Exploring early discontinuation of mental health outpatient treatment: language, demographics and clinical characteristics among migrant populations in Japan

Clinic setting

This study uses electronic health record (EHR) data obtained from migrants who received mental health services between 1 April 2016 and 31 March 2019, at three primary hospitals covering the central areas of the Tokyo-Yokohama metropolitan region that accept international patients residing in Japan.15 At these hospitals, translation services are accessible through on-site medical interpreters in English and Chinese, or via telephone for more than 30 languages, and handheld voice translator devices such as ‘Pocketalk’. Non-profit organisations also provide low-cost or free medical interpreters who could accompany patients to their visits.

Data

Data source

The initial patient cohort consisted of 14 511 patients with psychiatric treatment visits during the study period. Patients with nationality as ‘Japanese’ or unknown were excluded, which resulted in the identified migrant patient cohort of 205 migrants (ages between 11 and 90 years), and the demographic and clinical characteristics of this cohort were described previously.15 The current study used a subsample of the prior study by including only patients aged 18 years and older who received outpatient mental healthcare (N=196) to examine self-discontinuation of treatment and its correlates (figure 1). No finding of the current study was reported previously.

Figure 1

Patient cohort and participant flow diagram.

Analyses

We used descriptive statistics to describe the patients’ demographic and clinical characteristics of the entire cohort, and by treatment continuation status with p values for univariate associations. We combined categories endorsed by 10 or less individuals to optimise sample size for subsequent analyses. To identify predictors of early treatment discontinuation, we conducted multivariable regression models using generalised estimating equations to adjust for hospital site correlations. Predictors selection used a combination of two approaches: (1) a priori guided by the literature and research objective; and (2) statistical grounds using backward selection. Age and sex were designated as a priori variables due to their prominence in the literature of treatment utilisation.8–11 Language was included because it is the primary predictor of interest. We conducted backward selection17 starting with a full model of all the covariates with hospital sites adjusted. We eliminated variables with p≥0.20 as recommended for smaller datasets, starting from the variable with the highest p value, one at a time while retaining the a priori variables. The final model included age, sex (female vs male), language (Japanese vs non-Japanese) and three specific ICD-10 diagnostic groups. The ICD-10 groups included were schizophrenia spectrum disorders (F20–29), mood disorders (F30–39), and neurotic, stress-related and somatoform disorders (F40–48). All covariates had variance inflation factor below 2 suggesting no concern with collinearity. Lastly, we explored interactions between language and each significant predictor. We used a statistical significance threshold of p<0.05 with two-tailed tests and IBM SPSS Statistics V.29.0 for data analyses.

Early treatment discontinuation by language

Of the study sample, 25.5% (n=50; 95% CI: 19.4%, 31.6%) self-discontinued treatment within the first 3 months. Of these 50 patients, 88.0% (n=44) discontinued within the first month, including 68.0% (n=34) who discontinued after one visit. The early discontinuation rates of Japanese-speaking migrants (26.5%; 95% CI: 19.5%, 33.5%) and non-Japanese speakers (22.2%; 95% CI: 10.1%, 34.3%) were statistically similar.

Predictors of early discontinuation: results of multivariable models

In the main effect model (table 2), older age has an average of 2.6% reduced odds per year of age increase for early treatment discontinuation (adjusted Odds Ratio (aOR)=0.974, p=0.016). Migrant patients who had a diagnosis other than a schizophrenia spectrum disorder had nearly four times higher odds of discontinuing treatment early compared with those who had the disorder (aOR=3.996, p=0.012). Those with diagnoses other than neurotic, stress-related and somatoform disorders had twice the odds for early discontinuation (aOR=2.792, p=0.025). We observed several significant interaction effects involving language (table 2). Among Japanese speakers, older patients had lower odds (aOR=0.966, p=0.003) of discontinuing treatment early, and such effect was not observed among the non-Japanese speakers (p=0.794; figure 2). Japanese speakers with diagnoses other than a schizophrenia spectrum disorder had seven times the odds of dropping out (aOR=7.220, p=0.013) when compared with Japanese speakers with schizophrenia diagnoses. Japanese speakers with diagnoses other than neurotic, stress-related and somatoform disorders had nearly three times the odds of early discontinuation (aOR=2.947, p=0.030). Online supplemental figures 1 and 2 provide graphical illustrations of these interactions. No other demographic or clinical characteristics were associated with early discontinuation.

Figure 2

Interaction effect between language and age on early discontinuation (within first 3 months) of mental health outpatient treatment: mean predicted probability of early discontinuation by age and language.

Early discontinuation rates: a narrative comparison

Our observed early discontinuation rate of 25.5% within 3 months of mental health outpatient treatment aligns with reported averages in the literature. It is comparable with the dropout rate of 19.1% (95% CI: 14.7%, 24.6%) among adult refugees in high-income countries,16 and falls within the estimated range between 20% and 47% from psychotherapy trials including both migrant and non-migrant populations.18 19 These observations are consistent with findings from systematic reviews suggesting no differences in treatment dropout rates between migrants and non-migrants.16 20 Nonetheless, the factors that impact treatment retention could be different between migrants and non-migrants.20 Notably, among those who discontinued treatment early, two-thirds did so after the first visit. A deeper understanding of treatment dropouts among migrants is critically needed.

The role of language in early discontinuation

Contrary to our hypothesis, early treatment discontinuation rates were similar between Japanese and non-Japanese speakers. Our study determined language groups based on language assistance used during intake, rather than objectively measuring language proficiency. Neither measuring patients’ Japanese language proficiency objectively nor obtaining self-reported proficiency was a standard clinical practice, and such information was not consistently available from EHR. The language grouping based on ‘observed fluency’ in the study may have masked potential differences between the two groups. Some Japanese speakers who were capable of basic conversation and did not use language assistance might lack the necessary fluency in Japanese to fully engage and communicate effectively with their mental health providers. Documented instances of misdiagnosis among migrant patients in Japan, attributable to inadequate Japanese proficiency and miscommunication, underscore the complexities.14 Miscommunications with clinicians, including difficulties expressing symptom experiences,21 were barriers to treatment engagement and resulted in early dropouts.22

Notably, the non-Japanese speakers in the study received language assistance from a family member or a friend, or a medical interpreter at their treatment visits. The additional social support received may have contributed to better treatment engagement. On the contrary, for some non-Japanese-speaking patients, the reliance on family members or friends who were not professionally trained interpreters could negatively impact communications between patients and clinicians, which may adversely affect treatment engagement and retention.23 These factors might have obscured the association between language and treatment discontinuation.

The absence of a direct association between language and early discontinuation challenges the widely acknowledged belief that language proficiency is a significant barrier to mental healthcare access and utilisation.12 Our findings reveal a moderation effect of language through age and clinical diagnosis in early discontinuation. Notably, our study discovered patient-level factors that increase the vulnerability of Japanese-speaking migrants to early treatment discontinuation. Understanding the nuanced role of language in specific contexts, such as age and diagnosis, remains crucial for tailoring effective treatment retention strategies.

Age and early discontinuation among migrants who spoke Japanese

Our findings revealed a negative association between age and early discontinuation, which was particularly pronounced among Japanese-speaking migrants. Previous psychotherapy research has noted younger patients are at risk of dropping out18 ,24; some research noted heightened social and self-stigma experienced by younger patients.25 For migrant young adults, seeking help for mental health may reinforce mental health stigma leading to dropping out early.26 Moreover, younger patients may perceive a lower need for mental health treatment due to better physical health. Indeed, population-based data showed that younger migrants between ages 20 and 59 years had lower mortality rates when compared with Japanese citizens.27

While prior research has found younger age increases the risks of treatment discontinuation, this association was observed only among Japanese speakers but not non-Japanese speakers. For the non-Japanese speakers, because of their language barrier, they might have to entrust their decision-making about treatment to others, often their family members who were able to speak the language. Similar association between language dependency and medical decision-making has been observed in non-psychiatric settings28 and feelings of disempowerment in healthcare settings were commonly reported in patients with limited English proficiency.29 In contrast, Japanese-speaking migrant patients who are more independent of their treatment-related decision-making because of fewer language barriers may be more vulnerable to the risks related to younger age as described earlier. Regardless of underlying reasons, our findings underscore the vulnerability of younger Japanese-speaking migrants to early treatment discontinuation. This supports the necessity for targeted strategies in engaging younger Japanese-speaking migrants in mental health treatment.

Clinical diagnosis and early treatment discontinuation among Japanese-speaking migrants

Our findings suggested that Japanese-speaking migrants with diagnoses other than schizophrenia spectrum disorders or neurotic, stress-related and somatoform disorders were more likely to self-discontinue treatment early. These associations were not observed among non-Japanese-speaking patients. Among refugees and asylum seekers, a meta-analysis examining dropouts from psychological interventions found no significant association with interpreter use, and indeed there was no significant predictor identified at patient level including clinical diagnoses or treatment-related characteristics.16 Thus, the observed association between clinical diagnosis and treatment dropouts in the context of patients’ language proficiency is novel.

Among the migrant patients in our study, we observed that 8.7% (95% CI: 0%, 20.2%) of the Japanese-speaking cohort with schizophrenia spectrum diagnosis discontinued treatment early, or >90% continued treatment beyond 3 months. The observed treatment dropout rate appeared to be lower than that reported among immigrant groups with these diagnoses (24%).20 With respect to neurotic, stress-related and somatoform disorders, the dropout rate observed among the Japanese speakers was 20.9% (95% CI: 8.7%, 33.1%), which was comparable with the ranges (16–22%) reported in meta-analyses among patients with anxiety disorders.30 Our findings revealed that having a primary diagnosis of schizophrenia spectrum disorders or neurotic, stress-related and somatoform disorders was protective of early dropouts among Japanese-speaking patients. One plausible reason could be related to medication adherence, a robust predictor for treatment continuation in patients with schizophrenia spectrum disorders24 and those with anxiety disorders.30 Japanese speakers might have better comprehension of medication rationale and its significance, and better communication with providers, all of which may contribute to their higher treatment continuation rates.

Japanese-speaking migrants with diagnosis other than neurotic or schizophrenia spectrum disorders had almost three to seven times greater odds of dropping out early. The observed rate showed one in three (34.1%; CI: 25.2%, 43.0%) discontinued early. The impacts of clinical diagnoses on early treatment discontinuation were not observed in non-Japanese speakers. The elevated dropout rate may imply challenges in fully engaging these individuals, despite their ability to communicate in Japanese. Symptoms of schizophrenia, such as excitement, withdrawal and bizarre behaviour, are externally expressed and can be recognised by clinicians through observation. For neurotic, stress-related and somatoform disorders, presentations with physical symptoms such as shortness of breath, palpitations, dizziness, headaches and gastrointestinal discomfort are common, which are more concrete to describe and quantify allowing better progress tracking of symptom relief to engage patients. Other disorders, such as mood disorders, however, are recognised and understood by others through language communication, which exacerbate language barriers. These differences in symptomatology likely contribute to some of the observed results. Further research will be needed to replicate these findings and to explore the reasons for dropouts in relation to language barriers and clinical diagnoses among migrants.

Limitations and future directions

The findings of our study should be interpreted with caution. Specifically, our dataset lacked several measures that could have influenced early discontinuation outcomes, including patients’ language preferences, language proficiency ratings, length of stay in Japan and acculturation. Additionally, the investigation did not account for important individual factors such as education, socioeconomic status, mental health literacy, needs perception and stigma related to mental health, which are recognised correlates of mental health services utilisation.8–11 Nonetheless, our study is among the first quantitative investigations of the predictors of early discontinuation of mental health treatment with consideration of language use, among migrants in Japan presenting with a diverse range of clinical diagnoses. Clinical characteristics such as levels of impairment as well as patient–clinician factors such as therapeutic alliance were not considered in the investigation. The study focused on the Tokyo-Yokohama metropolitan region of Japan, which limits generalisability to other regions or countries. While our study revealed moderation effects from language, we did not identify predictors of early discontinuation among non-Japanese speakers. The small sample size of non-Japanese speakers limits the statistical power. This knowledge gap underscores the need for in-depth interviews with migrants receiving mental health treatment for a deeper understanding of facilitators or barriers to treatment engagement such as concerns with pharmacotherapy.8 For future research to understand language-related factors and their roles in treatment discontinuation and retention among migrant populations, objective measures of language proficiency and language preference should be considered.