Epidemiology of Schizophrenia
Assen Jablensky
Introduction
Epidemiological research into schizophrenia aims to answer four essential questions.
What is the ‘true’ population frequency of the disorder in various populations and how is it distributed across the various groups within populations?
Do the incidence, manifestations, and course of schizophrenia vary in relation to factors of the physical and social environment?
Who is at risk and what forces determine or influence the risk of developing schizophrenia?
Can the answers to the above questions help explain what causes the disorder and how to prevent it?
Table 4.3.5.1 Historical landmarks in the epidemiology of psychoses | |||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
The hallmark of the epidemiological method (see Chapter 2.7) is the referral of a measure (numerator) of the occurrence of a disorder, or of any associated characteristics, to a population base (denominator), such as person-years at risk. The epidemiological study of diseases usually proceeds from a description of its frequency and associations (establishing rates of occurrence) to testing hypotheses about risk factors and causes by analysing ratios between rates.
Schizophrenia has been studied extensively from an epidemiological perspective since Kraepelin(1) introduced the concept of dementia praecox in 1896. In the first half of the twentieth century, epidemiological research into schizophrenia took two divergent paths. While European studies tended to focus on population distributions and genetic risks, North American researchers investigated the social ecology of the disorder. A variety of methods were explored and successfully applied by the pioneers of psychiatric epidemiology, and the contours of the epidemiological map of schizophrenia in Europe and North America were effectively laid down between the two World Wars. The early studies were carried out by dedicated researchers who often spent months or years collecting data ‘door-to-door’ in small communities. Close knowledge of the respondents, access to multigenerational records from the local parish registers, and the cooperation of the community resulted in studies that remain landmarks of psychiatric epidemiology (Table 4.3.5.1).
During the last several decades, the scope of epidemiological studies of schizophrenia has expanded to include populations in Asia, Africa, and South America about which little had been known previously. The World Health Organization (WHO) International Pilot Study of Schizophrenia and its successor, the WHO 10-country epidemiological study(9,10) were the first systematic investigations of the comparative incidence, clinical manifestations, and course
of schizophrenia in both developing and developed countries. The WHO programme was an impetus for similar studies in India, China, the Caribbean, and Australia. Two major studies of psychiatric morbidity in the United States, the Epidemiological Catchment Area project,(11) and the National Comorbidity Survey,(12) generated data on the prevalence of DSM-III/IIIR schizophrenia and related disorders in representative population samples. In the 1980s and 1990s, epidemiological studies increasingly utilized existing large databases such as cumulative case registers or birth cohorts to test hypotheses about risk factors, and began to include methods of genetic epidemiology. There is a current tendency towards integrating epidemiological approaches with other types of aetiological research in schizophrenia. This predicts an important role for epidemiology in the era of molecular biology of mental disorders.
of schizophrenia in both developing and developed countries. The WHO programme was an impetus for similar studies in India, China, the Caribbean, and Australia. Two major studies of psychiatric morbidity in the United States, the Epidemiological Catchment Area project,(11) and the National Comorbidity Survey,(12) generated data on the prevalence of DSM-III/IIIR schizophrenia and related disorders in representative population samples. In the 1980s and 1990s, epidemiological studies increasingly utilized existing large databases such as cumulative case registers or birth cohorts to test hypotheses about risk factors, and began to include methods of genetic epidemiology. There is a current tendency towards integrating epidemiological approaches with other types of aetiological research in schizophrenia. This predicts an important role for epidemiology in the era of molecular biology of mental disorders.
Epidemiological methods and instruments in the study of schizophrenia
The measurement of the prevalence, incidence, and disease expectancy of schizophrenia depends critically on the sensitivity of the case-finding method (i.e. its capacity to identify all affected persons in a given population) and the availability of a diagnostic instrument or procedure that selects ‘true’ cases (i.e. those corresponding to an established clinical concept).
Case-finding
Case-finding designs fall into three broad groups: case detection in clinical populations, door-to-door surveys of population samples or whole communities, and birth cohort studies. Each method has its advantages and limitations.
While case-finding through the mental health services provides a relatively easy access to a substantial proportion of all persons with schizophrenia, the cases in treatment may not be fully representative of all individuals with the disorder. Bias related to gender, marital status, socio-economic factors, culture, or ethnicity are known to affect the probability of being in treatment at a given time in a given setting, and generalizations about schizophrenia from hospital or clinic samples are liable to error. Some of the deficiencies of case-finding through service contacts are avoided in cumulative national or regional psychiatric case registers, which cover large well-defined populations and can be linked to other population databases (e.g. birth records). This makes registers efficient research instruments in low-incidence disorders such as schizophrenia.
Surveys involve accounting for every person at risk within a defined community or a population sample in terms of either being or not being a case. Face-to-face interviews (and follow-up) of all residents in defined communities has been a feature of some high-quality research, especially in the Scandinavian countries. However, since the size of the populations surveyed in this way is limited, the number of detected cases of schizophrenia is usually too small to generate stable estimates of epidemiological parameters. Surveys of large populations involve two basic designs: a single-phase survey of a probability sample drawn from the general population, and a two-phase survey where a validated screening test is first applied to the entire population and only those scoring as screen-positive proceed to a full assessment. In the instance of schizophrenia, logistics dictates a choice between assessing large numbers less rigorously and investigating a smaller sample in greater depth. In the absence of a simple and valid screening procedure for schizophrenia, such as a biological or psychological test, the advantages of the two-phase survey may be offset by poor sensitivity or specificity of the screening device which is usually a questionnaire or checklist.
The study of birth cohorts at ages when their members have passed through the greater part of the period of risk for onset of schizophrenia is usually done by direct interviewing or by analysing available case register data. Well-characterized birth cohorts are among the best tools for the study of the incidence of schizophrenia and associated risk factors. However, even in settings where the population is stable and mortality and morbidity are adequately monitored, the size of birth cohorts with prospectively collected data may not be sufficient for conclusive epidemiological inferences.
All this suggests that there is no single ‘gold standard’ of case-finding for schizophrenia that could be applied across all possible settings, and the assets and liabilities of particular case-finding procedures need to be evaluated in the context of each study. This makes the detailed reporting of case-finding methods a mandatory prerequisite for an ‘evidence-based’ epidemiology of schizophrenia.
Diagnosis
Variation in diagnostic concepts and practices always explains a proportion of the variation in the results of schizophrenia studies, especially if they involve different populations or different periods. Until the 1960s, the diagnostic rules used in epidemiological research were seldom explicitly stated. In the late 1960s, the WHO International Pilot Study of Schizophrenia(10) examined diagnostic variation in schizophrenia across nine countries by comparing the diagnoses made by psychiatrists using a semi-structured clinical interview with diagnostic classification by a computer algorithm(13) utilizing the same interview data. The results demonstrated that in the majority of settings psychiatrists were using comparable diagnostic concepts in the Kraepelin–Bleuler tradition. The introduction of explicit diagnostic criteria and rules with the consecutive editions of DSM and the WHO’s ICD-10 improved further the reliability of diagnosis but did not resolve all diagnostic issues with implications for epidemiology. While ICD-10 and DSM-IV tend to agree well on the core cases of schizophrenia, they agree less well on the classification of atypical or milder cases. Such differences may be less important in clinical practice but they present a problem for epidemiological and genetic studies. By providing more restrictive criteria for schizophrenia, both classifications aim to identify clinically similar cases and to minimize false-positive diagnoses. This is not an unequivocal advantage for epidemiology. Applying such criteria at case-finding may result in the rejection of potential cases which fail to meet the full set of criteria at initial assessment. Therefore it is desirable to develop less restrictive screening versions of the DSM and ICD criteria for epidemiological research.
Instruments
The diagnostic instruments used in surveys which involve interviewing fall into two categories: fully structured interviews such as
the Diagnostic Interview Schedule (DIS)(12) and the Composite International Diagnostic Interview (CIDI)(14) both written to match exactly the diagnostic criteria of DSM-IIIR/IV and ICD-10, and semi-structured interview schedules such as the Present State Examination (PSE)(13) and the Schedules for Clinical Assessment in Neuropsychiatry (SCAN),(15) which cover a broad range of psychopathology and elicit data that can be processed by alternative diagnostic algorithms.
the Diagnostic Interview Schedule (DIS)(12) and the Composite International Diagnostic Interview (CIDI)(14) both written to match exactly the diagnostic criteria of DSM-IIIR/IV and ICD-10, and semi-structured interview schedules such as the Present State Examination (PSE)(13) and the Schedules for Clinical Assessment in Neuropsychiatry (SCAN),(15) which cover a broad range of psychopathology and elicit data that can be processed by alternative diagnostic algorithms.
The DIS/CIDI type of instrument is reliable and capable of generating standard diagnoses of common mental disorders in a single-phase survey design. Its clinical validity in schizophrenia is less certain because symptoms may not be reported accurately or impairment may be underestimated by the respondent. In contrast, the PSE/SCAN allows a greater amount of psychopathological data to be elicited in a flexible clinical interview format, but its use in epidemiological studies presupposes availability of clinically trained interviewers. While SCAN and other similar interviews are suitable as second-stage diagnostic instruments, there is still a need for a relatively simple and effective screening procedure for case-finding of schizophrenia in field surveys.
Persons, place, time: descriptive epidemiology of schizophrenia
The epidemiological description of schizophrenia draws on extensive evidence available today on its frequency, age, and sex distribution in relatively large populations or geographical areas. Less than complete information is available on variations in its epidemiological characteristics that may be found in unusual or isolated populations, or on the temporal trends in its occurrence.
Prevalence, incidence, and disease expectancy
(a) Prevalence
Prevalence provides an estimate of the number of cases per 1000 persons at risk present in a population at a given time or over a defined period. Point prevalence refers to the ‘active’ (i.e. symptomatic) cases on a given date, or within a brief census period. Since asymptomatic cases (e.g. persons in remission) will be missed in a point prevalence survey, it is useful to supplement the assessment of the present mental state with an enquiry about past episodes of the disorder to obtain a lifetime prevalence index. In schizophrenia, which tends to a chronic course, estimates of point and lifetime prevalence will be closer to each other than in remitting illnesses.
An overview of selected prevalence studies of schizophrenia spanning nearly seven decades is presented in Table 4.3.5.2. The studies differ in many aspects of methodology but the majority of them feature a high intensity of case-finding. Several studies are repeat surveys in which the original population was reinvestigated following an interval of 10 or more years (the resulting consecutive prevalence figures are indicated by arrows).
The majority of studies have produced point prevalence estimates in the range 2.1 to 7.0 per 1000 population at risk and lifetime prevalence of schizophrenia in the range 15.0 to 19.0 per 1000. The figures are not uniformly standardized, and should be compared with caution because of demographic differences between populations related to factors such as age-specific mortality and migration. A systematic review of 188 studies in 46 countries, published between 1965 and 2002,(16) estimated the median value for point prevalence at 4.6 per 1000 persons and for lifetime prevalence at 7.2 per 1000.
Certain populations and groups deviate markedly from the central tendency. Strikingly high prevalence of schizophrenia (two to three times the national or regional average) has been found in geographically and genetically isolated populations, including small communities in Northern Sweden and Finland, and several Western Pacific islands (see Table 4.3.5.2). At the other extreme, a virtual absence of schizophrenia and a high rate of depression have been claimed for the Hutterites of South Dakota, a Protestant sect whose members live in close-knit endogamous communities sheltered from the outside world.(33) Negative social selection for schizoid individuals who fail to adjust to the lifestyle of the majority and eventually migrate without leaving progeny has been suggested (but not definitively proven) as an explanation. Results of two surveys in Taiwan,(21) separated by 15 years, point to a falling prevalence of schizophrenia (from 2.1 to 1.4 per 1000) in the context of major socio-economic change and an overall increase in total mental morbidity in the population.
The question about the extent of true variation in the prevalence of schizophrenia across populations has no simple answer. Methodological differences among studies, related to sampling, case-finding, and diagnostic assessment are likely to account for a good deal of the observed variation. As an example, the high mean prevalence rate of DSM-III schizophrenia reported from the Epidemiologic Catchment Area study in the United States(25) is difficult to reconcile with inconsistencies, such as a 13-fold difference in the rates for age group 18–24 years across the various sites of the survey. One possible reason is that the principal diagnostic instrument of the survey (DIS), administered by lay interviewers, may produce both false-positive and false-negative diagnoses of schizophrenia in a number of cases. Similarly, computer-generated diagnoses of ‘non-affective psychosis’ in the National Comorbidity Survey,(12) based on a version of the CIDI administered by lay interviewers, were found to agree poorly with clinicians’ diagnoses when a subsample of the respondents were re-interviewed over the telephone.(34)
Notwithstanding such caveats in the interpretation of survey findings, the prevalence rates are fairly similar in the majority of studies, though certain specific populations clearly deviate from the modal value. Even in those instances, however, the magnitude of the deviation is modest compared with the 10- to 30-fold differences in prevalence observed in other multifactorial diseases (e.g. diabetes, ischaemic heart disease, multiple sclerosis) across populations.
(b) Incidence
The incidence rate (an estimate of the annual number of first-onset cases in a defined population per 1000 persons at risk) is of greater interest for the study of risk factors than prevalence since it represents the so-called force of morbidity (the probability of disease occurrence) in a given population, and is closer in time to the action of antecedent or precipitating factors. The estimation of incidence depends critically on the ability to determine reliably the point of onset of the disorder. In the case of schizophrenia, the long prodromal period and the fuzzy boundary between premorbid state and onset of psychosis make this particularly difficult. In the absence of an objective biomarker of the disease, onset is usually defined as
the point in time when clinical manifestations become recognizable and diagnosable according to specified criteria. The first hospital admission, which has been used as a proxy for disease onset in many studies, is not a robust indicator because of the variable time lag between the earliest appearance of symptoms and the first-admission across treatment facilities and settings. A better approximation is provided by the first-contact, i.e. the point at which any psychiatric, general medical, or alternative ‘helping’ agency is accessed by symptomatic individuals for the first time. A limitation common to both first-admission and first-contact studies is that they produce rates of ‘treated’ incidence and miss symptomatic cases that do not present for assessment or treatment. This limitation can be overcome by periodically repeated door-to-door surveys of the same population or by longitudinal cohort studies (though both are difficult to mount for reasons of cost and logistics).
the point in time when clinical manifestations become recognizable and diagnosable according to specified criteria. The first hospital admission, which has been used as a proxy for disease onset in many studies, is not a robust indicator because of the variable time lag between the earliest appearance of symptoms and the first-admission across treatment facilities and settings. A better approximation is provided by the first-contact, i.e. the point at which any psychiatric, general medical, or alternative ‘helping’ agency is accessed by symptomatic individuals for the first time. A limitation common to both first-admission and first-contact studies is that they produce rates of ‘treated’ incidence and miss symptomatic cases that do not present for assessment or treatment. This limitation can be overcome by periodically repeated door-to-door surveys of the same population or by longitudinal cohort studies (though both are difficult to mount for reasons of cost and logistics).
Table 4.3.5.2 Selected prevalence studies of schizophrenia | ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Table 4.3.5.3 summarizes the essential features of 12 selected incidence studies of schizophrenia. Studies using a ‘broad’ definition of schizophrenia (ICD-8 or ICD-9) estimate about three-fold difference in the variation of rates, in the range from 0.17 to 0.57 per 1000 population per year, for first-admissions or first contacts. Studies using more stringent criteria, such as the Research Diagnostic Criteria (RDC),(121) DSM-IV, ICD-10, or Catego S+,(13) have reported incidence rates two to three times lower than those based on ‘broad’ criteria. A systematic review of data from some 160 studies from 33 countries, published between 1965 and 2001,(35) yielded a median value of 0.15 and mean value of 0.24 per 1000, with a five-fold range of the rates and a tendency for more recent studies to report lower rates.
Considering the methodological differences among individual studies, generalizing about the incidence of schizophrenia from pooled data may be problematic. To date, the only investigation
that has applied a uniform design and common research tools to generate directly comparable incidence data for different populations is the WHO 10-country study.(9) Incidence counts in the WHO study were based on first-in-lifetime contacts with any ‘helping agency’ within defined areas (including traditional healers in the developing countries) which were monitored over a 2-year period. Potential cases and key informants were interviewed by clinicians using standardized instruments, and the timing of onset was ascertained for the majority of the patients. In 86 per cent of the 1022 patients the onset of diagnostic symptoms of schizophrenia was within the year preceding the first-contact, and therefore the first-contact incidence rate was adopted as a reasonable approximation to the ‘true’ onset rate. Two definitions of ‘caseness’, differing in the degree of specificity, were used to determine incidence: a ‘broad’ clinical definition comprising ICD-9 schizophrenia and paranoid psychoses, and a more restrictive definition of PSE/Catego S+(13) ‘nuclear’ schizophrenia manifesting with Schneiderian first-rank symptoms. The rates for eight of the catchment areas are shown in Table 4.3.5.4.
that has applied a uniform design and common research tools to generate directly comparable incidence data for different populations is the WHO 10-country study.(9) Incidence counts in the WHO study were based on first-in-lifetime contacts with any ‘helping agency’ within defined areas (including traditional healers in the developing countries) which were monitored over a 2-year period. Potential cases and key informants were interviewed by clinicians using standardized instruments, and the timing of onset was ascertained for the majority of the patients. In 86 per cent of the 1022 patients the onset of diagnostic symptoms of schizophrenia was within the year preceding the first-contact, and therefore the first-contact incidence rate was adopted as a reasonable approximation to the ‘true’ onset rate. Two definitions of ‘caseness’, differing in the degree of specificity, were used to determine incidence: a ‘broad’ clinical definition comprising ICD-9 schizophrenia and paranoid psychoses, and a more restrictive definition of PSE/Catego S+(13) ‘nuclear’ schizophrenia manifesting with Schneiderian first-rank symptoms. The rates for eight of the catchment areas are shown in Table 4.3.5.4.
Table 4.3.5.3 Selected incidence studies of schizophrenia | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
|
Related posts:
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
