HL7 Templates SIG 08/05/01 a Rossi Mori, F Consorti

HL7 Templates SIG 08/05/01 A Rossi Mori, F Consorti

Assembling clinical information

Note by Angelo Rossi Mori and Fabrizio Consorti 2001-04-19

with an analysis of the examples provided by Stan Huff

Summary

In this working document we characterise a process of wrapping clinical information into a sequence of nesting constructs, from the elementary expressions to the whole record, and its storage or communication.

The goal is not to provide a universal theory on clinical information, but:

§  to set a framework to discuss examples of different kinds of structured data that need to be represented by one or more formalisms decided by the Templates SIG

§  to relate the fragments of clinical information to the RIM attirbutes

§  to explore the different kinds of models involved on this topic (terminology models, information models, etc).

We initially describe a wrapping sequence that brings from a base concept to a set of statements, as a series of Russian dolls. We distinguish three major levels:

1.  semantic fragments within an attribute of the information model ("encapsulation" of terminological issues)

2.  representation of a clinical statement by means of attributes of the information model

3.  representation of a set of clinical statements

We use the examples contained in a powerpoint presentation by Stan Huff: "A Characterization of Terminology Models, Clinical Templates, Message Models, and other kinds of Clinical Information Models".

Contents

Assembling clinical information 1

What is a clinical statement ? 2

kernel 2

status concepts 3

instantiation context 3

context about collection of information — attribution information 4

Graphical representation of the clinical statement wrappings 4

Payload = set of clinical statements 5

1. Siblings 5

2. Associated statements — clinical context 5

3. Sections 6

Transaction unit = envelope + payload 6

Graphical representation of the transaction unit 6

Analysis of Stan's examples 7

Complex dissections of the statement's terminological part 7

Terminology instance models 7

Terminology fractal models vs information models 8

Meta terminology models — Categorial structures 9

SAM, Semantic frActal Metamodel 9

Categorial structures in MoSe and CEN/TC251 9

Categorial strucutree on medical devices 9

Metamodels on LOINC 10

Statement-level clinical template 11

A bridge between terminology and the RIM 11

Context dependent ? 11

Example from ENV 14032, Clause 5. System of concepts to support nursing actions 12

Assembling multiple statements 13

Further wrapping, up to the transaction level 14

Stan's original overheads 15

What is a clinical statement ?

First of all, we must discuss the idea of "clinical statement" from a semantic perspective.

A clinical observation is a collection of heterogeneous fragments of information, e.g.

AuditoryCanalEczema

EarRegion

AssociatedSignsAndSymptoms

ClinicalSeverity

ClinicalTrend

DateOfOnset

In this document we want to characterise the different kinds of fragments, in order to understand how templates are build and how they can be represented.

The statement is the basic unit of information in a record.

A statement is build in 4 steps, by gradually appending diverse kinds of information, as russian dolls.

We can distinguish at least 4 parts, each part is made of heterogenous fragments.

statement = kernel + status concepts + instantiation information + circumstances of collection

Each part is expressed in information models by attributes and in terminology models by semantic fragments.

The attributes do not correspond exactly to the semantic fragments listed below, i.e. an attribute could involve one or more semantic fragments. The semantic fragments that belong to each part are listed below.

kernel

·  base (i.e. the most significant semantic fragment of a disease, a finding,
a procedure, without their modifiers, e.g. AuditoryCanalEczema)

·  features

·  color

·  consistency

·  flexibility (of devices)

·  reusability (of devices)

·  spatial context

·  particular body location

·  particular body region

·  body side

·  operational context — circumstances, method, procedure ("protocol" in GEHR ?)

·  "challenge" in LOINC, preparation, kind of sample

·  patient position

·  observation aid, device

·  motivation context — purpose, to achieve

·  e.g. palliative, diagnostic

·  e.g. revascularisation (in a surgical procedure)

The kernel is often considered as the realm of terminology. The base is normally listed in coding systems; most coding system includes also precoordinated terminological details (i.e. modifiers, features, circumstances and purpose), but in LOINC, for examples, sometimes these additional details are represented by post-coordination.

kernel = base +terminological details

The kernel alone is an abstract concept: we don't know what is its relation with the patient.

Note that concepts are fractal, thus each semantic fragment (either base or terminological detail) can often be further decomposed in more elementary components (e.g. blood pressure = blood + pressure, AuditoryCanalEczema = eczema + auditory canal, hepatitis = liver + inflammation). We assume that the further decomposition of the semantic fragments is a topic for terminology developers, and it is not relevant for the HL7 templates.

status concepts

changing meaning

·  general prefix for situations (e.g. history of, status post)

·  for conditions (e.g. risk for, ruled out)

·  for procedures – life cycle status (e.g. planned, ordered, cancelled, done, etc)

·  object of information (e.g. relative of patient)

·  uncertainty

minor modifiers

·  clinical trend

·  clinical severity

·  chronicity

·  speed of onset

·  successfulness (of procedures)

The status concepts were introduced by UK-CCC (now Information Authority) and then were extracted and listed separately from the Read codes (now Clinical Terms). They were adopted by the CEN standard on EHCR (ENV 13606-2). They overlap with the mood in HL7.

The status concepts apply to the kernel as a whole, and together they make a "fully described expression", i.e. the terminological part of the statement. It is still a generic concept, but the terminological context is now explicit.

fully described expression = kernel + status concetps

The status concept can transform dramatically the meaning of the expression. In particular, a set of descriptors involve a negation (rule out, absence, not done, …), that is a big problem for terminology representation.

instantiation context

·  effective age (at the time of onset/procedure)

·  date of onset

·  associated date time (for a procedure)

·  repetition count (for a procedure)

·  scale

·  units + numeric value

·  qualitative value

By adding the instantiation information to a fully described expression, we obtain an instantiated expression.

situated expression = fully described expression + instantiation information

Most instantiation information are not considered as a topic for terminologists.

The qualitative values are a problem, bacause there are many ways to split a situated expression into fully described expression + qualitative values, according to the purposes of the representation (in HL7 and LOINC, this problem is called "variable vs value styles").

Vocabulary and structure must be coordinated to achieve an integrated whole and consistency in messaging.

a) Representation using “Value” Name style: name is generic, values are elements of a list

HLA Antigen present = {Aw43, B27, Cw1, Dw12, ...}

b) Representation using “Variable” Name style: list of all names, values are boolean

HLA Aw43 Antigen = {Present, Absent}

HLA B27 Antigen = {Present, Absent}

Interface A (Rheumatologist view) OBX|1|CE|4821-5^HLA-B27^LN|1|G-A203^Present^SNI|

Interface B (Paternity testing) OBX|1|CE|4694-6^HLA-TYPE^LN|1|F-C4327^B27^SNI|

context about collection of information — attribution information

·  collector of information (e.g. physician)

·  date of collection

·  location of collection

·  source of information (e.g. patient)

·  patient ID etc

The attribution information completes the statement and provides important information for the interpretation.

statement = situated expression + attribution information

The circumstances of collection are rarely considered as a topic for terminologists.

Usually this kind of information apply to a set of statements and is recorded only once (see below).

Graphical representation of the clinical statement wrappings

Payload = set of clinical statements

In this section we discuss how a set of clinical statements makes the payload of a document or a message.

A set of statements may show common fragments, that may be mentioned only once in the payload description, e.g.

·  the sample or the device may be the same for a set of observations,

·  the attribution information may be the same for a set of statements in a section

There are at least three ways to assemble multiple statements:

1. siblings = set of similar statements (e.g. battery of electrolytes)
2. associated statements = (subjective) relations between independent statements
3. a predefined sequence of sections/sentences with explicit or implicit statements

1. Siblings

In the first way a set of similar statements may be assembled according to various criteria. This topic is perhaps the most relavant for he Templates SIG. It includes all the batteries, panels, etc.

Note that a statement is usually assembling attributes from the Information Model, whereas the definition of a battery assembles types of statements, e.g. a set of LOINC codes.

·  same request, multiple actions with the same sample /method / device / property

device-dependent / e.g. same sample, multiple measurements (+ calculations)
purpose-dependent / e.g. battery of tests (for screening, orientation, confirmation)
sequence of values in a test / e.g. glucose tolerance test
interdependent values / e.g. set of percentages (electrophoresis, leukocytes count)
unique observation act / e.g. blood pressure, systolic + diastolic

·  same study product (signal, image), multiple observations/measurements

same signal / e.g. measurements on ECG
same image / e.g. description of artifacts and related mesures

·  same encounter (contact, hospital stay, a GP encounter)

same exam / e.g. Stan's vital signs + heart exam

2. Associated statements — clinical context

In the second way a set of (subjective) relations is established between independent statements, e.g.

·  reason for a procedure

·  etiology

·  associated signs & simptoms

·  exacerbating factors

·  alleviating factors

·  observations vs interpretation

The associated statements often have different instantiation and attribution information.

·  same decisional context (purpose / interpretation / process (propedeutic test) / problem)

dynamic context / i.e. links to other statement influencing the interpretation
·  statements needed/used for a decision
·  associated statements

3. Sections

In the third way —— there is a (pre-defined) sequence of sections or paragraphs, e.g. for imaging, something like: preparation, performance of a procedure, results, interpretations, recommandations

The relation among the statements is established by putting different statements into the same container (Section), i.e. in a static way.

·  same container

static context / i.e. chain of nested containers (sections) above a statement

The sections may be filled in with narrative or explicit statements, as described in the definition of battery in HL7 v2.3:

"Obstetrical ultrasound is a battery made up of traditional component measurements and the impression, all of which would be returned as separate results when returned to the requestor. "

"A test involving waveform recording (such as an EKG) can be represented as a battery made up of results of many categories, including digital waveform data, labels and annotations to the data, measurements, and the impression. "

Note that in the last sentence we consider the various measurements as siblings. They make up a section. Other sections are the waveform data, or the impression.

[from HL7 version 2, § 9.6]

history & physical = chief complaint + source + present illness + …

TXA|0001|HP^history & physical|TX^text| …

OBX|1|CE|2000.40^CHIEF COMPLAINT|| ...

OBX|2|ST|2000.01^SOURCE||PATIENT ...

OBX|3|TX|2000.02^PRESENT ILLNESS| |SUDDEN ONSET OF CHEST PAIN. 2 DAYS, PTA ASSOCIATED WITH NAUSEA, VOMITING & SOB. NO RELIEF WITH ANTACIDS …

Transaction unit = envelope + payload

This topic is not within the scope of the templates SIG and is mentioned here only for completeness.

The clinical information is stored as document or database record, or exchanged as message.

The payload is stored or exchanged together with dynamic information, i.e. an envelope.

An envelope is the transaction context. It describes, for example:

people / author / sender, receiver and their organizations
patient matching information / name, sex, age
topic and goal / kind of message / document
handling dates / of creation, update, transmission, arrival

·  envelope for messages = sender, receiver, patient id, kind of payload, etc

·  envelope for documents = directory entry (receiving date, address of the file) + document header

·  document header = author ID, attestation date, access rights, patient ID, kind of payload

Graphical representation of the transaction unit

Analysis of Stan's examples

Here we discuss the examples provided by Stan in his slides (see them in the Annex), using the framework just described above.

We also revise the Stan's framework on terminology models vs information models.

Complex dissections of the statement's terminological part

Independently of the formalism used, the examples below are based on fractal terminology models that allow to generate complex expressions by nesting suitable building blocks.

It is unlikely that such expressions are enumerated in a static nomenclature (i.e. they will be generated by the users each time they need them).

If we really need to represent systematically and process this kind of expressions, the fractal explosion will happen within one or more RIM attributes (i.e. it is considered by the information model as ONE code), it is encapsulated within terminology.

Terminology instance models

In GALEN, we called this construct a "dissection" (borrowed from Cotè in SNOMED II).

Note that we usually remain inside the statement's terminological part (kernel + status concepts).

MAIN cardiac murmur

OCCURS_DURING diastole

HAS_FEATURE high pitch

HAS_FEATURE grade III

HAS_FEATURE blowing sound

IS_ACTED_ON_BY radiation

HAS_DESTINATION part

IS_PART_OF edge

IS_PART_OF manubrium sterni

HAS_SELECTOR left

HAS_SELECTOR lower

HAS_FEATURE volume

HAS_QUANTITY maximum

HAS_LOCATION third intercostal space

[Diastolic Murmur: x]-

->(characteristic)->[high pitched]

->(grade)->[Grade III/VI]

->(characteristic)->[Blowing]

->(characteristic)->[Decrescendo]

->(loudest)->[Third Intercostal region]