Background
The excess rates of psychotic disorders among migrant and ethnic minority groups have been called a ‘public health tragedy’ (Morgan & Hutchinson, Reference Morgan and Hutchinson2010), and these inequities have persisted for nearly a century with little progress toward prevention. The most recent meta-analytic estimates suggest a more than two-fold greater risk of psychotic disorder among first-generation migrants, with persistence of risk into the second generation (Selten, van der Ven, & Termorshuizen, Reference Selten, van der Ven and Termorshuizen2020).
One factor that may hold promise for improving our understanding of the relationship between migration and psychotic disorders is language. Indeed, language impairments are a key feature of schizophrenia and other psychotic disorders, with evidence for the central role of language spanning from etiology to diagnosis and therapeutics (Covington et al., Reference Covington, He, Brown, Naçi, McClain, Fjordbak and Brown2005). It has been hypothesized that schizophrenia arose as a manifestation of the genetic evolution towards the capacity for language (Berlim, Mattevi, Belmonte-de-Abreu, & Crow, Reference Berlim, Mattevi, Belmonte-de-Abreu and Crow2003), and disruptions in discourse are a key feature of thought disorder (Covington et al., Reference Covington, He, Brown, Naçi, McClain, Fjordbak and Brown2005). Among people who are bilingual, the clinical presentation of psychotic symptoms may differ based on the language used for assessment, with more severe symptoms present when assessments are in the first language (Erkoreka, Ozamiz-Etxebarria, Ruiz, & Ballesteros, Reference Erkoreka, Ozamiz-Etxebarria, Ruiz and Ballesteros2020). Advances in linguistic computational approaches suggest that language abnormalities are a potential biomarker for diagnosis and prognosis of psychotic disorders (De Boer, Brederoo, Voppel, & Sommer, Reference De Boer, Brederoo, Voppel and Sommer2020). Looking further upstream, early life socio-economic conditions also have a pervasive effect across multiple domains of language use in adult life (Rowe, Reference Rowe2018), and key social determinants include social class, ethnicity, migrant status, parental education, and size of a family's social network (Perkins, Finegood, & Swain, Reference Perkins, Finegood and Swain2013; Rowe, Reference Rowe2018). Many of these early life social conditions are also important markers of risk for psychotic disorders, yet the social pathogenesis of psychosis remains poorly understood (Shah, Mizrahi, & McKenzie, Reference Shah, Mizrahi and McKenzie2011).
Although it has been extensively studied for non-psychotic disorders (Montemitro et al., Reference Montemitro, D'Andrea, Cesa, Martinotti, Pettorruso, Di Giannantonio and Tarricone2021), evidence on the role of language as an explanatory factor for psychotic disorders among migrant groups is limited. A recent case-control study by Jongsma and colleagues used a binary indicator that combined fluency in the host country language and linguistic distance (Jongsma et al., Reference Jongsma, Gayer-Anderson, Tarricone, Velthorst, van der Ven, Quattrone and Kirkbride2021), which is defined as the degree of relatedness between a migrant's first language and the dominant language in the host country, and is an important determinant of language acquisition (Isphording & Otten, Reference Isphording and Otten2014). Adjustment for this indicator attenuated the association between migrant status and psychotic disorders, and there was a nearly two-fold greater odds of psychosis among people with linguistic distance and/or low fluency in the majority language (Jongsma et al., Reference Jongsma, Gayer-Anderson, Tarricone, Velthorst, van der Ven, Quattrone and Kirkbride2021). Similarly, we have previously shown that first-generation migrants to Ontario who spoke neither of Canada's official languages (English and French) at arrival had a 13% higher risk of psychotic disorder, relative to people who spoke English at arrival, and this effect was specific to psychotic disorders and not a marker of risk for mood and anxiety disorders among migrant groups (Anderson, Le, & Edwards, Reference Anderson, Le and Edwards2022). Finally, it has also been hypothesized that diglossia may be important to understanding the etiology of psychotic disorders (Alherz, Almusawi, & Barry, Reference Alherz, Almusawi and Barry2019), which refers to a linguistic context where there is a ‘high’ form of language used for more formal communication – such as educational or employment settings – and a ‘low’ form of language that is used for everyday discourse (Schiffman, Reference Schiffman and Coulmas2017). Migration imposes an induced diglossic environment, whereby the first language becomes the ‘low’ form of language, and the dominant language in the host country is used as the ‘high’ form of language in educational and employment contexts (Alherz, Reference Alherz2022). Diglossia has recently been shown to be associated with prodromal symptoms of psychosis among first-generation migrant groups (Alherz, Almusawi, & Alsayegh, Reference Alherz, Almusawi and Alsayegh2022).
To advance knowledge in this nascent field, there is a need for longitudinal research that considers the separate effects of linguistic distance and fluency in the host language, accounting for timing of language acquisition by considering factors such as age at migration (Alherz, Reference Alherz2022; Alherz et al., Reference Alherz, Almusawi and Barry2019). Thus, we sought to explore the association between linguistic distance of the first language and the risk of psychotic disorders among first-generation migrant groups in Ontario, Canada.
Methods
We followed the REporting of studies Conducted using Observational Routinely collected health Data (RECORD) guidelines (Benchimol et al., Reference Benchimol, Smeeth, Guttmann, Harron, Moher and Petersen2015) (online Supplement 1). The data were obtained from ICES, which is an independent, non-profit research institute whose legal status under Ontario's health information privacy law allows it to collect and analyze health care and demographic data, without consent, for health system evaluation and improvement. The use of the data in this project was authorized under section 45 of Ontario's Personal Health Information Protection Act (PHIPA) and does not require review by a Research Ethics Board.
Study design and source of data
We used linked population-based health administrative data from ICES to create a retrospective cohort of first-generation migrants who landed in Ontario over a 20-year period (1992 to 2011), which has been described previously (Anderson et al., Reference Anderson, Le and Edwards2022). The study cohort was predominantly based on data from the Immigration, Refugee, and Citizenship Canada Permanent Resident (IRCC-PR) database, which contains data on migrants who landed in Ontario after 1985 linked to the health administrative data (linkage rate = 86%) (Chiu et al., Reference Chiu, Lebenbaum, Lam, Chong, Azimaee, Iron and Guttmann2016). The IRCC-PR includes information from federal immigration records on the characteristics of migrants at the time of landing, including country of origin, migrant class, first language, and knowledge of Canada's official languages (English and French).
The cohort of first-generation migrants was linked to the health administrative data, which includes information on outpatient physician visits, emergency department visits, and hospitalizations covered under the universal Ontario Health Insurance Program (OHIP). Person-time follow-up began as of the 14th birthday for people aged 0–13 years at the time of migration, and as of the landing date for people aged 14+ years at migration, due to the low risk of psychotic disorder prior to age 14 (<3% of cases) (Solmi et al., Reference Solmi, Radua, Olivola, Croce, Soardo, Salazar de Pablo and Fusar-Poli2022). The cohort was followed to the end of 2016, age 65, death, or the end of OHIP eligibility, which would indicate a move outside of Ontario or emigration from Canada. We had 5–25 years of follow-up available, depending on the landing date. A complete description of the databases and variables used is available in online Supplement 2. These datasets were linked using unique encoded identifiers and analyzed at ICES.
Exposure variable
Language is a highly complex construct made up of different objects (e.g. sounds, words, syntax, etc.), and there are several approaches in the literature for classifying languages into categories reflecting linguistic distance (Jongsma et al., Reference Jongsma, Gayer-Anderson, van der Ven, Quattrone, di Forti, Arango and Kirkbride2022). To identify classification approaches for the purposes of the current study, we sought advice through an international email listserv of linguists (https://linguistlist.org), and received a number of suggestions for classification approaches. Based on the advice we received, we classified first language using one of four approaches:
i. Language Genealogies – Using a similar approach to the prior study by Jongsma et al. (Reference Jongsma, Gayer-Anderson, Tarricone, Velthorst, van der Ven, Quattrone and Kirkbride2021), we operationalized linguistic distance in relation to English using a genealogical language tree, which is based on the classical lexical-etymological method. Each language was assigned the following relatedness score: 0 = first language is English; 1 = first language is on the same ‘branch’ as English within the same language family (i.e. Other Germanic Languages); 2 = first language is on a different branch but within the same language family as English (i.e. Non-Germanic Indo-European Languages); 3 = first language is in a different language family from English (i.e. Non-Indo-European Languages).
ii. Estimated Acquisition Time – The Foreign Service Institute has developed an estimate of the time and difficulty associated with the acquisition of various languages (https://www.state.gov/foreign-language-training/). This approach classifies 66 languages into four categories based on the time required to reach a professional working proficiency. Categories range from I, which are languages closely related to English (e.g. Dutch, Afrikaans) to IV, which represent languages which are exceptionally difficult relative to English (e.g. Japanese, Arabic, Chinese, Korean).
iii. Parametric Comparison Method (PCM) Score – The PCM score calculates the distance between 54 different languages by comparing their properties on 94 binary syntactic parameters (Irimia et al., Reference Irimia, Ceolin, Crisma, Guardiano, Kazakov and LongobardiIn Press). PCM transcends the established genealogical distinctions and provides scores ranging from 0 to 1, with higher scores reflecting greater syntactic distance from English. The distribution of PCM scores was bimodal, therefore we divided the non-zero scores into quartiles, and used English (score = 0) as the reference category in our main analyses.
iv. Automated Similarity Judgement Program (ASJP) – The ASJP calculates the distance between two languages by comparing a set of core vocabulary. The score represents the normalized average of the number of additions, deletions, and substitutions required to transfer a word from one language to another – in other words, the similarity of synonymous words across different families (Wichmann, Holman, & Brown, Reference Wichmann, Holman and Brown2022). Scores ranged from 0 to 105, with higher scores reflecting greater lexical distance from English. The distribution of scores was highly skewed, therefore we divided the non-zero scores into quartiles, and used English (score = 0) as the reference category in our main analyses.
There were nearly 900 different languages, including dialects, listed in the IRCC database, and these were classified by a member of the research team (JAK), with dual coding of a random sample of 10% of the languages by a second member of the team (IW), with a high level of agreement. Some languages were unable to be classified by one or more of the four approaches (range 0.1–37.6%), and there were a small number of languages that were unable to be classified by any approach (e.g. language isolate, sign language). We also obtained information on whether migrants could speak English at the time of arrival (yes/no), which is the dominant language in Ontario (93%) (Government of Canada, 2023). This variable was based on an indicator in the immigration record of whether migrants could speak one of Canada's national languages (English and French), which did not consider level of fluency or language proficiency.
Outcome variable
We followed cohort members in the health administrative data to identify incident cases of non-affective psychotic disorder (schizophrenia, schizoaffective disorder, psychosis not otherwise specified) using a validated algorithm (Kurdyak, Lin, Green, & Vigod, Reference Kurdyak, Lin, Green and Vigod2015) (online Supplement 2). Cases were identified based on the presence of an inpatient hospitalization with a discharge diagnosis of non-affective psychotic disorder, or two outpatient physician visits for non-affective psychotic disorder within a 12-month period. A modified version of this algorithm has been previously validated against medical charts, and found to have high levels of sensitivity (94%) and adequate positive predictive value (62%) (Kurdyak et al., Reference Kurdyak, Lin, Green and Vigod2015).
Other variables
We obtained information on other factors previously shown to be associated with the risk of psychotic disorders among migrants in the study cohort (Anderson et al., Reference Anderson, Le and Edwards2022) (online Supplement 2). Age at migration was classified as infancy (0–2 years), early childhood (3–6 years), middle childhood (7–12 years), adolescence (13–18 years), early adulthood (19–29 years), and adulthood (30+ years) (Anderson & Edwards, Reference Anderson and Edwards2020). Binary sex was classified as male and female. Country of birth was classified as European (including Russia), African (excluding North Africa), Caribbean, South Asian, East Asian (including Southeast Asian), Latin American (including Central and South American), and North Africa & Middle East (Statistics Canada, 2010). Migrant class was categorized as economic, sponsored (including family reunification migrants), and refugee. We also obtained information on post-migration place of residence (urban and rural), and census-based neighborhood income quintile.
Data analysis
We summarized the characteristics of the cohort using counts and proportions for categorical variables, and means and standard deviations for continuous variables. There was a low proportion of the sample with missing data (<6%), and those with missing data were excluded.
We used modified Poisson regression models with robust variance estimators (Zou, Reference Zou2004) to estimate incidence rate ratios (IRR) for the association between each language variable and the risk of developing a psychotic disorder. We first estimated the unadjusted association between each language variable and the risk of psychotic disorder, followed by a partially adjusted model that included age at migration, sex, country of birth, migrant class, rurality of residence, and neighborhood income quintile. Finally, we estimated a fully adjusted model that included all variables from the partially adjusted model, in addition to knowledge of English at the time of arrival.
We conducted subgroup analyses by age at migration (<19 years v. 19 + years), as age is responsible for approximately 30% of the variance in second language acquisition (Granena & Long, Reference Granena, Long, Granena and Long2013). We also stratified our analyses by migrant class (economic, sponsored, and refugee), as economic migrants to Canada must demonstrate proficiency in English or French, whereas sponsored migrants and refugees do not have this requirement. We conducted sensitivity analyses, including (i) removing people whose first language was English from the sample, and using PCM and ASJP scores as a continuous variable; and (ii) restricting the sample to people with complete data across all four language variables and repeating the analyses. This was done to assess the impact of comparing models with different samples in the main analysis.
All analyses were conducted using SAS Version 9.4 (SAS Institute, Cary, North Carolina), and results are presented as IRRs with 95% confidence intervals (CI). There were few differences between the partially adjusted and fully adjusted models, therefore our presentation of the findings will focus on the unadjusted and fully adjusted models. Full parameter estimates from the multivariable models are presented in online Supplements 3 to 6.
Results
Our study cohort included 1 863 803 migrants who landed in Ontario, Canada between 1992 and 2011 (Table 1). The largest proportion of migrants came from South Asian (27.2%) and East Asian (25.6%) countries, followed by migrants from European countries (17.1%) and from North Africa and the Middle East (10.9%). There was a relatively small proportion of migrants from Latin America (6.6%), the Caribbean (5.2%), or from African countries (5.0%). Approximately half of the cohort came as an economic migrant (48.5%) and were over 30 years of age at the landing date (51.8%). Over 60% of the sample spoke English at the time of entry into Canada. We identified 16 771 incident cases of psychotic disorder over the follow-up period, which have been described in detail previously (Anderson et al., Reference Anderson, Le and Edwards2022).
Sociodemographic characteristics of the cohort of first-generation migrant groups who landed in Ontario, Canada between 1992 and 2011 (n = 1 863 803)
Language genealogies
Nearly the entire sample (99.9%) was classified using the classical language genealogy approach grounded in lexical etymologies. In the unadjusted models, all non-English first language categories were associated with a lower risk of psychotic disorder, relative to English (Fig. 1). This difference remained in the fully adjusted model for other Germanic languages (IRR = 0.63, 95% CI 0.50–0.80), however there was no longer a significant effect for non-Germanic Indo-European languages (IRR = 0.98, 95% CI 0.92–1.05). People whose first language was in a non-Indo-European language family had significantly higher rates of psychotic disorder, relative to those whose first language was English (IRR = 1.08, 95% CI 1.01–1.16) (Fig. 1).
Distribution of first languages in the study sample, and results of the Poisson regression models for the four different approaches to classifying linguistic distance on the risk of non-affective psychotic disorder among first-generation migrant groups.
Estimated acquisition time
We were able to classify 87.3% of the sample based on estimated acquisition time of the first language. Nearly all categories of language acquisition time had a lower risk of psychotic disorder in the unadjusted models, relative to English (Fig. 1). In the fully adjusted models, both Category I languages (approximately 600–750 class hours, IRR = 0.88, 95% CI 0.81–0.95) and Category II languages (approximately 900 class hours; IRR = 0.86, 95% CI 0.71–1.04) were associated with lower rates of psychotic disorder, relative to English, although the latter includes the possibility of a null effect. Category III languages (approximately 1100 class hours) were associated with a higher risk of psychotic disorder (IRR = 1.18, 95% CI 1.10–1.27), relative to English, however there was no increased risk for Category IV languages (approximately 2200 class hours; IRR = 0.96, 95% CI 0.88–1.06) (Fig. 1).
Parametric comparison method (PCM) score
PCM scores were available for 62.4% of the study sample. Nearly all quartiles of PCM score were associated with a lower risk of psychotic disorder in the unadjusted models, relative to English (Fig. 1). In the fully adjusted model, languages with the closest syntactic distance to English (i.e. Quartile 1) were associated with a lower risk of psychotic disorder, relative to English (IRR = 0.92, 95% CI 0.85–0.99). We did not observe significant effects for the other quartiles of PCM score in the fully adjusted models (Fig. 1). In sensitivity analyses, where we used the PCM score as a continuous variable, we again did not find significant effects for PCM score in the fully adjusted model (IRR = 1.37, 95% CI 0.94–1.99).
Automated Similarity Judgement Program (ASJP) score
ASJP scores were available for 99.0% of the study sample. In the unadjusted models, all quartiles of ASJP score had a lower risk of psychotic disorder, relative to English (Fig. 1). In the fully adjusted model, languages with the greatest lexical distance from English (i.e. Quartile 4) were associated with a higher risk of psychotic disorder, relative to English (IRR = 1.15; 95% CI 1.06–1.24). We did not observe significant effects for the other quartiles of ASJP score in the fully adjusted models. In sensitivity analyses, where we used the ASJP score as a continuous variable, we found that each one-point increase in ASJP score was associated with a 1% increase in the risk of psychotic disorder (IRR = 1.01, 95% CI 1.01–1.02).
Subgroup and sensitivity analyses
The findings of our subgroup analysis stratified by age at migration (<19 years v. 19 + years) were largely aligned with our main analyses. Of exception, in adjusted analyses the protective effects of having a first language with lower acquisition times were only observed in people 19 years of age and older at migration, and not among those who were under the age of 19 at migration (online Supplement 7). We also conducted subgroup analyses stratified by migrant class (economic, sponsored, refugee). These were again largely aligned with our main analyses, with some evidence of a stronger magnitude of effect among sponsored migrants for some indicators of linguistic distance (online Supplement 8).
We conducted a sensitivity analysis restricted to people with complete data on all four indicators of linguistic distance, and the results were aligned with our main analyses (online Supplement 9).
Discussion
Canada has a large and diverse migrant population from a wide range of geographic locations with different migratory patterns relative to other countries, thus providing a unique perspective on the epidemiology of psychotic disorder among migrant groups. Our analysis of a large, population-based cohort of migrants with detailed information on language enabled us to conduct a prospective analysis of the role of first language on the risk of psychotic disorders, accounting for knowledge of English at the landing date and a range of other markers of psychosis risk. Our findings suggest that linguistic distance of the first language may play some role in the excess risk of psychosis among migrants, although effects were varied across the different indicators of linguistic distance. In addition, the magnitude of effect was small, and marked differences in psychosis risk persisted for African and Caribbean groups after multivariable adjustment (online Supplement 3 to 6). If linguistic distance has an etiological role in psychotic disorder among migrant groups, we might have expected to see a gradient effect with higher psychosis risk with increasing linguistic distance, which was not observed. Furthermore, there may be opposing mechanisms at play – for example, East Asian languages such as Chinese, Japanese, and Korean have some of the highest scores for linguistic distance, but we have previously shown that East Asian groups have lower rates of psychotic disorder than the general population (Anderson et al., Reference Anderson, Le and Edwards2022). Conversely, Caribbean migrant groups have some of the highest rates of psychotic disorder, both in Canada and internationally (Anderson et al., Reference Anderson, Le and Edwards2022; Selten et al., Reference Selten, van der Ven and Termorshuizen2020), and English and other languages with a smaller linguistic distance from English, such as Dutch and Spanish, are predominant in the Caribbean. Thus, if linguistic distance is a contributing factor to the excess risk of psychotic disorder among migrants, the relationship is likely complex and multifactorial, and unlikely to fully account for the elevated rates of psychotic disorders among some migrant groups.
We note that lexico-semantic differences between languages (i.e. genealogical and ASJP approaches) were more influential than syntactic variations (i.e. PSM) from English. As the similarities in vocabulary are more critical for comprehension than syntactic structure per se (Longobardi & Guardiano, Reference Longobardi and Guardiano2009), linguistic factors that impair understanding of the message communicated by a speaker may be a more relevant risk factor for psychosis. Prior research has found an independent effect of fluency or proficiency in the language of the host country on the odds or risk of psychotic disorders among migrant groups (Haasen, Lambert, Mass, & Krausz, Reference Haasen, Lambert, Mass and Krausz1998; Jongsma et al., Reference Jongsma, Gayer-Anderson, Tarricone, Velthorst, van der Ven, Quattrone and Kirkbride2021; Tarricone et al., Reference Tarricone, D'Andrea, Jongsma, Tosato, Gayer-Anderson, Stilo and Morgan2022), whereas others have found no association (Garrido-Torres et al., Reference Garrido-Torres, Alameda, Cristóbal, Padilla, Robles, Canal-Rivero and Ruiz-Veguilla2022). These findings have typically been interpreted from a sociocultural perspective, whereby a lack of proficiency in the language of the host country increases marginalization, impedes social and occupational functioning, and prevents full participation in society (Jongsma et al., Reference Jongsma, Gayer-Anderson, Tarricone, Velthorst, van der Ven, Quattrone and Kirkbride2021). Furthermore, a lack of proficiency in English may increase the likelihood of experiencing racism and discrimination (De Souza, Pereira, Camino, De Lima, & Torres, Reference De Souza, Pereira, Camino, De Lima and Torres2016), which has been shown to increase the risk of psychotic symptoms and disorders among migrant and ethnic minority groups (Bardol et al., Reference Bardol, Grot, Oh, Poulet, Zeroug-Vial, Brunelin and Leaune2020; Pearce, Rafiq, Simpson, & Varese, Reference Pearce, Rafiq, Simpson and Varese2019). We did not find that adjusting for knowledge of English at the landing date had an appreciable impact on our observed estimates, although it was associated with a lower risk of psychotic disorder across all models (online Supplement 3–6).
From a neurobiological perspective, switching between languages engages multiple brain regions and cognitive control plays a crucial role in proficient bilingual language processing. The extent of the impact of switching languages on cognitive performance varies based on task difficulty (Köpke et al., Reference Köpke, Howells, Cortelazzo, Péran, De Boissezon and Lubrano2021; Luque & Morgan-Short, Reference Luque and Morgan-Short2021). The cognitive demands associated with languages that are more distant may lead to a different brain activation pattern (Cargnelutti, Tomasino, & Fabbro, Reference Cargnelutti, Tomasino and Fabbro2022). Difficulties in predicting the communicative intent of a foreign language speaker may increase the computational cost in a social discourse, and reduce proactive processes that normally aid in comprehension (Kuperberg, Ditman, & Choi Perrachione, Reference Kuperberg, Ditman and Choi Perrachione2018). It has been hypothesized that cultural misunderstandings in verbal communication exacerbate experiences of social exclusion and discrimination among migrants, thus leading to stress-associated impacts on dopaminergic neurotransmission (Henssler et al., Reference Henssler, Brandt, Müller, Liu, Montag, Sterzer and Heinz2020).
Our measure of English knowledge was assessed at the time of migration and was limited to a binary indicator that did not consider level of fluency or language proficiency, nor did we consider longitudinal changes in English language capacity after migration. It has been hypothesized that the cognitive impairments that characterize the early course of psychotic illness, in addition to genetic variations (Vaughn & Hernandez, Reference Vaughn and Hernandez2018), could impede second language acquisition, and therefore a lack of language proficiency could represent a pre-migratory marker of vulnerability to psychotic disorder (Montemitro et al., Reference Montemitro, D'Andrea, Cesa, Martinotti, Pettorruso, Di Giannantonio and Tarricone2021); however, prior research did not find that second language acquisition was impaired in people with schizophrenia (Bersudsky, Fine, Gorjaltsan, Chen, & Walters, Reference Bersudsky, Fine, Gorjaltsan, Chen and Walters2005). Evidence from non-psychotic mental disorders suggests that language proficiency may have differential impacts at different time points in the post-migration phase (Montemitro et al., Reference Montemitro, D'Andrea, Cesa, Martinotti, Pettorruso, Di Giannantonio and Tarricone2021), and a lack of dominant language proficiency has been shown to correlate with symptoms of paranoia, particularly persecutory ideation (Thomas, Bentall, Hadden, & O'Hara, Reference Thomas, Bentall, Hadden and O'Hara2017). Future research should explore how second language acquisition and proficiency throughout the post-migration period may impact the risk of psychotic disorder among migrant groups.
Limitations
The most notable limitation of our findings is the high degree of complexity associated with quantifying language-related variations. There is likely extensive heterogeneity in English fluency, pre-migration English exposure, and individual aptitude for second language acquisition among the migrants in our cohort. Furthermore, there is no validated measure to classify linguistic distance, and we used a binary indicator of English language knowledge. We have attempted to minimize the impact of this by exploring multiple approaches to classifying linguistic distance. We were also unable to classify all languages using each approach. We explored the effect of this in sensitivity analyses, with little effect on our findings, but caution should be exercised when interpreting the findings from the PCM scores. Linguistic distance per se is a coarse measure of challenges in comprehension and production that may be faced by migrant groups; sociolinguistic differences (e.g. phonology, vocabulary, and dialectic differences) are critical factors (Palaniyappan, Reference Palaniyappan2021) that we were not able to examine using an administrative database.
Additionally, our classification using estimated acquisition time was based on the time required for English speakers to learn a second language, rather than for people who speak other languages to learn English; we have assumed that the acquisition times would be similar, which may not be valid. Low English proficiency or a lack of translation services may be a barrier to accessing mental health services, which may have led to a detection bias or a differential likelihood of misdiagnosis for some groups. We are missing information on non-physician mental health services, such as psychologists and community mental health providers, as well as important contextual factors, such as same-group ethnic density (Baker, Jackson, Jongsma, & Saville, Reference Baker, Jackson, Jongsma and Saville2021). We also did not adjust for education at landing in our multivariable models, which has been shown to be associated with language acquisition among migrant groups (Chiswick & Miller, Reference Chiswick and Miller2001). Finally, we are unable to identify affective psychotic disorders in the health administrative data due to a lack of specificity in the outpatient diagnostic codes.
Conclusions
There has been little progress toward prevention of the excess rates of psychosis among migrant groups. Our findings suggest that proficiency in the language of the host country may hold some promise for informing our understanding of psychotic disorders among migrant groups. Importantly, this represents a modifiable marker of risk for psychosis that could be the target of public mental health strategies. Further research to elucidate the role of linguistic distance and host language proficiency in the excess risk of psychotic disorders among migrant groups is warranted.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S003329172400117X.
Data availability statement
The dataset from this study is held securely in coded form at ICES. While legal data sharing agreements between ICES and data providers (e.g. healthcare organizations and government) prohibit ICES from making the dataset publicly available, access may be granted to those who meet pre-specified criteria for confidential access, available at www.ices.on.ca/DAS (email: das@ices.on.ca). The full dataset creation plan and underlying analytic code are available from the authors upon request, understanding that the computer programs may rely upon coding templates or macros that are unique to ICES and are therefore either inaccessible or may require modification. The ICES analyst (Britney Le) had full access to the study database.
Acknowledgements
This study was supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health (MOH) and the Ministry of Long-Term Care (MLTC). This study also received funding from Department of Psychiatry at Western University. Kelly Anderson is supported by a Tier 2 Canada Research Chair in Public Mental Health Research. Lena Palaniyappan is supported by the Monique H. Bourgeois Chair in Developmental Disorders and Graham Boeckh Foundation (Douglas Research Centre, McGill University) and a salary award from the Fonds de recherche du Quebec-Sante ́ (FRQS).
This document used data adapted from the Statistics Canada Postal CodeOM Conversion File, which is based on data licensed from Canada Post Corporation, and/or data adapted from the Ontario Ministry of Health Postal Code Conversion File, which contains data copied under license from ©Canada Post Corporation and Statistics Canada. Parts of this material are based on data and information compiled and provided by: Ontario Ministry of Health, and the Canadian Institute of Health Information (CIHI). The analyses, conclusions, opinions, and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement is intended or should be inferred. Parts or whole of this material are based on data and/or information compiled and provided by Immigration, Refugees and Citizenship Canada (IRCC) current to March 17, 2020. However, the analyses, conclusions, opinions and statements expressed in the material are those of the author(s), and not necessarily those of IRCC. We thank the Toronto Community Health Profiles Partnership for providing access to the Ontario Marginalization Index, and Dr Wolfram Hinzen of the Grammar and Cognition Lab at the Universitat Pompeu Fabra (Barcelona) for early discussions on classifying language families.
Competing interests
L. Palaniyappan has received personal fees from Janssen Canada, Otsuka Canada, SPMM Course Limited UK, and the Canadian Psychiatric Association; editorial stipend from the Canadian Medical Association Journals (jpn.ca); book royalties from the Oxford University Press; and investigator-initiated educational grants from Sunovion, Janssen Canada and Otsuka Canada. No other competing interests were declared.
Open access
