Scott Henderson's lucid editorial, as its title implies, is firmly anchored in the present day. As a yardstick of progress, however, it may be useful to compare his review with one written nearly 50 years ago, which likewise took as its starting point Morris' list of seven ‘uses of epidemiology’ (Shepherd & Cooper, Reference Shepherd and Cooper1964). Although its authors managed to cite mental health research on each of the seven, their quest took them to the edges of psychiatric practice (e.g. mental handicap, accident-proneness, tension headache and psychosomatic aspects of peptic ulcer). The present editorial, in contrast, can draw upon national prevalence surveys in a score of countries as well as causally oriented research focusing on schizophrenia, affective disorder and other central concerns of today's clinical specialty. The simple comparison points to an impressive growth in volume and scope of psychiatric epidemiology over the past half century.
How far has this expansion been reflected in substantive scientific achievement? In order to address the question, let us consider briefly each in turn of Henderson's ‘four significant questions’.
1. Can epidemiology contribute to diagnostic classification?
The issue has become topical because repeated revisions of the standard classifications have multiplied the numbers of available diagnoses, leading to an explosive, though probably artefactual, increase in reported co-morbidity, and there is uncertainty as to how this situation should be dealt with. Pincus et al. (Reference Pincus, Tew and First2004) have argued forcefully that ‘the current strategy of diagnosing “maximal” co-morbidity may not reflect “optimal” co-morbidity’. With each new edition, the diagnostic pie has been divided into smaller slices, so that more and more patients qualify for multiple labelling. This trend threatens to overwhelm clinicians and health information systems, until the wood can no longer be seen for the trees and the target for treatment becomes indistinct. Analysis of population-based symptom distributions, it is argued, could resolve this dilemma by providing a basis for alternative strategies; notably (a) the building of diagnostic hierarchies and (b) a dimensional approach.
In each instance, however, there are weighty objections to be overcome. Imposing diagnostic hierarchies at the present stage of knowledge would require too much theorizing about underlying pathogenetic processes and so weaken the empirical basis of classification. Dimensional models of classification, for their part, have been applied more successfully to personality traits than to clinical states. A prototypical trait is an attribute of the person: innate, stable and enduring; whereas a prototypical state is a characteristic: essentially time-limited, changeable and subject to external conditions (Chaplin et al. Reference Chaplin, John and Goldberg1988). Insofar as human psychopathology is defined in terms of illness (or morbidity), it is better regarded as a state.
2. How should psychiatric morbidity in populations be measured?
Mensuration lies at the core of epidemiology. The data in question must be clearly defined and collected by means of reliable tools, while standardised methods are required to ensure comparability of findings. The editorial succinctly reviews these ‘instruments’ from brief rating scales via the development of complaint inventories such as the General Health Questionnaire (GHQ), and on to more detailed and searching structured interviews such as the Schedule for Clinical Assessment in Neuropsychiatry (SCAN) and Composite International Diagnostic Interview (CIDI) (see editorial references). It then makes a fundamental point, namely that ‘(psychiatric) case-finding necessarily uses a different method from elsewhere in medicine: the presence of morbidity is determined exclusively by the verbal responses of individuals about their recent symptoms’. Here a caveat must be entered: the statement holds true principally for ‘third generation’ studies of the past 30 years or so, in which large population samples have been questioned by teams of lay interviewers using closed-question schedules.
Standardised interviews of this kind differ widely from clinical ‘case-taking’ methods in psychiatry, which do not include systematic checking of a symptom inventory, but start from the presenting condition and proceed to assessment of the patient's mental status from his or her appearance, behaviour, talk, mood and thought content; any abnormal beliefs or experiences, and cognitive status in terms of memory, orientation and general knowledge (Cooper & Singh, Reference Cooper and Singh2000).
Some earlier semi-structured interviews (e.g. the Mental Status and Clinical Interview Schedules) did respect the need to rate both reported symptoms and abnormalities noted at interview, but the distinction was largely abandoned as sample sizes and reliance on lay interviewers increased. Today there is a growing awareness that clinical abnormalities should also be recorded in research interviews, though no consensus as to how this might best be achieved. The two-phase survey design summarized in the editorial may offer the best hope for a compromise solution to this dilemma.
3. What have we learned about the prevalence and distribution of mental illness in populations?
This question requires a review of its own. In essence, there has been an enormous increase in descriptive information from prevalence surveys, including international collaborative projects. However, as the editorial points out, the variation among countries in reported prevalence estimates is disturbingly large, raising doubt about the validity of diagnostic findings. This can also apply to differences reported in a single country over short time-periods, as with the Epidemiologic Catchment Area (ECA) and National Co-morbidity Survery (NCS) projects in the United States (Cooper & Singh, Reference Cooper and Singh2000). Henderson infers that the chief utility of prevalence surveys probably resides in the impact each may have in its own country, by highlighting unmet needs and raising both medical and lay-public awareness of the suffering and dangers these entail. Equally important in this context, however, is the guidance they can provide for service planning.
Here the descriptor ‘mental health’ seems preferable to ‘psychiatric’ epidemiology, since given one-year prevalence estimates around a mean of 12% of the general population, it is apparent that the great mass of ‘common mental disorders’ can never be dealt with by specialist services alone. A logistical basis for service planning may be a needs-assessment model, in which care packages are specified in terms of psychiatric in-patient or ambulant treatment, primary health care (with specialist back-up when needed), community support, or some combination of these. The hierarchy of ‘levels’ and ‘filters’ proposed by Goldberg & Huxley (Reference Goldberg and Huxley1992) typifies a flexible system of this kind, which can be adapted to each country's existing service structures.
4. What pathogenic exposures in the environment have been identified by epidemiology?
Straightforward head counts may yield clues to aetiology where an epidemic spread or rapid increase in frequency is revealed. Thus, the explosion of illicit drug injecting in developing countries since the 1980s can be linked to social and economic consequences of globalisation (Wodak et al. Reference Wodak, Sarkar and Mesquita2004), as can also the contemporary growth of suicide by pesticide swallowing in India and other Asian countries (Konradsen, Reference Konradsen2007).
Studying fluctuations in the population frequencies of endemic conditions and linking them to pathogenic agents is a much more difficult exercise. It may be confounded both by the secular trend towards increased case identification and by the effects of pharma-marketing campaigns on diagnostic rates (currently, e.g., autistic spectrum disorders in children and bipolar disorder). In this context, prevalence surveys are too crude and must be supplemented by prospective incidence, cohort or case-control designs.
The need to focus on relatively specific pathogens will also call for measurement of environmental exposures (White et al. Reference White, Armstrong and Saracci2008), which in relation to mental disorders can involve a very wide range of causal or risk factors ranging from natural or man-made disasters to developmental trauma, neurotoxic and infective hazards and on to macrosocial conditions, family dynamics and negative life events (Cooper, Reference Cooper, Cheng and Cooper2001). Viewed in this light, indeed, mental health epidemiology takes on the aspect of a monitor of human distress.
Here it is necessary to distinguish between distal predisposing factors that can establish one or other form of lasting vulnerability, proximal provoking causes in the subsequent lifespan and possible interaction between the two categories. Thus, of the four factors for schizophrenia cited in the editorial, developmental trauma belongs to the first group and cannabis use to the second, while both ethnic status and urban residence may relate to either.
In conclusion, Henderson suggests that prospects for this sub-specialty will be brightest where data on pathogenic exposures can be harnessed to advances in molecular genetics, so that their pathways of transmission to the brain can be traced. It may be so, but in global perspective at least, a balanced appraisal might regard ‘molecular epidemiology’ as simply a travelling companion (McMichael, Reference McMichael1994), and see in closer conjunction with social psychiatry on one side, and public-health epidemiology on the other, the best hopes for future progress.
You have
Access
Scott Henderson's lucid editorial, as its title implies, is firmly anchored in the present day. As a yardstick of progress, however, it may be useful to compare his review with one written nearly 50 years ago, which likewise took as its starting point Morris' list of seven ‘uses of epidemiology’ (Shepherd & Cooper, Reference Shepherd and Cooper1964). Although its authors managed to cite mental health research on each of the seven, their quest took them to the edges of psychiatric practice (e.g. mental handicap, accident-proneness, tension headache and psychosomatic aspects of peptic ulcer). The present editorial, in contrast, can draw upon national prevalence surveys in a score of countries as well as causally oriented research focusing on schizophrenia, affective disorder and other central concerns of today's clinical specialty. The simple comparison points to an impressive growth in volume and scope of psychiatric epidemiology over the past half century.
How far has this expansion been reflected in substantive scientific achievement? In order to address the question, let us consider briefly each in turn of Henderson's ‘four significant questions’.
1. Can epidemiology contribute to diagnostic classification?
The issue has become topical because repeated revisions of the standard classifications have multiplied the numbers of available diagnoses, leading to an explosive, though probably artefactual, increase in reported co-morbidity, and there is uncertainty as to how this situation should be dealt with. Pincus et al. (Reference Pincus, Tew and First2004) have argued forcefully that ‘the current strategy of diagnosing “maximal” co-morbidity may not reflect “optimal” co-morbidity’. With each new edition, the diagnostic pie has been divided into smaller slices, so that more and more patients qualify for multiple labelling. This trend threatens to overwhelm clinicians and health information systems, until the wood can no longer be seen for the trees and the target for treatment becomes indistinct. Analysis of population-based symptom distributions, it is argued, could resolve this dilemma by providing a basis for alternative strategies; notably (a) the building of diagnostic hierarchies and (b) a dimensional approach.
In each instance, however, there are weighty objections to be overcome. Imposing diagnostic hierarchies at the present stage of knowledge would require too much theorizing about underlying pathogenetic processes and so weaken the empirical basis of classification. Dimensional models of classification, for their part, have been applied more successfully to personality traits than to clinical states. A prototypical trait is an attribute of the person: innate, stable and enduring; whereas a prototypical state is a characteristic: essentially time-limited, changeable and subject to external conditions (Chaplin et al. Reference Chaplin, John and Goldberg1988). Insofar as human psychopathology is defined in terms of illness (or morbidity), it is better regarded as a state.
2. How should psychiatric morbidity in populations be measured?
Mensuration lies at the core of epidemiology. The data in question must be clearly defined and collected by means of reliable tools, while standardised methods are required to ensure comparability of findings. The editorial succinctly reviews these ‘instruments’ from brief rating scales via the development of complaint inventories such as the General Health Questionnaire (GHQ), and on to more detailed and searching structured interviews such as the Schedule for Clinical Assessment in Neuropsychiatry (SCAN) and Composite International Diagnostic Interview (CIDI) (see editorial references). It then makes a fundamental point, namely that ‘(psychiatric) case-finding necessarily uses a different method from elsewhere in medicine: the presence of morbidity is determined exclusively by the verbal responses of individuals about their recent symptoms’. Here a caveat must be entered: the statement holds true principally for ‘third generation’ studies of the past 30 years or so, in which large population samples have been questioned by teams of lay interviewers using closed-question schedules.
Standardised interviews of this kind differ widely from clinical ‘case-taking’ methods in psychiatry, which do not include systematic checking of a symptom inventory, but start from the presenting condition and proceed to assessment of the patient's mental status from his or her appearance, behaviour, talk, mood and thought content; any abnormal beliefs or experiences, and cognitive status in terms of memory, orientation and general knowledge (Cooper & Singh, Reference Cooper and Singh2000).
Some earlier semi-structured interviews (e.g. the Mental Status and Clinical Interview Schedules) did respect the need to rate both reported symptoms and abnormalities noted at interview, but the distinction was largely abandoned as sample sizes and reliance on lay interviewers increased. Today there is a growing awareness that clinical abnormalities should also be recorded in research interviews, though no consensus as to how this might best be achieved. The two-phase survey design summarized in the editorial may offer the best hope for a compromise solution to this dilemma.
3. What have we learned about the prevalence and distribution of mental illness in populations?
This question requires a review of its own. In essence, there has been an enormous increase in descriptive information from prevalence surveys, including international collaborative projects. However, as the editorial points out, the variation among countries in reported prevalence estimates is disturbingly large, raising doubt about the validity of diagnostic findings. This can also apply to differences reported in a single country over short time-periods, as with the Epidemiologic Catchment Area (ECA) and National Co-morbidity Survery (NCS) projects in the United States (Cooper & Singh, Reference Cooper and Singh2000). Henderson infers that the chief utility of prevalence surveys probably resides in the impact each may have in its own country, by highlighting unmet needs and raising both medical and lay-public awareness of the suffering and dangers these entail. Equally important in this context, however, is the guidance they can provide for service planning.
Here the descriptor ‘mental health’ seems preferable to ‘psychiatric’ epidemiology, since given one-year prevalence estimates around a mean of 12% of the general population, it is apparent that the great mass of ‘common mental disorders’ can never be dealt with by specialist services alone. A logistical basis for service planning may be a needs-assessment model, in which care packages are specified in terms of psychiatric in-patient or ambulant treatment, primary health care (with specialist back-up when needed), community support, or some combination of these. The hierarchy of ‘levels’ and ‘filters’ proposed by Goldberg & Huxley (Reference Goldberg and Huxley1992) typifies a flexible system of this kind, which can be adapted to each country's existing service structures.
4. What pathogenic exposures in the environment have been identified by epidemiology?
Straightforward head counts may yield clues to aetiology where an epidemic spread or rapid increase in frequency is revealed. Thus, the explosion of illicit drug injecting in developing countries since the 1980s can be linked to social and economic consequences of globalisation (Wodak et al. Reference Wodak, Sarkar and Mesquita2004), as can also the contemporary growth of suicide by pesticide swallowing in India and other Asian countries (Konradsen, Reference Konradsen2007).
Studying fluctuations in the population frequencies of endemic conditions and linking them to pathogenic agents is a much more difficult exercise. It may be confounded both by the secular trend towards increased case identification and by the effects of pharma-marketing campaigns on diagnostic rates (currently, e.g., autistic spectrum disorders in children and bipolar disorder). In this context, prevalence surveys are too crude and must be supplemented by prospective incidence, cohort or case-control designs.
The need to focus on relatively specific pathogens will also call for measurement of environmental exposures (White et al. Reference White, Armstrong and Saracci2008), which in relation to mental disorders can involve a very wide range of causal or risk factors ranging from natural or man-made disasters to developmental trauma, neurotoxic and infective hazards and on to macrosocial conditions, family dynamics and negative life events (Cooper, Reference Cooper, Cheng and Cooper2001). Viewed in this light, indeed, mental health epidemiology takes on the aspect of a monitor of human distress.
Here it is necessary to distinguish between distal predisposing factors that can establish one or other form of lasting vulnerability, proximal provoking causes in the subsequent lifespan and possible interaction between the two categories. Thus, of the four factors for schizophrenia cited in the editorial, developmental trauma belongs to the first group and cannabis use to the second, while both ethnic status and urban residence may relate to either.
In conclusion, Henderson suggests that prospects for this sub-specialty will be brightest where data on pathogenic exposures can be harnessed to advances in molecular genetics, so that their pathways of transmission to the brain can be traced. It may be so, but in global perspective at least, a balanced appraisal might regard ‘molecular epidemiology’ as simply a travelling companion (McMichael, Reference McMichael1994), and see in closer conjunction with social psychiatry on one side, and public-health epidemiology on the other, the best hopes for future progress.