Three British case-register studies showed either steady or increasing rates due to changes in the composition of their populations: Bamrah and others in Salford found increasing prevalence rates from 1974 to 1984, but during this period there was a substantial decrease in the population and an increase in social deprivation and unemployment (9). Harrison and others found no change in the rate of first-onset schizophrenia from 1975 to 1987 in Nottingham but attributed the lack of a decline to an increase in the proportion of migrants and their children in the population at risk (10). In Camberwell, Castle and others found an increase in first-contact rates of schizophrenia from 1965 to 1984 and attributed this difference from the overall decline in schizophrenia rates in England to the large influx of Afro-Caribbeans, whose rates of schizophrenia are 4 to 8 times that of their White counterparts (11).

Studies based on official statistics consistently show a decrease in the administrative incidence rate of schizophrenia. Eagles and Whalley, in Scotland, report that between 1969 and 1978 there was a consistent decrease in first-admission rates, regardless of sex and age (12). Joyce found a decline in first-admission rates of schizophrenia in New Zealand from 1974 to 1984 and noted a rise in first-admission rates for mania, but this was insufficient to account for the observed change (13). Der and others, using official statistics from England and Wales from 1952 to 1986, found a decrease in the incidence of schizophrenia from the mid-1960s, while there was no corresponding increase in the rate of mania (14). Takei and others reported a marked decrease in first-admission rates for schizophrenia from 1966 to 1990 for all of Scotland (15). Although detailed analysis suggested that this was mostly due to period effect, birth-cohort effects also played a role.

In North America, there are no large databases from which to produce studies comparable to those already described. Several authors have reported on studies that suggest possible changing rates of schizophrenia, but the methodologies do not allow strong epidemiological conclusions. In Canada, Nicole and others determined first-admission rates in the mid-1980s by chart review and compared their results with reported European incidence rates (16). They concluded that their data supported decreasing incidence over time but urged the collection of longitudinal data to confirm this finding. Stoll and others reported discharge rates for schizophrenia from 1972 to 1988 for 6 teaching hospitals in the US and Canada (17). The results were varied, with the US hospitals showing a decrease in the number of discharges and the single Canadian facility showing a small increase. Since this study used event data, rather than person-based data, and was not geographically based, it cannot be concluded that the changes represented a true change in the incidence of schizophrenia. Torrey analyzed results of several US studies using data from case registers, the Epidemiological Catchment Area (ECA), the National Comorbidity Survey (NCS), and National Institute of Mental Helath (NIMH) sample surveys (18). Despite

the methodological problem of making such comparisons, he concluded that there is a trend toward an increasing incidence of schizophrenia. He emphasized the lack of an adequate US database for properly studying this issue. Although several mental health community studies have been carried out in North America, they are cross-sectional, not longitudinal, and exclude institutional and homeless populations. Kendler and others, using clinical follow-up interviews with respondents in the NCS, found that the NCS diagnoses of psychosis agreed poorly with clinical diagnoses (19). Offord and others did not report schizophrenia and related psychosis in the Mental Health Supplement to the Ontario Health Survey because the number of respondents meeting the criteria was too small for reliable estimates (20).

Methods

Kingston, Ontario, is an ideal location for epidemiological study. It is the tertiary care university-based medical centre for Eastern Ontario, serving a population of about 600 000. It is relatively isolated from major centres, and the population tends to be quite stable, having maintained a similar demographic profile over the last 15 years.

The Kingston Psychiatric Record Linkage System (KPRLS) is a psychiatric case register established in 1984 (21,22). Since then, the KPRLS has collected and linked demographic, diagnostic, and service use information for all inpatients and outpatients seen by the psychiatric services of the 3 Kingston hospitals (2 general and 1 provincial psychiatric facility). Since 1994 the system has also included referrals to psychiatry in the 2 emergency departments of the general hospitals. The psychiatric hospital does not provide emergency services. Ethics approval was received from Queen’s University in 1984 and is updated annually.

Each facility provides demographic and diagnostic information about patients on every admission, at first outpatient attendance at that facility, and for psychiatric consultations in the emergency department. Contact data is supplied for each subsequent outpatient visit. Periodic spot checks are performed to verify the accuracy of the information. To protect confidentiality, the identifying data are stored in a computer file separate from the clinical information. Linkage is performed by computer-assisted manual procedures. Lifetime diagnoses are assigned using a DSM-IV hierarchical system applied to cumulative diagnostic information in the KPRLS. As of September 30, 2000, 412 000 contacts by 28 250 individuals have been compiled into the system. As with any case register, or linked data system, a major advantage for epidemiological study is that information on all patients receiving specified services is included, allowing person-based as well as event-based analyses. A limitation of such systems is that the data about each individual are kept to a minimum so that  they can be reasonably and accurately collected.