Mental Health–Related Discrimination as a Predictor of Low Engagement With Mental Health Services

Service engagement is a broad and multifaceted concept encompassing how individuals interact with health and care services (1). Tait and colleagues (2) have conceptualized engagement in mental health services as having four components: availability, collaboration, help seeking, and treatment adherence. It is important to emphasize that low engagement is not necessarily problematic and may reflect appropriate choices about the degree to which individuals wish to interact with mental health services. Low engagement can, however, contribute to negative outcomes. For example, those who do not attend follow-up appointments are more likely than those who remain in contact to require hospital admission (3), including involuntary admission (4). Not seeking help in a crisis and treatment nonadherence may also lead to such outcomes. Missed appointments may result in increased provider frustration and decreased provider empathy (5), as well as increased costs (6).

Mental health–related stigma and discrimination may contribute to low engagement. A recent systematic review concluded that because most research has focused on initial access to care, more research is needed about how stigma and discrimination affect service engagement (7). This review found no quantitative studies of experiences of discrimination and engagement. There were few such studies of other forms of stigma or discrimination and their impact on engagement. One study showed that perceived discrimination and devaluation was related to lower engagement with psychotherapy services (8), and another did not (9). A third found a negative relationship between self-stigma and psychosocial treatment adherence (10).

In light of this evidence gap, our study aims were to test the hypothesis that experienced discrimination related to mental health is associated with low engagement among adults receiving care from community mental health teams (CMHTs) and to explore the pathways by which experienced discrimination contributes to low engagement with mental health services. We hypothesized a number of potential pathways on the basis of a provisional a priori conceptual model that we developed, including pathways via anticipated discrimination, internalized stigma, mistrust in services, reluctance to disclose information about one’s mental health, and lack of social support. [A figure illustrating the conceptual model is included in an online data supplement to this article.]

Methods

Design

This research is part of the Mental Illness–Related Investigations on Discrimination (MIRIAD) study, a cross-sectional survey of individuals using secondary mental health services in an inner-city location in England (11). The MIRIAD study had primary aims to investigate anticipated and experienced discrimination based on mental illness (11); multiple discrimination (12); and the impact of discrimination and stigma on economic costs (13), suicidality, hospital admission (14), barriers to care, and service engagement. The study reported here focused on service engagement. It received ethical approval by the East of England/Essex 2 Research Ethics Committee.

Recruitment and Sample

Participants were recruited from 14 CMHTs in an inner-city area in England. Inclusion criteria were age at least 18; a clinical diagnosis of depression, bipolar disorder, or schizophrenia spectrum disorder; self-defined black, white, or mixed (black and white) race-ethnicity; in current treatment with a CMHT; sufficiently fluent in English to provide informed consent; and sufficiently well that participation did not pose a risk to health or safety. Clinicians were provided with a list of eligible service users and asked whether the service user was sufficiently well to participate. A letter and reminder were posted to eligible service users inviting them to contact the research team if they were interested in participating. The planned sample size was 200. [Further details are available in the online supplement.]

Data Collection and Measures

Participants were interviewed by research assistants in interviews spread over two sittings (range of one to four sittings) and received £15 per interview. The interview schedule contained a range of measures, and some data were also collected from clinical records and from the service user’s main professional caregiver.

The primary dependent variable of service engagement was measured with the Service Engagement Scale (SES) (2), adapted by our team for service user completion. The SES is a 14-item measure with a possible total score of 42. High scores indicate low engagement. It has four subscale domains: availability, collaboration, help seeking, and treatment adherence. The scale, which was originally developed for completion by clinicians, was completed by both the participant’s main professional health care provider and by the service user. For service users, the wording was changed from “the client” to “I,” with other minor wording changes and the reversal of two items (numbers 10 and 13) to make it suitable for service user completion. [The service user version is available in the online supplement.]

In this study, we elected post hoc to use only the service user–completed version of the SES (SES-SU) because 13% of the clinicians contacted (26 of 156) reported that they knew the service user only “a little.” Nine percent of service users (18 of 197) had had little contact (two or fewer appointments) with their community mental health team in the past 12 months. In addition, data from professionals were missing for 23% of those invited to participate (46 of 202), with consequent reduction in statistical power.

We undertook preliminary validation of the SES-SU in the study sample. Cronbach’s alpha was .75, indicating good internal reliability. As predicted, the SES-SU correlated with service users’ ratings of their therapeutic relationship with their main professional caregiver, which was measured with the Scale to Assess Therapeutic Relationships (15) (ρ=–.56, p<.001). It also correlated with the number of total perceived barriers to care on the Barriers to Access to Care Evaluation (16) (ρ=.28, p<.001) and with the professional-rated SES (ρ=.19, p=.019). Although the latter correlation is low, we would not necessarily expect a higher correlation because service users and providers may have different views of engagement, and one is not necessarily more valid than the other (17). In addition, research on a related concept found a lack of correlation between service users and providers’ views when alliance was measured (18). Furthermore the SES-SU subscale score for availability correlated, as predicted, with data from clinical records on the percentage of scheduled appointments kept by the service user (ρ=–.25, p=.004).

The Discrimination and Stigma Scale (DISC) (19) is an interviewer-delivered measure of experiences of discrimination (“unfair treatment”) in the past 12 months due to having a diagnosis of mental illness. Participants report experiences of discrimination across 21 items, such as employment, parenting, and mental health treatment, on a 4-point Likert scale (not at all, 0; a little, 1; moderately, 2; and a lot, 3). The DISC has good psychometric properties (16). A mean discrimination score is calculated (range 0–3) by adding each item score and dividing by the number of applicable, nonmissing items.

Other measures were the Questionnaire on Anticipated Discrimination (20), Internalized Stigma of Mental Illness Scale (21), Stigma Stress Appraisal (22), Scale to Assess Therapeutic Relationships (15), and Brief Psychiatric Rating Scale (BPRS) (23). Lack of social support was measured with a scale adapted from that used by Brohan and colleagues (24). Single items were used to assess mistrust in mental health services and discomfort disclosing information about one’s mental illness (25,26).

Analysis

We used a structural equation model to examine associations between experienced discrimination and service engagement on the basis of our a priori conceptual model, including the following variables: experienced and anticipated discrimination, internalized stigma, mistrust in mental health services, Stigma Stress Appraisal score, lack of social support, quality of the therapeutic relationship, and discomfort disclosing information about one’s mental illness. The model was adjusted for key sociodemographic and clinical variables: age, race-ethnicity, and symptomatology (BPRS). [Further details about the measures and the analysis are available in the online supplement.]

Results

A total of 4,233 service users were screened for eligibility; 1,345 (32%) were eligible and were invited to participate; and 207 (15%) provided written informed consent having had the study procedures fully explained to them. No differences were found between eligible consenting and eligible nonconsenting service users in terms of diagnoses, age, gender, and race-ethnicity. Five service users were excluded after being interviewed because of incorrect diagnoses or incomplete data, which left 202 participants. Among the 202 participants, diagnoses were as follows: schizophrenia spectrum disorder, N=96, 48%; bipolar disorder, N=41, 20%; and depression, N=65, 32%. Sixty-nine percent (N=139) had had a psychiatric hospital admission. About half (55%, N=110) were female. The mean age was 42 (range 19–67). Fifty-five percent (N=110) reported black or mixed race-ethnicity. Of the 197 participants for whom data on employment were available, 23% (N=46) were employed. Full details of participant characteristics are described elsewhere (11).

Table 1 shows a parsimonious model after pathways were removed when evidence was found that they significantly reduced model fit. Because lack of social support was found to be significantly associated with experienced discrimination but not with service engagement, this pathway was removed from the model. Discomfort disclosing information about one’s mental illness and the Stigma Stress Appraisal score were found to be associated with both anticipated discrimination and with internalized stigma but not with service engagement, and these pathways were also removed from the model. The reported model fit the data well. Results from the model estimated using maximum likelihood were similar (Table 2).

Effects associated with pathways from experienced discrimination to service engagement were summarized into direct effects (unmediated effects) and indirect effects (the effect of experienced discrimination through mediating variables in the model) (Figure 1). Total indirect effects refer to the accumulated indirect effect of experienced discrimination via either mistrust in mental health services and the therapeutic relationship or via internalized stigma. No evidence was found for a direct effect between experienced discrimination and service engagement, and thus the direct pathway was omitted from the model. A comparison of the nested models showed that omission of this direct effect did not significantly reduce overall fit. When the direct effect of experienced discrimination on service engagement was included, the size of the effect was –.138—that is, a 1-unit increase in experienced discrimination resulted in a .138-unit improvement in service engagement, which was a nonsignificant effect.

The total indirect effect of experienced discrimination on service engagement was statistically significant (coefficient=1.055, 95% confidence interval [CI]=.312–2.074, p=.019). This was mainly attributed to the mediated effect of experienced discrimination via mistrust in mental health services and the therapeutic relationship (coefficient=.804, CI=.295–1.558, p=.019). This indicates that a 1-unit increase in experienced discrimination resulted in .804-unit of deterioration in service engagement, which is a small effect (standardized effect=.063). The specific indirect effect of experienced discrimination via internalized stigma was not significant.

Discussion

In this exploratory study, we demonstrated a linear relationship between reported experiences of mental illness–related discrimination and low service engagement. This is the first quantitative investigation of this relationship (7). We found that the pathway between these two variables is not direct. Instead, it is indirect, via mistrust of mental health services and therapeutic relationships. Anticipated discrimination and internalized stigma did not play a significant role in the path model.

No previous studies have investigated mental illness–related discrimination as a predictor of service engagement (7). One study reported a negative association between self-stigma and adherence to psychosocial treatment (10). Other studies in this area have highlighted the complexity of the relationship between internalized stigma and engagement (27–29), which may explain our nonsignificant finding.

In the HIV field, trust in one’s care providers has been shown to be positively associated with more clinic visits and better medication adherence (30). Other research has found an association between anticipated racial discrimination and generalized mistrust in other people (25). Research in the mental health field has shown a relationship between discrimination or stigma and mistrust in health care professionals (14,31). However, as far as we are aware, ours is the first research to demonstrate a pathway from discrimination to mistrust in services and then to low engagement. The importance of therapeutic relationships in promoting engagement is well established (5).

The sample was not sufficiently large to support the inclusion of a wide range of sociodemographic and clinical confounders into the path model. Therefore, the model provides preliminary evidence on pathways from discrimination to lower engagement. It is also a limitation that perceived racial discrimination was not examined. Other limitations are that the study design was cross-sectional, and thus prospective research would be needed to establish the direction of relationships. Future research with larger samples should seek to rectify these limitations. To be included in the study, individuals needed to be currently registered with a CMHT, and thus the sample did not include those who had completely disengaged from services. A relatively small proportion of eligible service users took part in the study, and participants may have been more engaged than eligible service users who declined to take part, although they were similar in terms of diagnoses and sociodemographic factors (11). The eligibility restriction to three diagnostic groups and ineligibility of Asian individuals limited generalizability.

This research included a preliminary validation of a service user version of the SES. Such a measure has some practical benefits and will be a useful tool in studies seeking to assess service users’ perspectives in regard to their own levels of engagement. It is a limitation that this version has yet to be tested on an independent sample and fully validated. Also, because the SES-SU was derived directly from the clinician version, which appropriately taps behavior, the SES-SU focuses on behavior, which was a limitation in this study because it does not include issues such as how service users feel about the care they receive, which can be seen as part of the concept of engagement.

Conclusions

Implications for Research

A key area for further research is investigating how best to intervene to prevent or minimize the experienced discrimination associated with mistrust of services, which in turn was associated with low engagement. We recommend prospective studies to examine causal relationships. Replication of the structural equation modeling with a larger sample to permit the inclusion of potential moderators and other confounders is also required. It would be helpful to establish whether the relationships we found will hold in non–inner-city, non-U.K. contexts, in primary care or psychotherapy settings, and with different diagnostic groups. Qualitative research could help to elucidate the processes underlying the pathway from discrimination to low engagement. We recommend further psychometric work on the SES-SU to replicate our preliminary validation in an independent sample and establish its test-retest reliability and factor structure and its relation to existing similar measures.

Implications for Practice

It is important to highlight that low engagement may reflect something about the services as well as the client (2). Mistrust in mental health services may be a realistic consequence of poor treatment experiences, coercion (32), or discriminatory practices by staff (33). A qualitative study of people with serious mental illness who had completely disengaged from services suggested that services that did not meet service users’ needs and difficulties in relationships with mental health staff (including mistrust) were the main reasons for disengaging, with stigma rarely mentioned as a reason (34). The main strategy for increasing engagement with services suggested by this group was to offer caring, noncritical listening (34). A recent systematic review found that collaborative care was an effective way to increase engagement (35). Our finding that service users’ perceptions of the quality of their relationship with the primary professional caregiver predicted engagement underlines the vital importance of developing and maintaining good relationships with service users (36). Each of the aforementioned aspects is likely to reduce mistrust in services, a variable that we, and others (5,37), found to be important in the pathway to low engagement.

It will also be important to intervene at earlier points in the pathway by decreasing the experienced discrimination that may contribute to mistrust of services and poor therapeutic relationships. Campaigns and programs aimed at the public and specific target groups to end stigma and discrimination can reduce experienced discrimination (38). Challenging and changing structural discrimination (39), such as discriminatory policies, underresourcing of mental health services, and negative media reporting, may also ultimately have a secondary effect in levels on engagement. Eliminating any stigmatizing or discriminatory attitudes and practices within mental health services is also advocated (32,40), particularly because service users identify this as one of the more common life areas for experiencing discrimination (11,12,41).