Empirical approaches to the classification of schizophrenia

Many of the clinical concepts of schizophrenia summarized so far, selected from a range that could be broadened to include dozens more authors, have been unsatisfactory in several ways:

• their symptomatic and syndromic components overlap but are not identical;

• they cannot be stated in precisely reproducible terms;

• the weights given to individual symptoms when formulating a diagnosis are not specified but left to clinical interpretation;

• other criteria, such as course, are of uncertain value for classification; and

• until recently there has been little convincing evidence for specific pathologies or physical causes in the large majority of cases.

Testing clinical concepts

New syndromes can be derived from symptom lists by the application of statistical techniques. There is a long line of such studies, many of them initiated in the 1950s and 1960s in order to overcome the unreliability then being demonstrated in day-to-day clinical diagnosis (e.g. Lorr 1966). They were successful in achieving reasonable reliability, but the usefulness of the statistical syndromes was not demonstrated, except in the sense that the factors often looked very similar to the diagnoses they tried to leave behind.

Kendell (1989) addressed the problem in a more practical way. He suggested utilizing statistical methods to refine syndromes, which could then be tested against outcome and used to generate or test biological hypotheses. His own studies (Kendell & Brockington 1980) did not demonstrate a point of statistical discontinuity between schizophrenic and affective psychoses.

This could be due in part to fluctuation over time; cross-sections cannot display the clinical picture that eventually emerges. In addition, the hierarchies that run through psychiatric classification ensuring that disorders higher up, such as dementia, schizophrenia and bipolar disorder, which tend to be co-morbid with symptoms of disorders lower down, must also be considered.

A high proportion of people with schizophrenia in the International Pilot Study of Schizophrenia (WHO 1973) would have been classified by the PSE computer program known as CATEGO as having affective disorders if symptoms discriminating for schizophrenia had been left out (Wing et al. 1974).

Moreover, some at least of the affective symptoms so common in acute schizophrenia must be reactive to the stress of the primary experiences.

From concepts to classification

The clinical concepts, especially those of Kraepelin and Bleuler, can be recognized as early attempts to classify schizophrenia, but there were differences in their approach to defining schizophrenia as an entity. While Kraepelin emphasized the value of onset and course as well as descriptions in diagnosis and classification of schizophrenia, Bleuler and Schneider preferred a crosssectional approach based on patients' current mental state, emphasizing fundamental and characteristic distinguishing symptoms. Recent recognizable international classification systems such as DSM and ICD have attempted syntheses of these concepts to foster international consensus.