Keywords: Clinical Terminologies, Coding, Classification, Selection, Diagnoses

Citing this page:

Jones, P. (2000) Introduction to Coding and Classification in Care,
, Accessed

Introduction

People intuitively deploy various strategies to help make information handling easier for themselves. Data compression is most obvious in the saying - "a picture paints a thousand words." Our brains must make sense of a seemingly chaotic world. To do so data reduction must occur, some sort of filter or categorization is needed. Diagnoses function as a means to compress information.

The majority of lay people know what specific diagnoses entail: treatment; need for hospitalisation; recovery; time off work; after effects; necessary nursing care. They are also acutely aware of the stigma associated with certain diagnoses. In order to manipulate data, it is necessary to have it in a form that the computer can process. This is usually achieved via codifying the data to be captured.

Contrary to expectations coding has a long history; Black wrote these words in 1968!

'A good deal of effort is being put into the coding of clinical observations, and into the technical aspects of record linkage; it is likely that before long a detailed medical history of anyone who has come under medical observation may be obtainable from a national record store. When this happens, a living clinical epidiology will be possible, based on verificable morbidity statistics.'

Johnson's (1987) approach, was not widely adopted.

"From 1970 to 1981 I recorded digitally every symptom, sign, diagnosis, and treatment that passed through my hands on 103,018 eighty column punched cards." Johnson took the American ZIP code system - which is geographically based - and produced a medical version; mapping the body together with associated symptoms. The body was divided as follows: 10 = head; 20 = chest; 30 = arm; 40 = trunk; 50 = leg. The code is extended to three segments to produce meaningful phrases. For example; "pains in the chest" becomes 00-10-20 and "cough with pain in the chest" - 21-10-20 and sore throat with earache" - 22-10-13. These are derived from 10 = head; 10 = eye; 12 = nose; 13 = ear and so on.

Since the introduction of computers into health care, diagnoses have held the center stage of clinical information systems. This is not to say that diagnoses have no role to play in management information systems, just that they are used in a different way, as per the section on groupings.

Informatician meets clinician

Coding and classification as informatics activities are obviously important, but the background to this work and the status of diagnostic systems bears (this brief) review. Especially since in some areas the limits and social effects of diagnoses are felt.

One example is the status of schizophrenia as a diagnosis. The 'Diagnostic and Statistical Manual of Mental Disorders' (DSM), criteria raises several issues that are still subject to debate as explained by Carson (1996). These criteria must affect research whenever DSM (or diagnosis) is used:

‘[Psychiatry] resembles a field that has yet to come to grips with the natural and dynamic processes producing the (disordered) phenomena observed. The explanation of the disorder resides in the class or category to which the attendant observations are allocated, as in the lists of symptoms by which 19th-century medical students were taught to identify the various diseases then recognized. .... wherein the ultimate nature of a phenomenon is assumed to reside in the carefully discerned properties it shares with other phenomena (i.e., in its accurate categorization).’

Lackoff (1987) also challenges the traditional ‘category bucket’. Differences arise when classification schemes are used in research, and the nature of symptoms such as delusions and hallucinations be subject to research in their own right, we should not regard them merely as a diagnostic indicator. Carson (1996) also acknowledges the difficulty of ongoing revision of classifications such as DSM.

Strauss (1994) sums up succinctly:

‘Psychiatry is a psychobiosocial field, but parts of the field focusing on psychological and social factors need to progress. We need to develop a dynamic descriptive psychiatry to complement the dynamic models of modern neuroscience.’ p.42

Defining 'terminology', 'nomenclature', 'classification'

At this point it may help some visitors to review some important definitions, Nowlan (1993) provides the following:

Codes for diagnoses can help in many ways, including:

• Codes are compact
• Codes offer a form of shorthand
  
• Codes can form a nomenclature
  
• Codes assist (or potentially hamper!) information retrieval
  
• Codes help ease storage problems
  
• Codes assist in automated processing, and reporting
  
• Coding promotes critical thought in the target domain

SEVERAL CLASSIFICATION SCHEMES (MEDICAL & PAMs)

International Classification of Diseases 10 (ICD-10)

The ICD is the oldest, most frequently cited in the literature. In order that the World Health Organisation can monitor the incidence of disease; injuries; and causes of death across international boundaries, standardisation is required. ICD is a standard originating in 1977. ICD 10 the latest revision came into use in April 1994. It extends - not surprisingly - ICD 9 which is limited to diseases alone. ICD 9 is still used by many countries.

In the ICD system there are chapters, divided by major anatomical systems or aetiology. Coding employs a three digit number with an optional fourth digit being separated by a decimal point. For example: 564.2 post gastric surgery syndrome.

Perhaps contrary to expectations, ICD codes are often described as not being unique, therefore several concepts may code to the same number. In addition to the above example 564.2 may also refer to: dumping syndrome; post vagotomy syndrome; post gastrectomy syndrome.

This general coverage is fine as far as the epidemiologist is concerned, but extending ICD to cover purely clinical applications creates problems. More specific coding is needed. For example, although the above terms are similar in their anatomical relation, the diagnostic processes, treatment and subsequent care may differ radically. ICD cannot provide this detail.

The validity of the information derived from the ICD is dependent upon diagnostic procedures and these change according to trends in medical practice and socio-epidemiological factors. Such evolutionary change complicates further the ability to cross map ICD with other schemes, notably READ.

Systematized Nomenclature of Medicine - Clinical Terms (SNOMED-CT)

Snomed is a division of the College of American Pathologists (CAP) dedicated to the development and support of (Systematized Nomenclature of Medicine (SNOMED). SNOMED adopts a multiaxial approach in an effort to overcome problems associated with enumerative classifications (College of Medical Pathologists, 1977) cited by Nowlan (1993). Produced by the College of American Pathologists (CAP), SNOMED CT (Systematized Nomenclature of Medicine--Clinical Terms) was formed by the convergence of SNOMED RT® and the United Kingdom's Clinical Terms Version 3 (formerly known as the Read Codes).
The key axes used comprise:

Topography (T)     Function (F)          

Morphology (M)     Disease (D)           

Etiology (cause)   Procedure (P)

The SNOMED terminology is reader dependent. Nowlan interprets (in italics) the terms of this compositional approach as:

granuloma in lung caused by M. Tuberculosis together with Fever

    T                E                 F                    D          

Lung      +   M. Tuberculosis  +   Fever      +    Tuberculosis        

 T-28000           E-2001           F-03003              D-0188        

tuberculosis in lung caused by M. Tuberculosis together with Fever

granuloma in lung caused by M. Tuberculosis together with Fever which is tuberculosis (p.32) (1)

Within SNOMED there is no mechanism that organises the classification of terms and the way that terms across axes are used. The combination of terms could be analogous to a 'fruit machine' - the wheels flying and stopping at any combination.

'It is crucial to appreciate that SNOMED is not a general classification of the things it is sensible to say in medicine. It is not in anyway capable of preventing the creation of medically meaningless utterances.' Ibid. p.33

NHS Clinical Terms (formerly READ Classification)

The READ clinical classification was designed by Dr James Read, specifically for use by clinicians (not just epidemiologists). The codes form a structured hierarchy of medical terms. Version 3 (Dec 1995) has five hierarchical levels, but each can be coded using upper and lower case letters. This gives 58 variations within a level and offers - 656,356,768 possibilities within the READ coding system. The 58 variations stem from use of the characters 0-9, A-Z, a-z are used although "i", "o", "I" and "O" are excluded to avoid misinterpretation.

THE READ CODES - AN EXAMPLE

Incorporation of synonyms allows use of preferred terms without information loss. The synonym list is cross-referenced to enable translation or mapping within the READ codes and other coding schemes e.g., ICD-10. Given the ample capacity it is possible to incorporate other classification schemes such as nursing. (Source: NHS Information Management Group - now the NHS-Information Authority.)

When search procedures are employed attention must be paid to phenomena of combinatorial-explosion. In its present form READ suffers from this phenomena. For example, if we wished to add severity values such as:

slight moderate severe null

to a list of 10,000 diseases, then the number of terms to represent these is multiplied to 40,000 (assuming we may also wish to say nothing on severity). This enumerative phenomena happens each time new values are added, and leads to immense picking lists with hundreds of rubrics.

More recently an agreement between the College of American Pathologists (CAP) and the United Kingdom’s National Health Service Executive will lead to the combination of the READ Version 3 and SNOMED® RT - the CAP’s Systematized Nomenclature of Medicine, under the heading of SNOMED Clinical Terms. A release is due in April 2003.

UNIFIED MEDICAL LANGUAGE SYSTEM (UMLS)

PURPOSE AND FOCUS

The UMLS project is a long-term research and development effort by the National Library of Medicine in the USA. UMLS is designed to facilitate the retrieval and integration of information from multiple machine-readable biomedical information sources. These include:

• clinical records
• descriptions of the biomedical literature
• directories of people and organisations
• factual data banks
• and knowledge based systems

UMLS focuses on the variety of available vocabularies, and classifications used in different sources, and the barriers these pose to the use of machine-readable sources by health care professionals and biomedical researchers. The continued diversity of terminological approaches also hinders the development of effective search interfaces and tools that can assist users.

UMLS COMPONENTS

The 11th Edition (2000) of the Metathesaurus includes about 730,000 concepts and 1.5 million concept names in different source vocabularies, an increase of approximately 16% from the previous edition. The components of UMLS are now so numerous that any attempt at summary here would be difficult. Suffice to say here that the main constituents are three UMLS knowledge sources:

1. The Metathesaurus (Section 2) contains semantic information about biomedical concepts, their various names, and the relationships among them.
2. The Semantic Network (Section 3) is a network of the general categories or semantic types to which all concepts in the Metathesaurus have been assigned.
3. The SPECIALIST lexicon (Section 4 ) contains syntactic information about biomedical terms and will eventually cover the majority of component terms in the concept names present in the Metathesaurus.

Further information is obtainable at: Nat. Inst. Health

UMLS aims to aid and promote the development of products that can assist in establishing a conceptual link between the user's question and relevant information held in machine readable form.

SEMANTIC NETS

UMLS is an enormous undertaking, the scale of the project is similar in extent to READ and SNOMED, but the methods must differ with UMLS's drive towards 'automated intelligence'. The representational form to facilitate this is the semantic network formalism.

It has long been recognised that some of the shortcomings of other systems may be addressed via semantic networks. It needs to be noted that there is a range of discrete purposes and each approach may(?) boast respective strengths and weaknesses.

Semantic nets have their foundation in research in language expressions. They are employed to convey meaning between concepts and the relationships between them. For their application in expert systems, Jackson (1986) prefers the description of 'associative nets'. This term:

'being more neutral with regard to what the network will be used for.' p.53

Woods (1975) definition of semantic nets still applies today:

'A semantic network attempts to combine in a single mechanism the ability to not only to store factual knowledge but also to model the associative connections exhibited by humans which make certain items of information accessible from certain others. It is possible presumably to model these two aspects with two separate mechanisms such as for example, a list of the facts expressed in the predicate calculus or some such representation, together with an index of associative connectives which link facts together.'

A semantic network is a fusion of these two requirements. Within the network are pointers relating classes of facts or objects to others. Many authors, including Woods, point out that the application of semantic networks in artificial intelligence remains unproved. This is because such is the complexity of human affairs - be that language, politics, law, medicine and any others we care to mention - that no single representation may be possible. This debate will continue.

NURSING TERMINOLOGY APPROACHES

ICNP & North American Nursing Diagnosis Association (NANDA)

An ongoing initiative organised under the auspices of International Council of Nurses.

The ICNP® (from whose site these notes were obtained) is a classification of nursing phenomena, interventions and outcomes. It serves as a unifying framework into which existing nursing vocabularies and classifications can be cross-mapped to enable comparison of nursing data. The development of the ICNP® paralleled profound changes worldwide in health care, nursing practice and the definition of nurses roles. As such it serves as a stepping stone into the future.

The initial objectives of the ICNP®, established by ICN and outlined it the Alpha version, were reviewed during the developmental of the Beta Version. These objectives continue to direct the aims of the ICNP® Programme and include:
 

• To establish a common language for describing nursing practice in order to improve communication among nurses, and between nurses and others;
• To describe the nursing care of people (individuals, families and communities) in a variety of settings both institutional and non-institutional;
• To enable comparison of nursing data across clinical populations, settings, geographic areas and time;
• To demonstrate or project trends in the provision of nursing treatments and care and the allocation of resources to patients according to their needs based on nursing diagnoses;
• To stimulate nursing research through links to data available in nursing information systems and health information systems;
• To provide data about nursing practice in order to influence health policy-making.

WHAT IS THE ICNP®?

The ICNP® is a classification of: nursing phenomena nursing actions nursing outcomes

The ICNP^® can help guide the process of nursing diagnosis, which is composed of concepts contained in the Classification of Phenomenon axes.

Grouping: (DRG)

Coding and classifications are useful - up to a point. Various interested parties need to grasp the 'bigger picture', so aggregation of patients is needed. This is facilitated by grouping, and was driven initially by economic considerations - billing for care in the USA. One of the first groupings is the DIAGNOSTIC RELATED GROUP (DRG).

• DRG stands for Diagnosis Related Group, and refers to a patient classification system that provides a way of describing the types of patients a hospital treats (its case mix).
• DRGs were developed by a group of researchers at Yale University in the late 60s as a tool to help clinicians and hospitals monitor quality of care and utilization of services. They proved to be so useful that in 1983 they became the system used by Medicare in the United States to pay hospitals.
• Briefly, the DRGs work by taking the 10,000+ ICD-9-CM codes and grouping these into a more manageable number of meaningful patient categories (close to 500 now). Patients within each category are similar clinically and in terms of resource use. (The DRG grouper uses administrative data to group patients.) University Manitoba (1997)

Grouping: (HRG)

A further grouping is the Health Resource Group:

‘HRGs are a method of categorising NHS acute in-patients and A&E patients into a manageable number of groups (565) which contain cases that are similar, both clinically and in the resources they are expected to use. Work is in progress to extend the coverage of HRGs to other areas, including community healthcare and outpatients. Work is also in progress to finalise Health Benefit Groups (HBGs). These will categorise the population in terms of its need for healthcare and, given appropriate care, its ability to benefit. HRGs are known as ‘iso-resource’ groups for this reason.’ source - cams.co.uk (no longer available)

Grouping: Case-Mix Groups(CMG)

‘Casemix is the term used to define various ways of describing the mix of different types of people, patients and/or treatments in a particular area, i.e. a town, hospital or community. In other words, the mix of cases.

The aim of casemix is to help health professionals to improve the delivery of healthcare and thereby the health of the local population, by using information in meaningful and useful ways.’ Source: NHS-IA Casemix Programme

Stock items are stock items (in health also), but clinical problems are not necessarily as simple to capture as widgets, especially if you are not sure from the outset of just what it is you want the (computer) system to do. COMPLEXITY
The health record is explored - what does it contain? What form does it take? What options are there for information management innovators?

© Peter Jones 2000

References

Coding and classification links

Black, D.A.K. (1968) The Logic of Medicine, Oliver & Boyd, Edinburgh, 71.

Carson, R.C. (1996) Aristotle, Galileo, and the DSM Taxonomy: The Case of Schizophrenia, Jour. of Consulting & Clinical Psychology,64,6,1133-39.

Jackson, P. (1986) Introduction to Expert Systems, Wokingham, Addison-Wesley, 53.

Johnson, B. (1987) Health Code, The Guardian, 23 July, 16 (see also (1990) Journal of Health Care Computing).

Lackoff, G. (1987) Women, Fire and Dangerous Things, Chicago University Press, Chicago.

Neisser, U. (1967) Cognitive psychology, NY, Appleton-Century-Crofts.

Nowlan, A. (1993) Structured Methods of Information Management For Medical Records, University of Manchester.

SAGNIS (1993) Report of Nursing Terms Scoping Project, NHS IMG CCC, Loughborough.

Strauss, J.S. (1994) Is Biological Psychiatry Building on an Adequate Base? Schizophrenia from Mind to Molecule, Andreasen, N. (Ed.) AP Press Inc., Washington.

Woods, W.A. (1975) Foundations for semantic networks, Representation and Understanding, Bobrow Collins (Eds),London, Academic Press, 44.

top : back : home