Abstract
Assen Jablensky comprehensively reviews the obstacles to developing sound methods that allow us to better identify people at risk of schizophrenia. When you consider how long that improvement in identification has been stalled, this reads as pessimism. This pessimism is a consequence of a particular way of thinking, not only about schizophrenia but also about how epidemiological knowledge is to be converted into preventive action.
The examples of cholera and pellagra instruct us on how preventive action may precede full understanding. However, these examples mislead us because they draw on illnesses where a single factor is necessary and (almost) sufficient and where an action that broke a single link in a causal chain effectively ended an epidemic. It is no accident that cholera and pellagra are ancient examples. Modern epidemics and, more importantly, endemic illnesses (like psychoses) have not yielded to those techniques that have made 20th century medicine triumphant. The better examples for our thinking in epidemiology come from modern preoccupations. Cardiovascular disease and suicide have required a different line of epidemiological investigation. I will argue that our successes and failures here show us how we may progress even when we cannot identify actual or potential cases.
Second, I will argue that categorical diagnosis has trapped us in a conceptual dead end and there may be more productive ways to conceive psychotic ill nesses, including schizophrenia.
A different epidemiological and preventive approach to psychotic illness?
Identifying the people at high risk for the development of psychosis confronts the problem of a condition with a low base rate and no identifiable causal agent or pathognomic sign. It is similar to identifying the people who are at high risk for suicide [1]. Where base rates are low, even tests of unimaginably high sensitivity and specificity have problematic misclassification rates. For a concept as muddy as schizophrenia an adequate level of correct classification cannot be attained. If we keep trying to convince ourselves that it is just a matter of getting better predictors and better detection, we will continue to beat our heads against this brick wall.
Genetic risk, which is our largest single predictor in psychosis, is very junior in prediction to a jumble of difficult-to-measure environmental risk factors [2]. Risk factors that prefigure the development of psychosis will be shared with many people who do not become psychotic. Each individual who becomes psychotic will have combinations of risk factors that are not shared with others. Major psychotic disorders, like suicide, reflect a distribution of liability. At some point in life, people will move beyond the threshold which gives us a diagnosis. There will be no identifiable causal chain of the disease, no reliable precursors, no reliable premorbid identification of the individuals, no germ of schizophrenia. We will have only the enumeration of risk factors of varying power and individuals of varying risk. This paralyses the conventional epidemiological approach of locating and researching the affected individual.
There is an alternative to identifying and treating the vulnerability of threatened individuals. The alternative approach was put forward by the late Geoffrey Rose [3] but is yet to make inroads into psychiatric epidemiology. He proposed that, for conditions such as cardiovascular disease (and, by extension, psychotic disorder), we should consider that the whole population shares some liability and some (those who are ill) are over the threshold at which illness is manifest. We should attempt to move the distribution of liability to illness in the whole population rather than trying to identify the individuals at higher risk. In this way, fewer people will cross the threshold even though we cannot, necessarily, predict who they are. We look at the same risk factors but we look at them and use them in a different way. In the case of suicide we can identify things like school exclusion and family breakdown [4]; cardiovascular disease reflects diet among other factors [5]. Many people with the risk factor do not have the illness and many with the illness do not have the risk factor. This should not discourage the attack now on the risk factors as they occur in the population.
The identification of individuals at risk has failed and continues to fail. Rates of psychosis vary because the population liability varies due to the sum of obscure factors of which we have only partial knowledge. We need to identify those ‘partial’ factors that affect population liability to the illness or its expression and we should then work on those factors in the whole population rather than trying to identify the high risk individual.
Population rather than individual approaches have been the effective principle in the decline in cardiovascular disease [6] and road trauma [7], which faced the same paralysis in prevention just a few years ago. We have not yet addressed the risk factors for psychoses in this way but some already stand out, such as birth trauma, drug misuse, poverty and social disadvantage. These will not be tackled in the name of psychosis prevention, yet, as they are improved, psychosis rates may change. Some risk factors may be ethically unapproachable (such as a eugenic response to the genetic risk), yet the psychiatric epidemiologist should investigate risk in a way that may be translated into population risk as well as individual risk. By this approach, risk factors can be tackled as they are found in the population rather than waiting for them to be placed in a causal chain for, verily, no general causal chain will be found.
The categorical concept of schizophrenia
When I was studying psychiatry, I was wedded to a concept of schizophrenia as a disease. Its reality was the mental hospital of its origin, which already then hardly existed. I was a good medical positivist and I believed that our diagnostic efforts and research were groping through the phenotypic veils toward the chaste prize of the true nature of schizophrenia. I may have presumed virtue in a conceptual trollop.
Expanding the concept of schizophrenia to the schizophrenia spectrum disorders is an authentic attempt to get around the extraordinary variability of the disorder while leaving the central disease concept undisturbed. I do not know that this can be maintained. Once we talk of a spectrum we have weakened that hook – that schizophrenia represents a singular truth – on which we have hung our approach to investigation. As our grasp on the underlying ‘truth’ weakens, we begin to move toward the ‘postmodern’ world. Fortunately, others have been there before us even though it is unfamiliar to medicine [8]. In the ‘postmodern’ world, the value of an idea is gauged by how useful the concept is, not by how close it takes us to the ‘truth’. Value is measured by utility, not by correspondence with truth. Once we talk of utility we must ask the questions, ‘Useful for what purpose?’ and ‘Useful to whom?’. The diagnosis of schizophrenia which is useful to an epidemiologist may not be useful to a neurochemist.
I am a psychiatrist in private practice for whom schizophrenia has ceased to be a useful concept. It no longer provides a useful description of the people sitting in my offices; it does not provide the portrait of psychopathology; it does not give us prognosis; and it does not tell us how to treat the disorder. Knowing the rate of schizophrenia with or without its spectrum does not give me useful information. In fact, this diagnosis reduces the data beyond what is useful either for the practitioner or for the patient because it hides important psychopathological differences behind diagnostic generalisations.
Can we look at these serious psychiatric disorders in another way that gives us a more useful portrait of psychopathology?
There are alternate approaches to conceptualising the heterogeneity of what we call schizophrenia. One way to do this is to atomise the diagnosis with subtypes. This has not been useful because the presentation is not only variable but also unstable over time; people can move from one subtype to another in very short periods.
Another approach is to move away from categories altogether. How then do we describe our patients in any useful way? Recent attempts at this have redescribed patients in terms of positive and negative symptoms (which spawned the SAPS and the SANS) [9]. Even though these instruments were actually predicated on a two-factor model, the instrument's own results were better explained by three factors; negative symptoms, positive symptoms and disorganisation. The factor analysis industry has generated four, five and six factor solutions. What is important is that we are taken a step away from the categorical disease diagnosis (with its assumption of single entity and single cause) and closer to the ‘domains’ of disturbed psychological functioning, which give us a better handle on what we are dealing with in disorders. For example, the prominence of negative symptoms gives information about prognosis about which is not to be found in the schizophrenia label itself. Dysphoria says something about the risks and treatment needed, which is absent from either categorical schizophrenia or its spectrum.
A number of researchers are looking back to the symptoms and summarising the data using modern statistical technology [10–14]. One might argue that they are doing with current techniques what Kraepelin did with the scientific technique of classification that dominated the sciences a century ago (and in which his own brother was pre-eminent).
If we take the factor-analytic approach and apply it across the field of psychological illness beyond the boundaries that we have thrown around schizophrenia, we find that the factors we find in schizophrenia appear in other serious conditions. We have done this for the symptom data from the survey of low prevalence (psychotic) disorders done as part of the National Survey of Mental Health and Wellbeing, Australia 1997–1998 [15]. The factor-analytic solution that fits the overall data yields five factors: (i) positive symptoms; (ii) substance abuse; (iii) dysphoria; (iv) negative symptoms; and (v) elevation and incoherence. The same factors appear in patients with all the ICD-10 diagnoses seen in the study (Fig. 1). These included the affective and schizoaffective psychoses, atypical and depressive psychoses, and severe depressions. The factors come out the same, except in a slightly different order reflecting the different weights that they have.
Scale scores for patients with severe depression (n = 79), bipolar disorder (n = 112), schizoaffective disorder (n = 102), schizophrenia (n = 510) and those who are delusional or have other symptoms (n = 145). (
) Dysphoria scale; (
) positive symptom scale; (
) substance abuse scale; (
) mania scale; (
) negative symptom/incoherence scale
Pathognomonic symptoms do not separate the diagnoses. We are seeing differing levels of disturbance in domains of psychological function. If we convert the factors into scales, the diagnoses appear as differences of the balance of disturbance in the domains of psychological functioning.
The balances of disturbed functioning are what are important for us to know. The factors rather than the diagnoses give the clinically useful information about description and treatment for the practising psychiatrist. These factors can be used for clinical report or resource allocation decisions. They can be examined for validity and reliability. Their biological, psychological and sociological origins and relationships can be plumbed. In short, descriptive factors can be used in the same way as a diagnosis, lacking only the feeling of substantial reality with which we invest our diagnoses.
Now this is not news, of course; in fact many psychiatrists have been behaving this way for years. As a fresh graduate I was scandalised when an experienced psychiatrist said to me that private psychiatrists did not have much use for diagnosis. I think now I understand what he meant. In practice we treat the symptom pictures we see and the formal diagnosis rarely clarifies that picture. Should our epidemiological work follow this principle rather than following the categorical diagnosis line?