126 Connecting the System to Enhance the Practitioner and Consumer Experience in Healthcare

E. Cummings et al. (Eds.)

© 2018 The authors and IOS Press.
This article is published online with Open Access by IOS Press and distributed under the terms
of the Creative Commons Attribution Non-Commercial License 4.0 (CC BY-NC 4.0).
doi:10.3233/978-1-61499-890-7-126

Representation of ICD-10-AM/ACHI Using

Classification Markup Language (ClaML)
Robert David SMITHa,1, Jeewani Anupama GINIGEa and Christos BOULAMATSISa
a
School of Computing, Engineering and Mathematics, Western Sydney University,
Australia

Abstract. This paper outlines the Australian experience in adopting the

Classification Markup Language (ClaML) to represent two clinical classification
systems: ICD-10-AM (International Statistical Classification of Diseases and
Related Health Problems, 10th Revision, Australian Modification) and ACHI
(Australian Classification of Health Interventions). The primary goal of this process
is to share classification data efficiently with relevant parties in a consistent format.
This paper outlines extensions implemented in ClaML to facilitate the
representation of the above classifications, with validation of the resulting ClaML
files verified through XML stylesheet transformation (XSLT) to render the data on
a browser, in the same format of printed ICD-10-AM and ACHI books.

Keywords. ClaML, XML, Clinical Classification, ICD, ICD-10-AM, ACHI

Introduction

ICD-10-AM (International Statistical Classification of Diseases and Related Health

Problems, 10th Revision, Australian Modification) and ACHI (Australian Classification
of Health Interventions) are two of the clinical classification systems used in acute care
setting, in public and private hospitals in Australia. These classification systems are used
to code acute care episodes covering diagnosis and healthcare interventions. ICD-10-
AM is the country specific version of the international standard ICD-101 and is used for
coding diagnosed diseases. ACHI is a unique Australian classification for coding
healthcare interventions, for which currently there is no approved equivalent
international classification. The ICD-10-AM and ACHI coded healthcare data, is used
for a variety of purposes such as: statistical tracking across all regions, territories and
states, funding, insurance claim processing and research.
In Australia, biennial releases of ICD-10-AM and ACHI are disseminated to various
classification users as printed hardcopy books by IHPA (Independent Hospital Pricing
Authority)2 or through the software systems used within hospitals for clinical coding such
as TurboCoder3, Codefinder4 and Codexpert5. Currently, the National Centre for
Classification in Health (NCCH) at the University of Sydney leads the Australian
Consortium for Classification Development (ACCD), in collaboration with Western
Sydney University (WSU) in the development of the ICD-10-AM and ACHI, on behalf
of IHPA. The current method of sharing the latest classification systems with the vendor
organisations, is through CSV files, database scripts or PDF files. These formats are not
easy to be parsed by systems, in an automated manner, which uses this clinical

1
Corresponding Author.
 R.D. Smith et al. / Representation of ICD-10-AM/ACHI Using ClaML 127

classification data. Due to ongoing evolving nature of the classification systems,

classification data is heavily subject to change. Further, the sharing of ICD-10-AM and
ACHI classifications as database scripts leads to misinterpretation of tables and data
fields that leads to errors in the classification system software. Therefore, based on
feedback from the vendors, the clinical classification developers have identified a
requirement for providing a standard that would facilitate the proper structuring of
classification data, which is machine consumable and human readable.
ClaML (Classification Markup Language)6, is an XML based mark-up language
recommended by WHO-FIC (World Health Organization Family of International
Classifications Network) for the maintenance, publishing and sharing of clinical
classifications. The ClaML schema was developed by DIMDI (Deutsches Institut für
Medizinische Dokumentation und Information)6 and approved as an ISO standard7. By
definition, ClaML is suitable for the presentation of a range of WHO classifications such
as ICD (International Statistical Classification of Diseases and Related Health Problems),
ICF (International Classification of Functioning, Disability and Health), and ICHI
(International Classification of Health Interventions) the ACHI equivalent international
standard, currently under development. The developers of the ClaML specification have
provided an example of the representation of the base ICD-10 classification with ClaML8.
This paper presents details on how ICD-10-AM and ACHI, the two most commonly
used clinical classification systems in Australia, are represented using ClaML. The paper
highlights the various extensions suggested to be used in ClaML when representing the
ICD-10-AM and ACHI classification systems. By using ClaML for representing ICD-
10-AM and ACHI, better opportunities and benefits are expected to be seen from the
self-describing nature of XML, and its suitability for complex hierarchically structured
data.

1. Methodology

This three-phased project identifies the usage of ClaML to capture the complex data
associated with ICD-10-AM and ACHI.
The first phase initiated a focus group involving the clinical classification experts
and IT specialists to understand the structures of ICD-10-AM, ACHI and ClaML. Two
subgroups within this focus group took the initiative to understand how ICD-10-AM and
ACHI classification system structures (e.g. chapters, blocks, code, etc.) can be mapped
to the tags of ClaML (e.g. classes, modifiers, rubrics, etc.). This exercise was carried out
through workshop mode over three months. As part of the workshop activities, relatively
simple chapters were manually entered, into a prototype XML that follows the ClaML
standard. The purpose was to simulate the migration of ICD-10-AM and ACHI
classification data and evaluate any potential issues that would occur during the transition
to ClaML. In this exercise, ClaML 2.0 was used as there is more material available to
utilise, in comparison to the newer ClaML 3.0 standard.
Once the nuances of ICD-10-AM and ACHI to ClaML were determined, scripts
were written to automatically generate the ClaML version of these two classifications
from the databases. Since these classification systems are large, it required additional
verification to ensure that the representations were accurate. To do that, XSLT
(eXtensible Stylesheet Language) files were created to consume the ClaML data, and
format the classification data, similar to the format it is represented in printed books (see
figure 1). This format is more familiar for human-users which allows for visual
128 R.D. Smith et al. / Representation of ICD-10-AM/ACHI Using ClaML

verification and quality assurance of the generated data. The XSLT contains the
instruction on how the ClaML should be parsed, in generating HTML content,
representing the full ICD-10-AM and ACHI classifications on a web page. The XSLTs
use multiple templates to format each part, based on pre-existing formatting rules.
In the third phase, the clinical classification specialists checked the accuracy of
representation of this formatted data, through a web-interface. This process revealed
certain missing and inaccurate representation of data elements, which led to revising of
the ClaML representations as well as the XSLTs. At the time of this writing, the
classification specialists are checking the generated content through the said web-based
system. Once the classification specialists are satisfied with the ClaML representation,
ClaML files for ICD-10-AM and ACHI will be distributed to vendor organisations to
test those against their products. The feedback received from this process also, will be
used for the further refinement of ClaML representation of ICD-10-AM and ACHI.

Figure 1. ClaML data rendered on a web page after applying XSLT.

1.1. Business Drivers for the Use of ClaML

The reasons for the use of ClaML and XML are that there was no current standard in
place to share the classification to those that needed access to the classification data, such
as the vendors of ACCD, one of the only forms of communicating the complete
classification data to the vendors was via a database export that had to be done by IT
staff which is not good as the export takes time to prepare and requires on-going IT
support for the exports if they need to be re-exported or updated.

1.2. ClaML Export versus Existing Methods

With the ClaML export, after any changes are made to the classification an export will
be able to be done by any staff member of ACCD. With the database export, IT support
would have to be contacted to re-export the classification by running 23 SQL scripts on
the production SQL server to generate the Vendor database this would then be copied to
removable media and given to the requesting staff. Also to note, ClaML is not used for
modification of the classification which is done within an intranet application on the
ACCD web site and publication for human use is provided by books printed by IHPA
and software applications created by the vendors of ACCD. As a result of our efforts we
find that there is no unified methodology for modification, distribution and publication
for human use of the classification as each needs different levels or types of access to the
classification.
 R.D. Smith et al. / Representation of ICD-10-AM/ACHI Using ClaML 129

2. Structure and Complexity of Clinical Classification Systems

2.1. Structure of ICD-10-AM and ACHI

In ICD-10-AM, there are 22 chapters and each chapter represents diseases or injuries
associated with a major body system such as: respiratory, gastrointestinal or peripheral
nervous system. These chapters in total have 16,875 exported entities. The chapters (e.g.
Chapter 2 - Neoplasms) are organised in a hierarchical manner, where the top most level
is represented by the chapter name: Chapter 2 - Neoplasms, followed by sub-chapters.
The sub-chapters (e.g. Malignant Neoplasms) are further specialised areas of the chapter.
Under the sub-chapters there are sections (e.g. malignant neoplasm of lip, oral cavity and
pharynx) which group similar codes under a specified afflicted body part. Categories
(e.g. C00 – malignant neoplasm of lip), defined under sections, are 3 character codes
which serves as a grouping of codes under a very specific area. Assignable codes exist
under categories, and may be in 2 levels: four digit codes (e.g. C00.0) or five digit codes
(e.g. C88.00).
In ACHI, there are 20 chapters and each chapter deals with a different body system
similar to ICD-10-AM. In ACHI, there are 8,644 exported entities. Each chapter
contains: anatomical site axis, procedural type axis, blocks and codes. The anatomical
site axis is the area of the body (e.g. Peripheral nervous system) and contains the
procedural type axis, which describes the types of procedure that could be performed on
that body system. Under the procedural type axis, there exists blocks (e.g. 370 -
Examination procedures on nose) that further divides the procedural type axis into the
various sub-types of the procedure. Furthermore, blocks contain codes (e.g. 41764-00 -
Nasendoscopy), at the lowest level that represents a procedure performed on a patient.
Both ICD-10-AM and ACHI have two sets of printed books, one representing the
alphabetic index and the second that is called the tabular list that shows the full details
of the code listing. When sharing classification data with vendors, it is important to
provide data associated with the alphabetical indices and tabular lists, for the accurate
re-production of electronic versions of these books.

2.2. Structure of ClaML and Mappings into ICD-10-AM and ACHI

Table 1 illustrates how each ClaML element is used in representation of entities

associated with ICD-10-AM and ACHI. Additional information exists in the ICD-10-
AM classification that is not in ICD-10 (base version upon which ICD-10-AM is created).
Table 1. Illustrating how ICD-10-AM, ACHI and index elements relate to ClaML elements.
ClaML ICD-10-AM Tabular ACHI Tabular ICD-10-AM/ACHI Index
Class Chapter , sub-section, code Chapter , sub-section, code N/A
Rubric Attributes Attributes Attributes
Label Attribute text content Attribute text content Attribute text content
Meta Link to other data Link to other data N/A
SubClass Child code links Child code links N/A
SuperClass Parent code link Parent code link N/A
Modifier Attribute, to be moved N/A N/A
ModifierClass Attribute, to be moved N/A N/A
Index N/A N/A Index with type set
Term N/A N/A Term text
Reference N/A N/A Term to code (Class) link
TableItem N/A N/A Multiple term to code link.
IndexKinds N/A N/A List of index types
Tables N/A N/A List of table types
130 R.D. Smith et al. / Representation of ICD-10-AM/ACHI Using ClaML

2.3. Compatibility of ClaML with Alphabetic Index

The ClaML structure allows for the representation of ICD-10-AM tabular content, with
minor changes to the core structure. However the indices of both ICD-10-AM and ACHI
could not fit into ClaML as is, due to the use of parent and child links in ClaML as the
terms are not unique across the classifications. To provide a structure for the index,
ClaML had to be extended (see figure 2).

Figure 2. ClaML structure representing index Figure 3. Tabular XML Structure same as ClaML
elements. The green elements as extension to Standard, red not used in ACHI.
ClaML for index representation.

2.4. ICD-10-AM/ACHI Attributes vs. ClaML Rubrics

ClaML defines rubrics as a term to describe the extra information associated with a code,
such as ‘see also’ (which references to another closely associated code) and ‘definitions’.
The attributes in ICD-10-AM and ACHI predominantly consist of plain text, and
sometimes HTML (HyperText Markup Language) formatted content. The term attribute
in ICD-10-AM and ACHI is similar to what ICD-10 (the base version) refers to as a
rubric. During the process of translating ICD-10-AM to ClaML, the original ICD-10
ClaML rubric terminologies were used. However, the attributes in ICD-10-AM did not
perfectly correspond to the ICD-10 ClaML rubric types. ICD-10-AM has striven to
differentiate attributes from ICD-10 rubrics, providing more specificity. Thus, less
styling tags (resembling HTML) is required in ICD-10-AM compared to ICD-10’s
representation of ClaML. For instance, in ICD-10-AM, there exists ‘Code Also’, ‘Code
First’, ‘Inclusion’, ‘See’ and ‘Use Additional’ attributes. For ICD-10, there is only
‘coding-hint’, which corresponds to all five of these attributes. This poses an issue as
there are different styles used to display each of these attributes. For each attribute a
mapping to an existing ClaML rubric was created as identified in table 2. The varied
attribute styles will cause issues when attempting to use the XML as there would be no
means to instruct how to process each attribute.
 R.D. Smith et al. / Representation of ICD-10-AM/ACHI Using ClaML 131

Table 2. Illustrating how ICD-10-AM and ACHI attributes map to existing ClaML rubrics.
ClaML ICD-10-AM Attribute ClaML ACHI Attribute
coding-hint code also, code first, coding-hint codeAlso, codeFirst,
inclusional, see, use codeAlsoWhenPref,
additional inclusional, see
text text definition definition
footnote -- not used -- exclusion excludes
definition definition inclusion Includes
introduction introduction note note, styledNote
modifierlink see meta tag type ACS acsNo
note note, styled note preferredLong long code name
exclusion excludes preferred code name
inclusion includes
preferredLong long code name
preferred code name
meta tag type range range
meta tag type ACS ACS No
modifiers code subdivision

3. Conclusion

After several workshops and investigations into ClaML, it was found that it is highly
compatible to represent ICD-10-AM and ACHI data. This paper explored the use of the
ClaML ISO standard for the representation of ICD-10-AM and ACHI. However, this
exploratory work indicated that there is no one-to-one map between ClaML structures
and ICD-10-AM/ACHI. Several parts of ICD-10-AM classification: codes, code
attributes and term attributes were successfully mapped to the ClaML structure elements
such as Classes, Rubrics, and Modifiers. Conversely, the index representations of ICD-
10-AM and ACHI were not compatible with ClaML. For these outlined reasons,
extensions were required to facilitate the transition of ICD-10-AM and ACHI to ClaML
standard. In addition, a set of transformation files (XSLT) based on a set of classification
data rendering rules were developed, and when combined with the ClaML files, allowed
the generation of content to be published on a web browser in a readable manner.

References

[1] Icd-10 online versions. 2018. [accessed 2018 13-03-2018].

[Link]
[2] Ar-drg classification system product sales. 2014. [accessed 2018 02-03-2018].
[Link]
[3] Turbocoder. [accessed 2018 02-03-2018]. [Link]
[4] Health information systems. 2018. [accessed 2018 02-03-2018]. [Link]
au/applications/coding-and-reimbursement/.
[5] Codexpert™ - electronic coding reference application. 2017. [accessed 2018 02-03-2018].
[Link]
[6] Das dimdi - medizinwissen online. 2018. [accessed 2018 02-03-2018].
[Link]
[7] Health informatics -- syntax to represent the content of healthcare classification systems -- classification
markup language (claml). 2013. [accessed 2018 02-03-2018].
[8] Downloadcenter klassifikationen, terminologien und standards. 2015. [accessed 2018 02-03-2018].
[Link]