Characteristics of early- and late-diagnosed schizophrenia: implications for first-episode studies

Schizophrenia Research 33 (1998) 27–34

Characteristics of early- and late-diagnosed schizophrenia: implications for first-episode studies Ger Driessen a, Nicole Gunther a, Maarten Bak a, Margo van Sambeek a, Jim van Os a,b,* a Department of Psychiatry and Neuropsychology, European Graduate School of Neuroscience, Maastricht University, Maastricht, The Netherlands b Department of Psychological Medicine and Social, Genetic and Developmental Research Centre, Institute of Psychiatry, London, UK Received 12 January 1998; accepted 15 May 1998

Abstract First-episode studies of schizophrenia are being carried out in many places. However, previous work has suggested that only half of the patients with schizophrenia receive the diagnosis in the initial stages of the illness. We examined whether cases of early- and late-diagnosed schizophrenia differed with respect to key sociodemographic characteristics and indicators of service use that might bias first-episode studies. Individuals who (i) presented for the first time between 1983 and 1993 to psychiatric services in a defined urban area with a cumulative mental health case register; and (ii) received a diagnosis of schizophrenia at least once during their mental health career were identified (n=186). This sample was divided into those who received the diagnosis of schizophrenia for the first time within the first year of service contact (early-diagnosed schizophrenia; EDS), and those who received it for the first time after the first year of service contact ( late-diagnosed schizophrenia; LDS ). The 10-year incidence of EDS and LDS were 10.4 and 7.0 per 100 000 person-years, respectively. EDS and LDS did not differ in their pattern of association with sex, single marital status and higher levels of neighbourhood socioeconomic deprivation. However, EDS was more incident in the higher age groups, and the level of service use was higher for EDS cases in the first years of contact with mental health services, with LDS cases gradually catching up and exceeding EDS service use later in the illness course. Although differences between EDS and LDS were few, studies of patients with schizophrenia in the ‘first’ episode are likely to be most representative if patients who receive the diagnosis for the first time after previous episodes of care for non-schizophrenic episodes are also included. © 1998 Elsevier Science B.V. All rights reserved. Keywords: Schizophrenia; First-episode studies; Diagnosis; Case register; Incidence

1. Introduction A recent UK study found that only half of the patients who ever received a diagnosis of schizo-

* Corresponding author. Tel: ++-31-43-3299773; Fax: ++-31-43-3299708

phrenia did so at the first episode (Goldacre et al., 1994). Because the incidence of schizophrenia varies with age, gender, marital status and neighbourhood level of deprivation (Castle et al., 1993a,b; Van Os et al., 1995, 1997; Dauncey et al., 1993), the question arises whether early and late diagnosed schizophrenia cases differ from each other with regard to important demographic char-

0920-9964/98/$ - see front matter © 1998 Elsevier Science B.V. All rights reserved. PII: S0 9 2 0 -9 9 6 4 ( 9 8 ) 0 0 05 9 - 0

28

G. Driessen et al. / Schizophrenia Research 33 (1998) 27–34

acteristics that can confound aetiological and treatment research, and limit their generalizability. We examined cumulative case register data in an urban area to examine whether (i) around half of the individuals who ever received a diagnosis of schizophrenia would be diagnosed as such in the later stages of their illness; (ii) late-diagnosed and early-diagnosed individuals differed from each other with respect to key sociodemographic variables; and (iii) late-diagnosed and early-diagnosed individuals differed from each other with respect to mental health service use.

2. Methods The Maastricht Mental Health Case Register (MHCR; Hamers et al., 1986) cumulatively has collected, since 1981, data on all mental health contacts (psychiatric hospital, community mental health centre (CMHC ), psychiatric department of university hospital, community psychiatric emergency outreach team, psychogeriatric nursing homes, sheltered housing, child psychiatric services, services for the mentally impaired, alcohol and drug misuse services) and demographic and diagnostic data in a region of around 200 000 population (city of Maastricht: 120 000; surrounding areas: 80 000). For the current study, registered contacts with child psychiatric services, alcohol and drug misuse services (which are separate from general psychiatric services in the Netherlands and for which data were incomplete over the period of investigation), and services for the mentally impaired and for patients with dementia were excluded. Maastricht is a relatively small city (population 120 000), located in the extreme south of the Netherlands in the province of Limburg. There are strong local traditions and Limburg has its own, officially recognized, language. Compared with the densely populated and more industrialized areas of the north-west of the country, levels of immigration of foreign nationals over the past decades have been low. Mental health services are covered by a national insurance and a GP referral is not necessary for attending the CMHC. Access

to mental health services does not depend on the neighbourhood level of deprivation. The period of investigation for the current study was 1981–1993. Four-dimensional population data in the age range 16 years and older (age, sex, marital status and neighbourhood) for this period were obtained from the municipal authorities for each of the years of the period under investigation. The case sample for the current study was defined by four criteria: (i) age 16 years or older; (ii) having been coded as living in the city of Maastricht at all mental health contacts; (iii) an ICD9 diagnosis of schizophrenia and related disorders (ICD9 295.x and 297.x) recorded at least once during a psychiatric career; (iv) in order to skew the sample towards new cases of schizophrenia, subjects registered in the first two years of the register (many of whom would have been prevalent cases that were in treatment when the register opened ) were excluded, leaving subjects registered over the period 1983–1993 (2-year incident sample). By confining the analyses to the city of Maastricht, with the exclusion of the surrounding villages, any effect of distance to psychiatric services was minimized, as within the city of Maastricht all distances to mental health services can easily be covered by bicycle. The 2-year incident sample might still contain some cases who had first contact with services prior to the Register’s start in 1981, but were erroneously registered as incident in 1983 or later after a long period of remission without service contact. As this could bias the findings, consistency of the results was checked in a smaller, much more ‘incident’ sample of 10-year incident cases (cases presenting between 1991 and 1993 with no contacts in the period 1981–1990). Analyses were also conducted with the sample restricted to patients under the age of 66 years, in order to examine possible bias due to diagnostic misclassification of older patients presenting with psychotic symptoms in the context of an early, and not yet recognized, dementing illness. 2.1. Early-diagnosed and late-diagnosed schizophrenia Patients who received a diagnosis of schizophrenia within a year after first contact with ser-

G. Driessen et al. / Schizophrenia Research 33 (1998) 27–34

vices constituted the group of patients with earlydiagnosed schizophrenia ( EDS). Patients who received a diagnosis of schizophrenia for the first time more than a year after first contact with services comprised the group of patients with latediagnosed schizophrenia (LDS ). Of course, for the purpose of calculation of incidence rates, EDS and LDS were considered incident at the time of first contact with services. Time-dependent demographic characteristics (age, marital status) were also taken at first contact for both EDS and LDS cases. 2.2. Demographic variables The register routinely collects information on age, sex, marital status and neighbourhood. In order to classify neighbourhoods according to level of deprivation, we requested from the municipal authority information on six socioeconomic variables to characterize the neighbourhoods of Maastricht over the period 1991–1992: (1) number of persons dependent on unemployment benefit per 1000 population in the age range of the economically active; (2) number of persons dependent on social welfare benefit per 1000 population in the age range of the economically active; (3) the number of single parent families per 1000 families; (4) the number of non-voters per 1000 population in the voting age range; (5) the number of foreign born per 1000 population; (6) the rate of migration (moving in and moving out) per 1000 population. Very small neighbourhoods, or neighbourhoods consisting mainly of industrial compounds (n=5) were excluded, leaving 36 neighbourhoods with a median population size of 2685 in 1992 (interquartile range: 1706–4203). Principal component (PC ) analysis of these six variables yielded two main components with an eigen value greater than unity, and characterized by a clear elbow in the scree plot. Together, the two PCs explained 75% of the variance. The first component (PC1) explained 45% of the variance and had high loadings (≥0.5) in the proportion

29

of single parent families and the proportion of foreign born; the second component (PC2) had high loadings in the proportion of long-term unemployed and the proportion of non-voters. The PC scores served as exposures in the analyses and will hereafter be referred to as deprivation scores. 2.3. Analyses 2.3.1. Associations between incidence rates and demographic characteristics Incidence rates in the population aged 16 years and older were analysed in relation to sex, age (5-year age groups), marital status (married, single, divorced, widowed ) and neighbourhood level of deprivation using the Poisson regression procedure in the STATA statistical programme (STATACorp, 1995). In Poisson regression, the logarithm of the rate is the dependent variable, and the effect of the exposure variable (demographic variables) on the rate can be examined. The exponentiated regression coefficient for a given exposure variable equals the incidence rate ratio (RR), i.e. the rate in the ‘exposed’ group divided by the baseline rate. The two continuous PCs of neighbourhood deprivation were divided by their tertiles in order to assess linear trends in their associations with incidence rates (Breslow and Day, 1980), adjusting for age and sex. 2.3.2. Service use comparisons In order to compare service use in EDS and LDS patients, we calculated the Service Consumption Score (SCS) for each patient. The service consumption score ( Tansella et al., 1986) is a simple weighted measure of the intensity of service use, calculated using the following formula: ((in-patient days×3) + (day-patient days×2) + (out-patient days×1)). Because the length of the possible follow-up period differed for patients according to the year of their first contact, ranging from 10 years (first contact 1983) to 1 year (first contact 1992), the SCS was calculated for each patient for the first year after first contact and then through yearly extended periods up to the maximum follow-up period. Subsequently, comparisons of SCS between EDS and LDS were conducted for these progressively longer intervals.

30

G. Driessen et al. / Schizophrenia Research 33 (1998) 27–34

For example, for patients whose first contact occurred in 1983, the SCS was calculated for the period of 1 year after first contact, 2 years, 3 years and so on until the maximum period of 10 years, whereas for a patient whose first contact occurred in 1991, the SCS was calculated for the first year and the first two years. SCS comparisons between EDS and LDS were compared for each of the ten SCS periods: 1 year, 2 years, 3 years and so on until the maximum period of 10 years. Associations were expressed as the odds ratio from the logistic regression model with type of schizophrenia ( EDS/LDS ) as the dependent variable, and the service consumption score, divided by its tertiles in order to reduce the effect of outliers and facilitate interpretation of the effect size, as the independent variable. Comparisons between EDS and LDS were also conducted for (i) the amount of time, divided by its tertiles, between the first and the second episode of care, an episode being defined as a period of care preceded and followed by at least 90 days of no contact with services (see Tansella et al., 1995; Tricot, 1986); and (ii) the amount of time in care during the first episode, divided by its tertiles.

3. Results A total of 186 individuals who had their first contact between 1983 and 1993 received a diagnosis of schizophrenia. One-hundred-and-twelve of these (60%) were EDS. The 10-year total schizo-

phrenia incidence was 17.4 per 100 000 personyears; the 10-year incidences of EDS and LDS were 10.4 and 7.0 per 100 000 person-years, respectively. 3.1. Associations with age, sex and marital status The pattern of associations between EDS and LDS on the one hand, and demographic variables on the other (adjusted for each other and for deprivation score) revealed similarities and dissimilarities ( Table 1). EDS, but not LDS, was associated with older age at first contact with services ( EDS: RR=1.19, 95% CI: 1.12–1.26; LDS: 1.04, 95% CI: 0.96–1.12). There was no evidence that the relationship was extra-linear, as addition of the squared age term did not result in an improvement of the model ( p>0.2). Thus, median age in EDS was 50.5 (range: 17–93), whereas it was 35 (range: 16–88) in LDS, a highly significant difference ( p<0.001). If cases older than 65 years were excluded, the age difference was attenuated ( EDS: median age: 36 years, range 17–64; LDS: median 31 years, range 16–62; p= 0.01). Both EDS and LDS were strongly associated with single marital status ( EDS: RR=3.24, 95% CI: 1.97–5.34; LDS: RR=4.32, 95% CI: 2.23–8.39), and to a lesser extent with being divorced (EDS: 1.98, 95% CI: 0.88–4.48; LDS: 3.82, 95% CI: 1.57–9.32). Neither EDS nor LDS were associated with gender. For both EDS and LDS, significant interactions between gender

Table 1 Associations between demographic variables and early-diagnosed (EDS) and late-diagnosed schizophrenia (LDS) expressed as incidence rate ratiosa with 95% confidence intervals Demographic variable

EDS

Older age Female sex Married Single Divorced Widowed Men: single versus other Women: single women versus other

1.19 1.16 1b 3.24 1.98 1.43 6.13 1.57

LDS (1.12–1.26) (0.76–1.79) (1.97–5.34) (0.88–4.48) (0.75–2.74) (2.84–13.22) (0.82–3.02)

1.04 0.89 1b 4.32 3.82 1.13 7.15 1.70

(0.96–1.12) (0.54–1.47) (2.23–8.39) (1.57–9.32) (0.31–4.15) (2.74–8.67) (0.73–3.97)

aAssociations with demographic variables were adjusted for small area deprivation and for each of the other demographic variables. bReference category.

31

G. Driessen et al. / Schizophrenia Research 33 (1998) 27–34

and marital status were apparent ( EDS: LRS= 10.4, df=1, p<0.01; LDS: LRS=5.7, df=1, p= 0.02), in that the risk-increasing effect of single marital status was largely confined to men. The risk for schizophrenia was increased with a factor of 6–7 in single men, but was not significantly increased for single women (Table 1). 3.2. Associations with neighbourhood deprivation scores The incidence of EDS was higher in deprived areas, but the association between EDS and both deprivation scores failed to reach statistical significance (RR linear trend deprivation score one: 1.19, 95% CI: 0.94–1.51; RR linear trend score two: 1.16, 95% CI: 0.92–1.46; Table 2). The associations between both deprivation scores and LDS were larger and statistically significant (RR linear trend deprivation score one: 1.37, 95% CI: 1.02–1.84; RR linear trend score two: 1.33, 95% CI: 1.00–1.77; Table 3), but not significantly greater than the associations between the deprivation scores and EDS (as indicated by overlapping confidence intervals). 3.3. Results in the 10-year incident sample Although statistical power of the analyses with the 10-year incident sample was limited due to the

small number of cases (n=55), the pattern of results was the same. The associations between LDS and neighbourhood deprivation scores were of similar magnitude as in the 2-year incident sample (PC1: RR=1.38, 95% CI: 0.61–3.10; PC2: RR=1.36, 95% CI: 0.63–2.98), though not statistically significant due to the small sample size. No associations existed between EDS and deprivation scores (PC1: RR=0.90, 95% CI: 0.6–1.25; PC2: 1.13, 95% CI: 0.78–1.63). Older age was associated with EDS (RR=1.11, 95% CI: 1.03–1.20), but not with LDS (RR=0.94, 95% CI: 0.78–1.12). Similar large associations existed with single marital status (EDS: RR=5.76, 95% CI: 2.56–12.94; LDS: RR=6.00, 95% CI: 1.00–36.30) and being divorced ( EDS: RR=3.40, 95% CI: 1.06–10.90; LDS: RR=8.08, 95% CI: 1.12–58.10). Additional restriction of the 10-year incident sample to patients aged less than 66 years also did not change the pattern of results. 3.4. Service consumption patterns of EDS and LDS The pattern of service consumption (adjusted for age in view of the contrasting associations between EDS and LDS on the one hand, and age on the other) over progressively longer periods discriminated between EDS and LDS (Fig. 1). EDS patients had much higher service consumption in the first years of the illness career, but the

Table 2 Treated incidence, relative risks, and neighbourhood characteristics for early-diagnosed schizophrenia Principal component

lowest tertile

middle tertile

highest tertile

PC1

26 8.5

40 10.6

46 11.7

1e 36 9.2

1.21 (0.74–1.99) 40 10.1

1e

1.09 (0.69–1.70)

PC2

number of cases raw incidence (per 100 000 person-years) adjusted rate ratioa (95% CI d) number of cases raw incidence (per 100 000 person-years) adjusted rate ratioa

summary RRa linear trendb

summary RRc linear trendb

1.42 (0.88–2.30) 36 12.5

1.19 (0.94–1.51)

1.12 (0.86–1.46)

1.34 (0.84–2.13)

1.16 (0.92–1.46)

1.09 (0.84–1.41)

aAdjusted for age and sex. bSummary rate ratio linear trend: the increase in risk associated with moving one unit of the deprivation exposure variable cAdjusted for age, sex and marital status. d95% confidence interval: a confidence interval not including unity denotes statistical significance. eReference category.

32

G. Driessen et al. / Schizophrenia Research 33 (1998) 27–34

Table 3 Treated incidence, relative risks, and neighbourhood characteristics for late-diagnosed schizophrenia Principal component

lowest tertile

middle tertile

highest tertile

PC1

15 4.9

23 6.1

36 9.2

1e 22 5.6

1.27 (0.66–2.43) 24 6.1

1e

1.10 (0.61–1.94)

PC2

number of cases raw incidence (per 100 000 person-years) adjusted rate ratioa (95% CI d) number of cases raw incidence (per 100 000 person-years) adjusted rate ratioa

summary RRa linear trendb

summary RRc linear trendb

1.84 (1.01–3.36) 28 9.8

1.37 (1.02–1.84)

1.43 (1.03–2.00)

1.74 (1.00–3.04)

1.33 (1.00–1.77)

1.31 (0.95–1.79)

aAdjusted for age and sex. bSummary rate ratio linear trend: the increase in risk associated with moving one unit of the deprivation exposure variable. cAdjusted for age, sex and marital status. d95% confidence interval: a confidence interval not including unity denotes statistical significance. eReference category.

difference with LDS patients progressively ( pvalue for linear trend <0.001) decreased so that service consumption for EDS and LDS had become similar in the later years of their illness career ( Fig. 1). EDS patients spent much more time in care during the first episode (age-adjusted OR over the tertiles of time in care in the first episode: 1.60, 95% CI: 1.13–2.37), and spent much less time out of care between the first and the second episode (age-adjusted OR over tertiles: 3.33, 95% CI: 2.14–5.18).

4. Discussion Few sociodemographic contrasts were observed between EDS and LDS, although EDS cases tended to be older. Notably, evidence for interaction between gender and single marital status was apparent for both conditions, single men having a 6- to 7-fold increased risk, whereas single women were not more likely to develop either EDS or LDS. EDS and LDS had a contrasting pattern of associations with service use variables. EDS cases had a more intensive level of service use early in the course of the illness, with LDS patients gradually catching up and even exceeding EDS in later years. 4.1. Methodological issues

Fig. 1. Association between service consumption and timing of psychiatric diagnosis expressed as odds ratios (OR). OR<1: higher service consumption early-diagnosed schizophrenia. OR>1: higher service consumption late-diagnosed schizophrenia.

Diagnostic precision at the level of a case register is limited. However, the broad definition used in the current study would have reduced diagnostic misclassification. Results in the 2-year incident sample might be biased by the presence of non-incident cases of schizophrenia, whose first contact was before 1981 but who were recorded as incident in 1983 or later after a very long period of no contact with services. Such a bias might be invoked to explain, for example, the association between EDS and older

G. Driessen et al. / Schizophrenia Research 33 (1998) 27–34

age. We were able to rule out such bias, however, by showing that results were similar in a sample of 10-year incident cases, which is unlikely to be substantially contaminated with prevalent cases. Not all cases of schizophrenia are treated by mental health services. The combined inception rate of EDS and LDS was 17.4 per 100 000 personyears, which is slightly higher than the rates of 15.1 in men and 11.4 in women reported by Goldacre et al. (1994). This difference between the two studies may be related to our use of a broader definition of schizophrenia, and does not suggest that many cases were missed by the Register. In spite of the relatively long period of investigation, the number of cases of schizophrenia was small and the statistical precision low. This may have obscured differences between EDS and LDS that were not picked up by this study. Our findings should be interpreted in the light of our decision to include all individuals aged 16 years and older with a diagnosis of schizophrenia. Some previous studies have excluded older individuals with psychosis but such an arbitrary selection has been criticized on epidemiological grounds (Castle and Murray, 1993), and current diagnostic systems do not recognize an age cut-off for the diagnosis of schizophrenia (e.g. APA, 1994). Furthermore, exclusion of older individuals did not change the pattern of results, although the age differences between LDS and EDS were attenuated. Contacts with alcohol and drug misuse services could not be included in the analyses. However, individuals with alcohol/drug problems and comorbid psychopathology are referred to mental health services and would thus not have been missed. We can nevertheless not exclude differences between EDS and LDS regarding alcohol and drug problems prior to the onset of psychopathology. 4.2. The findings Our study confirmed earlier findings that a substantial portion of subjects with a diagnosis of schizophrenia receive the diagnosis later on in their illness career (Goldacre et al., 1994). This indicates that the diagnosis of schizophrenia is a dynamic

33

process, and that first-episode studies of schizophrenia only include a selection of cases whose symptoms and early course allow for a diagnosis to be made in the initial stages of illness. Both the early- and the late-diagnosed cases in our study had typical schizophrenia correlates, such as the strong association with single marital status (Hare, 1956; Van Os et al., 1995), and the interaction between gender and single marital status ( Tien and Eaton, 1992; Riecher-Ro¨ssler et al., 1992). Similarly, there was no evidence for a sex difference in the incidence of either EDS or LDS, which is what would have been expected given the broad ICD9 definition of schizophrenia and related disorders used in this study (Castle et al., 1993b). The association between schizophrenia mental health contact rates and level of small-area deprivation is well established (Jarman et al., 1992; Thornicroft, 1991; Harrison et al., 1995; Boardman et al., 1997). Both EDS and LDS were positively associated with neighbourhood level of deprivation, although this was statistically significant only in the LDS cases. EDS was associated with older age and more intensive service use in the early stages of the illness, whereas LDS cases displayed more intensive use of services later in the course of their illness. The relationship with age in EDS cases, although not exponential in nature, resembles the finding reported by Tien and Eaton (1992) in a one-year follow-up of a population sample aged 18 years and older. The association with older age in combination with more intensive service use early in the course in EDS may indicate that for LDS cases there is less delay between illness onset and first service contact, because of, for example, non-specific or prodromal symptoms that initially require less intensive service use than EDS cases presenting with more full-blown schizophrenic illness. However, other explanations are also possible, such as diagnostic variability related to factors that were not measured in the present study, and differences in the rate of comorbidity. 4.3. Conclusion The evidence presented here suggests that the sociodemographic profile of both EDS and LDS

34

G. Driessen et al. / Schizophrenia Research 33 (1998) 27–34

is compatible with that reported in the literature on schizophrenia. There appear to be few differences between patients with an early and patients with a late diagnosis of schizophrenia. However, patients diagnosed with schizophrenia early in the course of their illness were more incident in the older age groups, and initially had a higher rate of service use. Thus, studies of patients with schizophrenia in the ‘first’ episode are likely to be most representative if patients who receive the diagnosis for the first time after previous episodes of care for non-schizophrenic episodes are also included.

References American Psychiatric Association, 1994. Diagnostic and Statistical Manual of Mental Disorders, 4th edn. (DSM-IV ). APA, Washington, DC. Boardman, A., Hodgson, R., Lewis, M., Allen, K., 1997. Social indicators and the prediction of psychiatric admission in different groups. Br. J. Psychiatry 171, 457–462. Breslow, N., Day, N., 1980. Statistical Methods in Cancer Research, Vol 2. WHO, Lyon. Castle, D.J., Murray, R.M., 1993. The epidemiology of lateonset schizophrenia. Schizophr. Bull. 19 (4), 691–700. Castle, D.J., Scott, K., Wessely, S., Murray, R.M., 1993a. Does social deprivation during gestation and early life predispose to later schizophrenia? Soc. Psychiatr. Psychiatr. Epidemiol. 28 (1), 1–4. Castle, D., Wessely, S., Murray, R., 1993b. Sex and schizophrenia: effects of diagnostic stringency, and associations with premorbid variables. Br. J. Psychiatry 162, 658–664. Dauncey, K., Giggs, J., Baker, K., Harrison, G., 1993. Schizophrenia in Nottingham: lifelong residential mobility of a cohort. Br. J. Psychiatry 163, 613–619. Goldacre, M., Shiwatch, R., Yeates, D., 1994. Estimating incidence and prevalence of treated psychiatric disorders from

routine statistics: the example of schizophrenia in Oxfordshire. J. Epidemiol. Comm. Health 48, 318–322. Hamers, H.J.F.R., Romme, M.A.J., De Vries, M.W., 1986. Resistance, privacy and technology: comments on the negotiations of the case register in Maastricht. In: Horn, G.H.M.M., Giel, R., Gulbinet, W.H., Henderson, J.H. ( Eds.), Psychiatric Case Registers in Public Health. Elsevier, Amsterdam. Hare, E.H., 1956. Mental illness and social conditions in Bristol. J. Mental Sci. 102, 349–357. Harrison, J., Barrow, S., Creed, F., 1995. Social deprivation and psychiatric admission rates among different diagnostic groups. Br. J. Psychiatry 167, 456–462. Jarman, B., Hirsch, S., White, P. et al., 1992. Predicting psychiatric admission rates. Br. Med. J. 304, 1146–1151. Riecher-Ro¨ssler, A., Fatkenheuer, B., Lo¨ffler, W., Maurer, K., Ha¨fner, H., 1992. Is age of onset in schizophrenia influenced by marital status? Soc. Psychiatr. Psychiatr. Epidemiol. 27, 122–128. STATACorp, 1995. STATA Statistical Software: Release 5.0. College Station, Texas. Tansella, M., Micciolo, R., Balestrieri, M., Gavioli, I., 1986. High and long-term users of the mental health services. Soc. Psychiatry 21, 96–103. Tansella, M., Micciolo, R., Biggeri, A., Bisoffi, G., Balestrieri, M., 1995. Episodes of care for first-ever psychiatic patients. A long-term case-register evaluation in a mainly urban region. Br. J. Psychiatry 167, 220–227. Thornicroft, G., 1991. Social deprivation and rates of treated mental disorder. Developing statistical models to predict psychiatric service utilisation. Br. J. Psychiatry 158, 475–484. Tien, A.Y., Eaton, W.W., 1992. Psychopathologic precursors and sociodemographic risk factors for the schizophrenia syndrome. Arch. Gen. Psychiatry 49, 37–46. Tricot, L., 1986. An example of the standardization of terminology defining episodes of care. In: Horn, G.H.M.M. ten Giel, R., Gulbinet, W.H., Henderson, J.H. (Eds.), Psychiatric Case Registers in Public Health. Elsevier, Amsterdam. Van Os, J., Takei, N., Castle, D., Wessely, S., Der, G., Murray, R., 1995. Premorbid abnormalities in mania, schizomania, acute schizophrenia and chronic schizophrenia. Soc. Psychiatr. Psychiatr. Epidemiol. 30, 274–279. Van Os, J., McKenzie, K., Jones, P., 1997. Cultural differences in pathways to care, service utilisation and treated outcomes. Curr. Opin. Psychiatry 10, 178–182.