Home > Encyclopedia of Public Health > Classification of Disease

The word "classification" can refer to either a "thing" or an "activity." As a thing, a classification is a set of categories (pigeonholes) into which may be placed all the objects in the universe for which it has been designed. As an activity, classification is the process of placing the objects into the categories. This article shall deal only with the first meaning: a framework for organizing information.

Two terms are used in describing a classification: "universe" and "axis." The universe is the totality of the objects that are to be classified—all diseases, all automobiles, all causes of death, all reasons why people encounter the health system, all persons in a given population, and so on. An axis is an attribute or property shared by members of the universe. In health matters, there are many axes—ages of patients, causes of illness, disorders produced, physiological systems disrupted, reasons for encounters, and so on.

Every classification has basic attributes:

1. It deals with a defined universe.
2. It is designed for a specific purpose, which determines its scheme of organization.
3. It groups the objects, using as few groups as consistent with its purpose. In public health and epidemiology, classifications are designed primarily for compilation of statistics.
4. It uses a schema that depends on the logic of its author (which often is a committee).
5. It must accommodate all the objects in its universe and as a result always has one or more categories termed other, which are often called wastebasket categories.

In public health, diseases are only one of the kinds of "objects" that cause death or disability, so it is rare to develop classifications for them alone. The earliest use of classifications in public health was for presenting "causes of death," which of course included injuries. Later, as classifications were expanded to include morbidity as well as mortality, their titles were expanded to "diseases, injuries, and causes of death." In the latest version of the standard classification used in public health, the International Statistical Classification of Diseases (ICD), the universe has become even broader, and the title of the tenth revision (1992) includes Related Health Problems.

In the early twenty-first century, only ICD is in widespread use. Its universe is all individuals who have (or should have) any contact with health services—for prevention, rehabilitation, acute care, long-term care, behavioral problems, investigation of abnormal findings, or for any other reasons. It is not surprising that, to handle this diversity, a number of different axes are found in the classification.

An early need for multiple axes involved trauma. Early mortality statistics showed deaths by external causes. But it was equally valid to tabulate the same deaths according to the injuries sustained. These are two different axes that are used by ICD.

Certain diseases are caused by infectious agents, and one chapter in the classification uses infectious agents as the organizing axis. Other chapters use physiological systems as their organizing axes— respiratory and circulatory, for example. Conflict arises, of course, because a disease such as bacterial pneumonia is both infectious and respiratory. If it is classified both ways, it will almost certainly be counted twice in the statistics.

Largely because of the multiple axis nature of ICD, an extensive set of rules called conventions has been developed to instruct the classifier how to handle these and other conflicting demands. One convention (which the United States has resisted) is the use of dagger and asterisk coding in which a code marked with a dagger (†) indicates the underlying disease and that with an asterisk (*) indicates the manifestation.

It should be clear by this point that classifications in health care are not really classifications of diseases or injuries or causes of disability but are actually classifications of individuals who are of interest to the public health community. As a result, almost never can the classifying be done from a single factor, such as the diagnosis. Rather, the person's other attributes, such as age and other diagnoses, must at least be considered and taken into the classification decision when called for by the conventions.

Retrieval of information for making statistical tabulations or finding individual case records is done by referring to the codes that have been substituted for the labels of the categories; retrieval is code-dependent. It is essential, then, to know the limitations presented by this fact. Retrieval can never produce any more detailed information than the category level—the code is equivalent to the category label. This is a serious limitation when, for example, an epidemic appears and it is a condition that is hidden in a wastebasket category. For example, in the 1970s, Guillain-Barré syndrome was lost in "Polyneuritis and polyradiculitis," where it could not be separated from the other miscellany.

Also, in twenty-first-century information systems, neither the category label nor the category content can be known with certainty, because there is no method for determining the source of the code, that is, the classification from which the code was taken and the version of that classification. For example, code 395 was for Meniere's disease in ICD-6 and ICD-7. With ICD-8 and ICD-9 it was used for diseases of the aortic valve. Especially in ICD 's derivatives, such as the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) in North America, changes are made annually to reflect new diseases and new knowledge. The result may be to add a new disease to an existing category or to move a disease from one category to another. In the mid-1980s, code 279.1 (deficiencies of cell-mediated immunity) was the category to which AIDS (acquired immunodeficiency syndrome) was assigned; after 1986, AIDS was supposed to go to an infectious disease category. Whether it did or not is moot—279.1 looks just the same. This problem, the ambiguity of category labels and codes, will persist until the information systems are modified and standardized to tag each code with an unique source identifier.

In view of the requirement that a classification be designed with a purpose, it is no surprise that ICD is increasingly unsatisfactory. Beginning with the desire simply to tabulate mortality statistics, it has taken on the burden of trying to serve multiple purposes, to accommodate morbidity, health care reimbursement, quality review, epidemiological surveillance, evidence-based medicine, facility planning, public policy, and others. Developers of electronic medical records are expecting it to serve the needs for clinical care for individual patients as well. One classification cannot serve all masters equally well.

It will not be possible to have optimal classifications for public health and epidemiology, as well as for the other legitimate uses of health and health care information, until a simple but major modification of the information system is adopted. That modification is to capture and uniquely and permanently code the specific diagnoses (clinical entities) which go into the classification categories. When that is done, the entities can be distributed into a variety of classifications, each designed optimally for its intended purpose.

VERGIL SLEE

(SEE ALSO: International Classification of Diseases)