Both the FC and the LPR methods found that the age and sex adjusted IRR is significantly higher for all migrant groups compared to the native Dutch [for all migrants IRR 1.70 (95% CI 1.30–2.21) for the LPR method and 1.91 (95% CI 1.15–3.25) for the FC].
The IRR for Moroccans was significantly lower in the LPR [IRR 2.69 (95% 2.10–3.41)] than in the FC study [4.81 (3.41–6.68)]. The IRR estimates in the LPR were lower for the Turkish and higher for the Caribbean than in the FC study, but these shifts were not statistically significant.
Interpretation
In one population, the FC identified 254 onsets schizophrenia, and the LPR 843 onsets. The onsets identified only by the LPR had a different mix of migrants than the onsets identified by both methods. The LPR method identified a relatively large number of native Dutch and Turkish onsets with a long DPT, and Caribbeans engaging with mental health services at older ages. The FC method identified mostly migrants with earlier onsets (presenting at earlier ages and with shorter DPT than the native Dutch), which in practice resulted in overinclusion of Moroccans and, to a lesser extent, Turkish migrants.
The evidence on the relation between migration and incidence of schizophrenia is nearly exclusively based on the FC sampling frame [
1,
13]. Danish register studies [
8,
9] have used the LPR method, but in their region had no corresponding FC estimates available for direct comparison. Indirect comparisons of their findings with FC data in other countries [
11,
14] are complicated by methodological differences (e.g. other clinical populations, other migrant groupings).
The evidence on migrant differentials in pathways to diagnosis is difficult to interpret because the social, cultural and health service context vary widely between countries [
15], and because there is no standardized definition of pathways to- and through mental health services. Prior studies have used overlapping concepts such as ‘access to mental health services’ [
16‐
18], ‘duration of untreated psychosis’ (DUP) [
19], ‘negative pathways’ [
20] and (in our study) ‘age at first contact with mental health services’ or ‘duration of prior treatment’.
There is some evidence on migrant differentials in pathways through mental health services. Studies from the UK have reported that people from African descent with a first episode of psychosis (FEP) are more likely than other migrant groups to come into contact with mental health services through negative and adversarial routes [
15,
21]. Similar findings were later reported for Moroccans and Caribbean in Rotterdam [
22] and Amsterdam [
23].
Migrant differentials in pathways through services (sometimes resulting in overinclusion in FC samples) may help explain why FC studies report that certain migrant groups have a very high risk of schizophrenia [
15,
21,
24,
25]. This might be the case for Moroccans in the Netherlands [
3] and Black Africans and Black Caribbean in the UK [
5], because these groups are also known to have more negative (and in our study, shorter/earlier) pathways through services, compared to migrants with a lower risk of schizophrenia, and nonmigrants.
Various mechanisms may explain how migration is related both to a higher risk of schizophrenia and to earlier or shorter pathways through services. Higher levels of stress [
26,
27], related to factors such as social defeat [
28], discrimination [
3,
29] or ethnic density [
30] may not only increase the lifetime risk of schizophrenia, but also lead to earlier onsets and negative pathways. Such ‘precipitated onsets’ could be mediated by social processes related to culture, stigmatization, or (lack of) social support [
20], by causing more dysfunction or modifying the clinical presentation.
Migrant differentials in pathways through care do not necessarily distort schizophrenia IRR estimates, as long as all possible pathways to the index diagnosis are accounted for. This is not a problem for the LPR method. But for some groups in FC studies it may lead to inflated IRR estimates because the FC method over includes groups with early onsets and short DPTs.
Strengths and limitations
The main strengths of our study are that it was conducted in a well defined urban catchment area with a 45% share of migrants, that the FC study used in the comparison meets the highest quality standards [
3,
5,
10,
31], and that the LPR was based on a data warehouse, synchronized every day with data from virtually all mental health services in the catchment area. The longitudinal sampling frame covered all treatment pathways from 1980 to 2009.
Both methods were restricted to treated subjects, and typical limitations of treated incidence studies apply, such as the risk of overinclusion of cases (e.g. due to inmigration of prevalent cases into the catchment area, or diagnostic errors), and the risk of underinclusion (e.g. due to cases avoiding mental health treatment entirely). Sensitivity analyses showed that potential distortions by these factors were likely to be small: very few cases moved into the catchment area shortly before the index diagnosis was made, and the diagnostic process was robust [
1].
Migrant differentials in access to mental health care would affect both methods equally, and therefore, cannot account for the differences observed between them; furthermore, surveys of access to care from different countries [
16‐
18] and meta-analyses of DUP-studies [
19] reported no systematic differentials.
There is evidence that migrants drop out of mental health treatment more frequently than nonmigrants [
32,
33]. Some migrants may have dropped out before the onset of schizophrenia and then been missed by one or both methods. This would deflate the migrant IRR estimate. In the 20–54 working age bracket, access to welfare benefits would be an additional incentive for undiagnosed but disabled schizophrenia patients to reengage with mental health services. These and other cases who reengaged would be listed in the register and ultimately detected as incident cases. They may then have been classified in an older age group.
Cross-cultural diagnostic bias could also have confounded our IRR estimates [
34‐
37]. We did not estimate cross-cultural diagnostic bias directly in the present study. Indirectly, however, we found no migrant differentials in diagnostic validity or stability in either FC or LPR study samples. As noted above, clinicians were not more conservative in diagnosing schizophrenia with native Dutch than with migrant subpopulations.
Unfortunately, we had no reliable data to examine potential confounding from socioeconomic status (SES) at time of onset. In our study (Table
1), the incidence of psychotic disorders for Turkish immigrants was only modestly increased, while they have much lower income, educational and employment levels than Surinamese migrants, whose relative risk was high [
38]. In the literature, the strength and nature of the relation between SES and schizophrenia remains unclear [
38‐
41]. In line with two comparable studies [
42,
43], we expect that adjusting for individual SES in our data would attenuate the migrants’ IRR estimates but not explain them.
Our findings of overinclusion of subjects presenting at younger ages and/or with shorter duration of prior treatment probably apply to all FC studies of schizophrenia (i.e. first episode of schizophrenia or FES), but we have not shown that it applies to studies of all psychoses (i.e. first episode of psychosis, or FEP).
It seems prudent to assume that selective sampling also occurs in FEP studies. To assume otherwise, for migration as a risk factor, would imply that there are no migrant groups with FEP who present at systematically younger ages, or who have systematically shorter DPT, compared to other migrant groups or to nonmigrants. To our knowledge, this hypothesis has not yet been tested directly.
The indirect evidence is mixed. As noted above, Anderson et al. [
15] found that specific migrant groups such as Blacks with FEP had more negative pathways than nonmigrants. High quality FC studies in the UK [
4,
14] and in The Hague have reported migrant IRRs for both FES and FEP, and the patterns were similar. Finally, we speculate that overdiagnosis of psychosis among migrants (diagnostic bias) could translate into earlier diagnosis of psychosis among migrants. There is some evidence that diagnostic bias distorts FEP and FES differently [
29], but the direction and extent of this difference is unclear.