Developmental considerations assume major importance in the provision of a classification scheme for children and adolescents, and, indeed, for adults as well (14). Some disorders such as autism have their origin in a specific developmental period, whereas others are frequently associated with developmental problems (Tourette syndrome may be associated with attentional difficulties). At other times, the child’s overall level of development may have a major impact on the ways in which various disorders can be expressed (the child with mental retardation who also exhibits conduct problems). Classification systems must be able to encompass such issues without simultaneously making the disorder so developmentally specific that the utility of the category is compromised.

The developmental approach to classification is used whenever disorders are viewed in the context of the unfolding of basic developmental processes. The use of standard, developmentally based assessment instruments such as tests of intelligence or communication skills exemplifies this approach. In contrast, many categorical and dimensional classification systems rely on assessment of deviant behavior. The use of such an approach is often complicated because issues of how deviant behavior is to be evaluated and how instruments are to be “normed” become quite important, and reliability among examiners can be low. Both the ICD-10 and the DSM-IV systems include some categories in which the definition is fundamentally developmental (mental retardation, articulation disorders), whereas in others the deviant nature of the disorder predominates (autism, schizophrenia of childhood onset).

Theoretical models of psychological disturbance have developed from rather diverse historical traditions; they have considerable value for the individual clinician in understanding and treating children with emotional and behavioral problems. For example, Anna Freud proposed a developmental profile based on and applicable to psychoanalytic assessment of children (15). More phenomenologically based classification systems can be traced to Kraeplin’s delineation of schizophrenia and bipolar disorder (16). In the early “official” classifications, theoretical concerns were reflected in terms such as “schizophrenic reaction of childhood” or “obsessional neurosis.” Classification schemes that are driven by theory are limited because, by their nature, they are based on a set of assumptions and hypotheses not usually generally shared and may give rise to different terms used to describe the same clinical phenomena; for example, a learning theorist may invoke principles of conditioning to explain a child’s phobia, whereas a psychoanalytically oriented theorist may be more concerned with the child’s level of psychosexual organization. Particularly following the work on development of research diagnostic criteria (17), the phenomenologic approach to classification has predominated in the various “official” diagnostic systems. The more robust diagnostic concepts have typically emerged from clinical experience rather than from theory (18). In some instances, a theory has been invoked to account for a given set of phenomena, but it is the set of phenomena rather than the theory that has endured. For example, Langdon Down provided a complex theoretical explanation for children with the condition now known as Down syndrome. His theory, based on obsolete racial stereotypes (mongolism), was incorrect, but his observation of some element of commonality among a large group of children with mental retardation has proved enduring.

For clinicians with pronounced theoretical views, the more phenomenon-based approach can be a source of frustration. It is sometimes incorrectly assumed that in such an approach matters as history, course, and outcome, and, for that matter, etiology and theory are irrelevant to classification. Information on course and outcome may provide important data relative to external validation of diagnostic categories, and information on the development of the disorder may be highly relevant to differential diagnosis regardless of how similar, at one point, two different disorders may appear to be. For example, the syndrome of childhood disintegrative disorder clearly appears to resemble autistic disorder once it is established; however, patterns of early development and outcome differ in these conditions (19). Theoretical views of conditions and mechanisms remain highly relevant for both clinical work and research because they are more likely to generate truly testable hypotheses.

It is often assumed that classification systems are developed to approximate some ideal diagnostic system in which the cause could be directly related to clinical condition. This is not, in fact, the case, in that no single ideal system is waiting to be discovered and that cause need not be included in classification systems (12). Similarly, classification need not reflect a “disease” model (12). Different etiologic factors may result in rather similar conditions, and the same etiologic factor may be associated with a range of clinical conditions. Aspects of intervention may be more directly related to the clinical condition than to the cause. Remedial services for children with mental retardation are, for example, much more likely to be oriented around aspects of developmental level than around the precise origin of the specific mental retardation syndrome. With a few exceptions (reactive attachment or posttraumatic stress disorders in the DSM-IV), etiologic factors are not generally included in official diagnostic systems.

In certain situations and populations, contextual variables such as family, school, or cultural setting pose major complications to diagnosis. The attentional problems of a child whose difficulties arise only as a result of an inappropriate school placement would not, for example, merit a diagnosis of attention deficit disorder. Contextual variables are particularly problematic in disorders of infancy and early childhood in which the infant exerts effects on the parents, who, in turn, exert effects on the child; attributions of causality may be particularly difficult to make (20). A few of the traditional categorical disorders can be readily observed in infants and young children (autism), but generally clinical complaints in this age group are centered around problems that encompass the infant in the context of the family or life situation. These issues are particularly relevant for diagnosis of disorders in infancy— an area that remains somewhat controversial so that, for example, a diagnosis of autism can often be made in very young children (21), although in some cases not all features are exhibited until around age 3 (22). Although research on disorders of infancy is limited (23), it is clear that infants exhibit a tremendous
ability to react, even over relatively short periods, to their environment, and change, rather than stability, is often the rule (24). Clinical problems often relate more to issues of goodness of fit between parents and the infant than to a disorder in the infant (25). As children become slightly older, traditional diagnostic groupings become more readily applicable (26). Issues of developmental level also become important in specifying inclusion and exclusion criteria for diagnostic categories; for example, a diagnosis of pica may be appropriate for a 12-year-old child with profound retardation but is less appropriate for a normally developing 10-month-old infant. These will be important issues to consider for DSM-V and ICD-11.

Cultural differences may also affect diagnostic concepts and practice (27). Clearly, certain sociocultural factors are associated with certain types of problems (e.g., economic disadvantage is associated with conduct and attentional problems), but the meanings of such relationships often remain unclear (28).

It is particularly important that clinicians and researchers alike bear in mind that disorders, rather than children, are classified. This is a source of considerable confusion. Concerns have been raised about the possible effects of labeling children (3), and to some extent these concerns are valid. It is, of course, also the case that having an adequate label for a child’s disorder may be helpful, for example, in securing needed services. Thus, a diagnosis of mental retardation or learning disability may be associated with social stigma or other untoward effects, or it may be associated with more realistic expectations on the part of parents and teachers and provision of potentially more appropriate services. These tensions are also exemplified in the debate between those who advocate broad and encompassing definitions (to maximize clinical and educational service provision) and those who advocate narrow definitions (by defining more homogenous groups of research subjects).

Similar debates arise about aspects of social stigma related to mental illness and behavioral and developmental problems. In this regard, it is always important to refer to the child’s disorder, not to the child as the disorder. The term diagnosis refers both to the notion of assigning a label to a given problem and to the act of evaluation. In important respects, it is the diagnostic process (29) that is the most important of the two. Although diagnostic labels have considerable value, they do not provide information specifically about the individual person, who is unique and uniquely related to intervention.

Validity is the extent to which a classification system does what it purports to do in terms of facilitating communication, intervention, and research. Various types of validity have been identified, for example, face validity (a judgment about whether the description of a category appears to represent the diagnostic construct reasonably), predictive validity (whether some aspect of subsequent course or response to treatment is predicted), and construct validity (whether the category has meaning in terms of what it purports to assess). Generally, such concepts are most useful in measuring the validity of psychometric assessment instruments; their applicability to classification systems is somewhat different. In general, childhood psychiatric disorders have face validity but not necessarily predictive or construct validity (11). The validity of a given diagnostic category can be established on the basis of its association with various features other than those incorporated in the definition (response to treatment, natural history in the absence of treatment, family pattern, biological correlates, and developmental correlates such as age at onset and intelligence quotient).

The sensitivity and specificity of a given categorical diagnostic instrument can be assessed relative to the true presence or absence of a specific disorder. However, a general problem for both categorical and dimensional classification systems is the nature of the standard against which a given category or criteria set is to be judged. Given the usual absence of an unequivocal diagnostic marker for the various conditions, clinical judgment is often used as the standard against which new instruments or definitions are assessed. The issues of “caseness” and diagnostic thresholds are particularly important in the derivation and validation of diagnostic systems (30,31).

In addition to validity, classification systems should exhibit reliability; that is, users in different locations seeing rather similar disorders should be able to agree on the applicability of a specific category or criterion (32). Various kinds of reliability have been identified: interrater, test-retest, and internal consistency. If a given category is not used reliably, it has little value for purposes of communication. Some disorders, almost by definition, have limited test-retest reliability over a relatively short time period (adjustment disorder), whereas other highly stable disorders tend to have better test-retest reliability (profound mental retardation). Sources of unreliability in psychiatric diagnosis include differences in the kinds of information clinicians collect, theoretical biases in the clinician, and differences in internalized diagnostic thresholds, as well as, of course, the true differences that persons with disorders will exhibit at various points over the course of their condition.

High reliability does not guarantee validity. It is possible for a disorder to be reliably defined but have little or no validity. Conversely, a disorder may have validity, but criteria and diagnostic instruments designed to detect its presence may have little or no reliability. In providing diagnostic criteria and descriptions, there is often a tradeoff between the level of detail of a definition and its reliability. What appear to be relatively minor changes in the wording of a criterion can produce major changes in the way in which a diagnosis or diagnostic criterion is applied.