Changes in the age structure of the population have been discussed as an issue affecting rates of schizophrenia in all the review articles cited. If the age at first-admission increases, it produces a corresponding decrease in first-admission rates. Aging of the population will affect prevalence rates even more than incidence rates as more individuals with schizophrenia accumulate in the population. Many studies have attempted to minimize this confounding factor. Eagles and others (1,12), Geddes and others (2), and Kendell and others (3) all found decreased incidence even while using age-standardized rates. Jones and others employed birth-cohort analysis and also found decreasing rates (6). Suvisaari and others adjusted for age and found an age-cohort effect in the decreasing rates of schizophrenia (5), and Takei and others reported that the birth-cohort effect accounted for part of the observed decrease (15). Munk-Jorgensen and Mortensen suggested a possible correlation between decreasing first-admission rates, deinstitutionalization, and excess mortality rates in schizophrenia sufferers (4). In this study, there was little change in the age structure of the population in the Kingston area over the 10 years.

Changes in service delivery over time could contribute to declining rates of schizophrenia. Although it is well recognized that some schizophrenia cases are never treated, it is unlikely that this has contributed significantly to the results, given the comprehensive services in the countries in which the studies have been conducted. Using the ECA study of 5 sites in the US, Norquist and Regier report that, although people with schizophrenia were usually treated in the specialty mental heath sector, annually 14% are seen exclusively by the primary care sector (28). This factor is only relevant to ambulatory care and does not affect reported inpatient rates. A more significant service delivery change has been the move from inpatient to community care within the mental health sector. Der and others reported that, although fewer schizophrenia cases are being admitted, this cannot account for the decreased rates in schizophrenia since the decline in first-admission rates for schizophrenia is 3 times that of other diagnostic groups, while the readmission rates for these groups have remained stable (14). Some authors included more than inpatients in their studies to take into account the move to community care. Munk-Jorgensen and Mortensen included day patients as well as inpatients in their study (4). Eagles and others used all psychiatric contacts, not only admissions, and stated that it is unlikely that general practitioners manage a large number of schizophrenia patients without involving psychiatry (1,12). Geddes and others looked at yearly cohorts of individuals newly diagnosed with schizophrenia in Edinburgh and found that the percentage who were never admitted did not change between 1978 and 1988 (2). In their study, Kendell and others included day patients and outpatients (3). Oldehinkel and Giel found that even the nonsignificant decrease in first-admission rates was neutralized when other mental health services were added (7). In Kingston, psychiatric services for severe mental illness did not undergo significant modifications between 1986 and 1996, and therefore, this factor does not account for the observed decrease in treated prevalence rates. There was no increase in outpatient prevalence rates to explain the decrease in inpatient rates. The observed decline in rates for less severe psychiatric disorders reflects ongoing efforts to reduce admissions for this group of patients. All the reviewers cite changing diagnostic criteria and habits over the last half-century as a major confounding factor in this type of research, and the authors of many studies discuss this issue. Kendell and others state that 1 of the 2 major contributors to falling incidence rates of schizophrenia is changing diagnostic criteria over time (3).

They demonstrated this by applying a computer algorithm of symptoms recorded in charts in 1971 and 1989. Oldehinkel and Giel found no decrease in the incidence of schizophrenia when a broader definition of the disorder was used (7). Folnegovi  and others concluded that the change from ICD-7 to ICD-9 over the course of their study did not influence the recorded incidence rates (8). Nicole and others compared nonvalidated ICD-9 diagnoses of schizophrenia with diagnoses validated by case notes and found a lower rate when they applied their own stricter criteria (16). They concluded that the reported falling rates of schizophrenia may be largely attributable to changing diagnostic criteria. Some have postulated that changes in diagnostic criteria would produce a decrease in schizophrenia with a corresponding increase in other disorders. Munk-Jorgensen and Mortensen found a decrease in first-admission rates distributed across all diagnoses, including functional psychoses, other psychoses, and other nonpsychotic illnesses (4). Der and others found a significant but smaller decrease in the first-admission rates of affective disorder during the same period as an observed fall in schizophrenia rates (14). Eagles and others found no concomitant rise in other diagnostic categories accompanying the decreased rates of schizophrenia (1). In contrast, Jones and others observed a commensurate rise in the rate of bipolar diagnoses over the same period that the rates of schizophrenia declined (6). Suvisaari and others reported a small but significant increase in other nonaffective disorders and concluded that the narrowing clinical concept of schizophrenia may partially explain the observed decrease in rates (5). Stoll and others found large reciprocal shifts in the frequencies of diagnoses of schizophrenia and major affective disorders in North America (17). They attribute part of this to the availability of lithium, the narrowed definition of schizophrenia and broadened category of major affective disorder in the DSM-III, and reimbursement rates in the US favouring affective disorder diagnoses. In the period examined by this study, the diagnostic classification changed from DSM-III to DSM-IV, but the few substantive differences between the 2 systems cannot account for the observed changes in prevalence rates. The decrease in prevalence rates of schizophrenia was not accounted for by changes in other disorders since these either remained the same or decreased over the period of the study.

Other methodological issues include methods of data collection and definitions. Eagles and Whalley suggest that improvements in methods of data collection over time, even within the same setting, might influence the reported rates (12). The definition of what constitutes a first admission has been shown to change over time, and this affects reported incidence rates. Kendell and others found annual error rates for first-admissions for schizophrenia that ranged from 28% to 59% (3). Munk-Jorgensen and Mortensen reported a decrease in the percentage of false-positive first-admissions from 15% to 0% over the course of their study (4). Suvisaari and others state that the data on first-admissions and diagnoses in the Finnish register have been proven to be reliable (5). The timing of the diagnosis of schizophrenia also affects reported incidence rates. In most studies, the diagnosis made at the time of first-admission is used. By adding all patients who ever received a diagnosis of schizophrenia to his sample, Kendell and others increased their cohort by 43% (3). They emphasized the problem of relying on first-admission rates as an index of true inception rates. Munk-Jorgensen and Mortensen used 4 methodologies, ranging from first-admission diagnosis to latest diagnosis, in a fixed observation period (4). This