.

Nevertheless, it is anticipated that in some situations a system producing an instance of patient summary might not support one or the other of these primary terminologies, supporting only a local interface terminology instead. Similarly, it is also anticipated that the receiving system might in some cases not be able to use a code in a patient summary, either because this code belongs to a primary terminology that the receiving system does not support or because this code belongs to an interface terminology specific to the country of the producing system.

In order to maximize the international scope and usability of patient summaries, and also to accommodate the exceptional situations listed above, this specification makes these requirements:

• The Primary Code of a codeable element SHOULD be populated.
• If populated, the Primary Code of a codeable element SHALL be chosen from the primary terminology assigned to the value set bound to this element; unless exceptions have been agreed.
• The 'displayName' of the Primary Code SHALL be populated with a term representing this same code in the terminology used, in the language chosen for the current instance of the patient summary.
• When the primary 'code' element is not populated, an appropriate 'nullFlavor' value SHALL be used along with the 'originalText' element (referencing a textual expression in the narrative representing the meaning for the producer) and/or one or more coded 'translation' elements.
• One or more Alternate Codes from a local interface terminology MAY be provided, each with its associated 'displayName'.
• In case the primary code is derived from an alternate terminology the alternate code SHOULD be provided in the translation element.

Primary Code

In the data type for codeable elements (CD constrained by the CD.IPS template), the Primary Code is represented by the attributes @code, @displayName, @codeSystem, @codeSystemName, @codeSystemVersion.

Alternate Code

In the data type for codeable elements (CD constrained by the CD.IPS template), an Alternate Code is carried in a 'translation' sub-element.

Original Text

In the data type for codeable elements (CD constrained by the CD.IPS template), the Original Text is provided in the 'originalText' sub-element.

Representing "known absent" and "not known"

In line with the properties of minimalism and non-exhaustiveness for the IPS (see the IPS definition above), and benefiting from the experience acquired with the European cross-border services, this guide explicitly addresses two general situations:

1. condition or activity unknown
2. condition or activity known absent.

Other kinds of negations such as: (a) the negation of an allergy to a specific agent; (b) the absence of a particular disease; or (c) the fact that a specific vaccination has not been performed, have been considered beyond the set of essential data for an IPS.

This specification represents this core set of negations (“general condition/activity unknown” and “general condition/activity/known absent” ) using explicit coded elements rather than relying on specific mechanisms of HL7 CDA such as nullFlavor and negationInd attributes or human readable text (possibly not understood by the foreign country receiver).

In contrast to the practice to use negationInd or nullFlavor attributes on a section itself, we prohibit the use of these attributes on section level to express “unknown” or “no information” situations. A section holds the categorized (coded) narrative part of the documented activity and will never carry negationInd or nullFlavor attributes. Instead, we enforce by design, that “unknown” or “no information” expressions always go to the coded entry with a corresponding act code.

The main reasons for this choice are:

• @negationInd in CDA has been superseded in V3 later by two other negation indicators: @actNegationInd and @valueNegationInd.
• To make clinical content representation less dependent on a particular format or syntax, enabling a more practical path to transforming and exchanging data from one standard format (e.g. CDA R2) to another (e.g. FHIR).
• to provide one single method to express the presence or absence of a particular condition (e.g. an allergy) or activity (e.g. an immunization), or the lack of knowledge regarding this kind of condition or activity, resulting in a more robust and easily implementable specification.

For the other kinds of negations, not explicitly mentioned in this guide, it is suggested to apply – where possible – the same approach. Future versions of this guide may extend the number of cases covered and include new coded concepts for supporting them.

When needed, more specific statements such as the absence of a specific condition or activity, although considered as beyond the set of essential data for IPS, can still be expressed by using the native negationInd attribute of CDA R2.

Uncoded information

An IPS originator may not be able to value a coded element with an appropriate coded concept, but only with textual information. This may happen for two reasons:

1. the originator is not able to express the concept in the reference value sets
2. the originator is not able to express the concept in any known terminology.

The first case, assuming that the coding strength is Required (aka CNE, coded, no extensions), is represented in this guide with the following assertion:

<code codeSystem="2.16.840.1.113883.6.96" nullFlavor="OTH">
  <originalText>
    <reference value="#ref1"/>
  </originalText>
</code>

That is expressing that there are no codes applicable in the referred code system (in the example, SNOMED CT). Please note that according to this guide the text is documented in the section narrative and only referenced by the coded element.

Note: Data Types R1 doesn't allow specifying that there are no codes applicable in the referred value set, as instead is possible with Data Types R2. Future versions of this guide may consider extending the data type to better support this situation.

The second case, that applies both to Required (aka CNE, coded no extensions) and Extensible (CWE, coded with extensions) coding strengths, is instead here represented valuing the coded element with the most generic nullFlavor “NI” (No Information) and pointing the text in the section narrative:

<value xsi:type="CD" nullFlavor="NI">
  <originalText>
    <reference value="#ref1"/>
  </originalText>
</value>

Note: The most proper NullFlavor code to be used here would be "UNC" (Uncoded). This code is available in the current and other recent versions of the HL7 RIM, but it is not present in version 2.07 of the RIM on which the CDA R2 standard is based. In the absence of "UNC", the most appropriate NullFlavor code to use for representing that the data is unable to be coded is "NI".

Unmapped Coded Concepts

In several real world situations the records used as source for the Patient Summaries may use locally adopted terminologies mapped into the reference value sets. When the original coded concept cannot be mapped in one of the coded concepts included in the reference value sets, it is recommended to populate the original code in the IPS instance (in a 'translation' sub-element), with a nullFlavor indicating that the mapping didn’t occur. (See also the "Concept code mapping" in the section "Functional requirements and high-level use cases".). The nullFlavor value depends upon the coding strength of the binding.

Two circumstances are considered here: the case in which the coding strength is Required (CNE) and when it is Extensible (CWE).

In the case of coding strength Required (CNE), use nullFlavor "OTH":

<value xsi:type="CD" codeSystem="2.16.840.1.113883.6.96" nullFlavor="OTH"> 
  [
    <originalText>
      <reference value="#ref1"/>
    </originalText>
  ] 
  <translation code="A02.9" codeSystem="2.16.840.1.113883.6.3"
    displayName="Infezioni da Salmonella non specificate"/>
</value>

The square brackets [ ] are used to indicate that the originalText element may or may not be present

Note: It may happen that a mapping would be possible in the target code system, but not in the target value set; Data Types R1 does not allow the specification of this difference, that there are no codes applicable in the reference value set, which is possible with Data Types R2.

Future versions of this guide may consider extending the data type to better support this situation.

In the case of Extensible (CWE) coding strength, this guide recommends representing the original code in the <translation> sub-element and using a nullFlavor for the primary code. This highlights the fact that a mapping to the reference value set was attempted, but no suitable target codes were identified.

<value xsi:type="CD" codeSystem="2.16.840.1.113883.6.96" nullFlavor="NI"> 
  [
    <originalText>
      <reference value="#ref1"/>
    </originalText>
  ] 
  <translation code="A02.9" codeSystem="2.16.840.1.113883.6.3" 
    displayName="Infezioni da Salmonella non specificate"/>
</value>

The square brackets [ ] are used to indicate that the originalText element may or may not be present.

Mapped coded concepts

As mentioned above, in several circumstances an original coded concept is mapped into the reference value set. When this occurs, both the original and the reference codes should be reported in the IPS instance.

Functional requirements exposed in "Concept Code Mapping" (multi-coding, link to original text, mapping between codes of the same code system) are also detailed below.

Case 1: Single local code mapped to primary code from the reference value set.

<value xsi:type="CD" code="42338000" codeSystem="2.16.840.1.113883.6.96"
  displayName="Salmonella gastroenteritis">
  [
    <originalText>
      <reference value="#ref1"/>
    </originalText>
  ] 
  <translation code="003.0" codeSystem="2.16.840.1.113883.6.103"
    displayName="Gastroenterite da Salmonella"/>
</value>

The square brackets [ ] are used to indicate that the originalText element may or may not be present

Case 2: Multiple local codes mapped together using nested 'translation' elements, and mapped to the primary code from the reference value set.

<value xsi:type="CD" code="422479008" codeSystem="2.16.840.1.113883.6.96"
  codeSystemName="SNOMED CT"
  displayName="FEMALE BREAST INFILTRATING DUCTAL CARCINOMA, STAGE 2">
  [
    <originalText>
       <reference value="#problem4name"/>
    </originalText>
  ]
  <translation code="code-example" codeSystem="1.999.999"
    codeSystemName="this is only an example"
    displayName="FEMALE BREAST INFILTRATING DUCTAL CARCINOMA, STAGE 2">
    <translation code="174.9" codeSystem="2.16.840.1.113883.6.103"
      codeSystemName="ICD-9CM"
      displayName="Malignant neoplasm of breast (female), unspecified"/>
    <translation code="C50.919" codeSystem="2.16.840.1.113883.6.90"
      codeSystemName="ICD-10-CM"
      displayName="Malignant neoplasm of unspecified site of unspecified female breast"/>
 </translation>
</value>

Case 3: Original and the reference coded concepts belong to the same code system (distinct codes). This may be the result of a different level of granularity between the original term and the reference value sets.

<code code="60591-5" codeSystem="2.16.840.1.113883.6.1"
  codeSystemName="LOINC" displayName="Patient Summary">
  <translation code="60592-3" codeSystem="2.16.840.1.113883.6.1"
    displayName="Patient summary unexpected contact"/>
</code>

Note: The R1 data type definition identifies the <translation> as “a set of other concept descriptors that translate this concept descriptor into other code systems." There is however a common understanding that it may be more than one representation in a single code system where code systems allow multiple representations, such as SNOMED CT. Data Types R2 extended the possibility to provide translations also in the same code system.

Translation of designations

The capability of recording one or more designations, in different languages, for the exchanged Patent Summary is one of the functional requirements requested for “safety and liability reasons” by the European Cross-border services (see "Designations’ Translation" in section "Functional requirements and high-level use cases" for more details).

Given that the base CDA R2 standard which uses datatypes R1 does not offer a native way to convey the language translation of 'displayName', this guide introduces an optional 'ips:designation' extension to the CD datatype for that purpose.

Below are examples of usage of this extension.

No code mapping

<code code="60591-5" codeSystem="2.16.840.1.113883.6.1"
  codeSystemName="LOINC" displayName="Patient Summary">
  <ips:designation language="it-IT">Profilo Sanitario Sintetico</ips:designation>
  <ips:designation language="fr-FR">Patient Summary</ips:designation>
  <ips:designation language="en">Patient Summary</ips:designation>
</code>

Including code mapping

<value xsi:type="CD" code="42338000" codeSystem="2.16.840.1.113883.6.96"
  displayName="Salmonella-gastroenterit">
  <ips:designation language="da-DK">Salmonella-gastroenterit</ips:designation>
  <ips:designation language="en">Salmonella gastroenteritis (disorder)</ips:designation> 
  [
    <originalText>
      <reference value="#ref1"/>
    </originalText>
  ]
  <translation code="003.0" codeSystem="2.16.840.1.113883.6.103"
    displayName="Gastroenterite da Salmonella"/>
</value>

Narrative Translations

“Narrative translation” means both the translation of the original narrative text, that can be human curated or automatically performed, and the generation of a translated narrative based on the coded entries.

The functional requirements associated with this process are described in "Designations’ Translation" in section "Functional requirements and high-level use cases", and can be summarized in two main points : (a) language identification and (b) distinguishable original and translated narratives. Moreover, the methodology applied for the narrative translation (e.g. derived from the coded entries; translated by a generic service;..) should be noted.

Regarding the translation of section narrative <text>, this guide recommends providing this translation on purely textual subordinate sections (one per translation) as specified in the IPS Translation Section (2.16.840.1.113883.10.22.3.15) template.

An example of this is:

<section>
  <templateId root="2.16.840.1.113883.3.1937.777.13.10.5"/>
   <id root="..." extension="..."/>
   <code code="48765-2" codeSystem="2.16.840.1.113883.6.1"
      displayName="Allergies and adverse reactions"/>
   <title>Allergies and Intolerances</title>
   <text>No known Allergies</text>
   <!-- omissions -->
   <component>
      <section>
        <!--  subordinate section carrying a translation of the parent section -->
        <title>Allergie ed Intolleranze</title>
        <text>Nessuna Allergia Nota</text>
        <languageCode code="it-IT"/>
      </section>
    </component>
</section>

Provenance

In the development of this Implementation Guide, consideration was given to the HL7 CDA® Release 2 Implementation Guide: Data Provenance, Release 1 - US Realm Draft Standard for Trial Use (December 2015). That guide provides a matrix offering a thorough and systematic analysis of provenance characteristics of electronic health records. Given the agreed scope principle (see section "Introduction") that the IPS be minimal and implementable, and the variable maturity and operational methods of existing national patient summaries, the proposal is that this first version should not attempt to require the full detail of that provenance specification.

The approach proposed for this version of the IPS is to:

• Allow optional documentation of section-level provenance.
• Require document-level provenance.
• Define IPS document provenance as one of two types: human-curated or software-assembled, based on the authors recorded in the header.
  • The classification is based on whether the IPS document is constructed by a human or an automated process, regardless of whether the IPS contains some content of both kinds.
• Require the IPS source system to identify the IPS provenance type and author.
  • The author shall be a human, if the IPS provenance type is "human-curated", or a device or system if the IPS provenance type is "software-assembled".
  • In the case of a software-assembled IPS that is then verified by a human, the document provenance type shall be "software-assembled" and the author shall be the device or system that constructed the IPS document, but an additional verifier identity shall name the human who performed this check. For the avoidance of doubt, verifier is not the same as legalAuthenticator (the verifier is represented as a participant with typeCode "VRF"). However, in cases where the verifying person intentionally wishes to sign the document, this shall be recorded as a legalAuthenticator.
• Allow optional section level author, provenance type, verifier, and informant identification, for IPS source systems that can support this.
• Not attempt to implement the US Realm CDA data provenance templates.

Note: Discussions with the EHR work group suggest that a possible future project should be an IPS functional profile, once there is greater clarity and operational experience of using the IPS.

Representation of Names

This specification requires that any Person Name is represented including at least the given and family components and therefore is never documented as a single string.

Even though it is recognized that there is not in all cultures the same concept of “family name”, no evidence has been collected in analyzing the international context (e.g. Japan, Korea; China) that justifies the retirement of this requirement.

Moreover, due to the global scope of the International Patient Summary, the case of non-alphabetic representations of the names has also been considered.

In this case, to facilitate the global use of the IPS, at least one alphabetic representation of the name SHALL be provided.