Outcomes of a Psychoeducational Intervention to Reduce Internalized Stigma Among Psychosocial Rehabilitation Clients

Stigmatization of mental illnesses is common (1–7) and is associated with myriad negative effects for people with mental health problems. These effects include opportunity loss, demoralization, anger, isolation, and reluctance to be associated with mental health care (2–12). Efforts to reduce stigma have proliferated (13–18) and have had some success (14,17,19). Yet change has been slow and uneven, and stigmatization of mental illness remains pervasive.

As with other prejudices, stigma regarding mental illness may be absorbed into one’s thinking about oneself. When stigma is internalized, it has additional, manifold negative effects (9,20–23), including increased depression and demoralization (20,24–26), avoidant coping (20,24,27–29), exacerbated ambivalence about mental health care (22,28,30,31), and eroded hope and self-esteem (27,30,32–34). Resources are needed to help people protect themselves from the effects of stigma and discrimination, which persist despite efforts to eliminate them.

One response to self-stigma has been the development of interventions to reduce levels of internalized stigma (35,36). Most are group-based programs combining education and personal discussion with practical strategies. For example, Yanos and colleagues (37) developed Narrative Enhancement and Cognitive Therapy (NECT) to harness the influence of self-narrative on sense of self and identity. In contrast, Corrigan and associates’ (38) brief intervention, Coming Out Proud, focuses on the pros and cons of self-disclosure of one’s mental illness. Fung and colleagues’ (39) Self-Stigma Reduction Program emphasizes interactions among insight, stigma, and treatment engagement.

The Ending Self-Stigma Program

We developed Ending Self-Stigma (ESS) by drawing from the social-cognitive model of internalized stigma (21,22,34), cognitive-behavioral therapy, stigma research, and psychoeducation and recovery principles (28–31,35–42). ESS consists of nine weekly 90-minute manualized group sessions. Each group consists of six to eight individuals and is led by one or two trained group facilitators. Each session reviews the previous week’s key points and then uses discussion and practice exercises to present a strategy for reducing self-stigma. Participants are encouraged to demonstrate mutual support, individualize the strategies, and carry out personalized practice between sessions.

In 2011, 34 adults with serious mental illnesses participated in a noncontrolled pilot study of ESS (43). The study conducted assessments of internalized stigma, mental health recovery, empowerment, and perceived social support before and after the intervention. Participants were mostly male (81%), middle aged (mean±SD=54±9), high school graduates (81%), and African American (59%); 35% were white, and 6% were of multiracial background; and all were U.S. military veterans. Results were promising: internalized stigma decreased significantly, whereas recovery orientation and perceived social support increased significantly. Recently, other teams have also published intervention evaluations (44–50), with positive to mixed results. However, like the results of our pilot, their results were often limited by small sample sizes (<50) and the absence of a control or comparison group.

This study, a randomized controlled trial (RCT), tested whether ESS reduced internalized stigma and improved recovery-oriented variables among adults with serious mental illnesses. We hypothesized that ESS would result in significantly reduced scores on the agreement, self-concurrence, and self-esteem decrement portions of the Self-Stigma of Mental Illness Scale (SSMIS) (34,51) and on the total score and subscales of the 29-item Internalized Stigma of Mental Illness (ISMI-29) scale (24,52). Secondarily, we hypothesized that ESS participants would show greater increases in the sense of belonging, self-efficacy, and recovery orientation compared with control participants and that the benefits of ESS participation would be sustained at six months postintervention.

Methods

We tested ESS against minimally enhanced treatment as usual, with assessments at baseline, after completion of the intervention (postintervention), and at six-month follow-up, with data collected between June 2011 and April 2014. The study took place at five Maryland psychosocial rehabilitation programs—two urban, two suburban, and one rural. Although diverse, they share the typical structure of such U.S. programs (53). All serve adults meeting Maryland’s “severely mentally ill priority population” definition (54), which requires diagnosis of schizophrenia, bipolar, recurrent major depressive, schizotypal personality, borderline personality, or other delusional or psychotic disorder and documented functional impairments (54). Procedures were preapproved by the University of Maryland Medical School Institutional Review Board (IRB), the Sheppard Pratt Health System (home of several sites) IRB, and management at the study sites.

Eligible individuals were 18 to 90 years old, receiving services at a recruitment site, able to give full informed consent, and willing to say they would attend the ESS course if they were assigned randomly to the intervention. Exclusion was permitted only for persons with severe or profound mental retardation.

From 12 to 16 participants were recruited via flyers, announcements, and clinician recommendation at each site so that each ESS course contained six to eight people after randomization. A trained research assistant screened interested individuals. Eligible participants provided written consent, preceded by a brief assessment to verify study comprehension.

Intervention Condition: ESS

ESS was delivered as described above. Participants were randomly assigned to condition at the conclusion of the baseline assessment, stratified to create ESS classes of eight or fewer people at their study site each round. Two master’s-level therapists, one of whom was an author (CW), completed a full day of ESS training with the principal investigator (PI) before the study. One conducted ESS groups at two sites and the other at three sites. They also tracked attendance, contacted and sent materials to absent participants, and attended bimonthly group supervision with the PI.

To check ESS fidelity, one session was randomly chosen from sessions 1 to 5 before each course. A trained team member observed that session, completing a fidelity scale of ten content items and ten process items. If more than four items were rated “unacceptable” (<80% fidelity), a second session, randomly chosen from sessions 6 to 9, was observed and rated. If it too was below 80% fidelity, data from that course were excluded from the study.

Control Condition: Minimally Enhanced Treatment as Usual

Participants randomly assigned to the control group continued their usual psychiatric rehabilitation services at their study site. This involved individualized combinations of skill and illness management groups, individual counseling, care coordination, medication management, and wellness activities (53,55,56). Treatment as usual was minimally enhanced by a brochure about internalized stigma, which was reviewed with control-condition study participants by a trained research assistant after the conclusion of the person’s baseline interview.

Interviews, Randomization, and Measures

Assessment interviews were conducted at baseline, after completion of treatment (three months after baseline [postintervention]), and at follow-up (six months postintervention) by a trained research assistant blind to participants’ study condition. After baseline assessment, participants were randomly assigned to ESS or to the control group by using site-stratified, permuted blocks of 12 to 16 participants.

Assessments used the following measures: the SSMIS (34,51), a measure with four ten-item subscales (stereotype awareness, stereotype agreement, self-concurrence, and self-esteem decrement) corresponding to the social cognition model and with acceptable internal consistency (α=.72–.91) and test-retest reliability (r=.68–.82); the ISMI-29 (24,52), a measure of internalized stigma with five subscales (alienation, stereotype endorsement, discrimination experience, social withdrawal, and stigma resistance); the Maryland Assessment of Recovery in People With Serious Mental Illness (MARS) (57), a 25-item self-report measure of recovery orientation among people with serious mental illnesses with excellent internal consistency (α=.95), test-retest reliability (r=.898), and good face and content validity; the General Self-Efficacy Scale (58,59), an eight-item measure with good psychometrics that assesses a person’s beliefs or expectations about his or her capabilities; the Sense of Belonging Instrument (SOBI) (60), a 32-item measure of the psychological experience of belonging that has strong validity and reliability (61,62); the Brief Symptom Inventory (BSI) (63), a multidimensional symptom inventory derived from the Symptom Checklist-90–Revised containing 53 items for evaluating nine symptom dimensions and calculating three global indices of distress; the Experiences of Stigma Survey (4), which includes nine items regarding stigma-related disrespect experiences and 11 items regarding discrimination experiences; and the Beck Cognitive Insight (BCI) Scale (64–66), a 15-item questionnaire on which the self-certainty subscale score is subtracted from the self-reflectiveness subscale score to yield a cognitive insight score. In addition, we administered the Repeatable Battery of Neuropsychological Status (R-BANS) (67,68), which assesses list learning and story recall, at baseline. Results for the R-BANS subtests are combined into the R-BANS Immediate Memory Index, which measures encoding and learning of verbal information. We collected information about age, race-ethnicity, gender, education, living situation, military service, and employment and relationship status at baseline. Diagnosis was assessed via review of the participant’s clinical record and confirmed by a treatment provider.

Data Analysis

Frequencies, histograms, and normal-quartile plots were used to examine variable distributions and check for errors. Descriptive statistics of participant characteristics at baseline were calculated by condition and overall. For the primary analyses of whether ESS reduced internalized stigma compared with the control condition, we fit a repeated-measures mixed model to each of the SSMIS subscales and the ISMI inventory subscales by using data from all three time points. All participants who were part of the randomization process were included, regardless of follow-up assessment attendance, in keeping with intent-to-treat analysis. Within-individual correlation was accounted for by specifying correlated errors among repeated measures. Random effects were specified for correlation between individuals at the same site.

Several outcome scales were log transformed to correct right skew. Independent fixed effects in the model included time point and condition (ESS or control) × time point interaction terms. The interaction terms allowed for testing of differential changes between mean scores for the groups from baseline to postintervention. Because participants were randomly assigned to condition, differences in participants’ baseline characteristics were assumed to be chance variation, statistical tests comparing characteristics between the two groups were not conducted, and no baseline covariates were entered into the primary model (69).

If the condition × time interaction term indicated a significant change in an outcome between the baseline and postintervention assessments, a test was conducted using the condition × time interaction term at six-month follow-up to determine whether an intervention effect persisted. The effect size (ES) was calculated by dividing the condition × time beta coefficient by baseline raw standard deviation. In secondary analyses, models were rerun after removing ESS condition participants who attended zero classes.

Additional secondary analyses added interaction terms for baseline covariates to assess whether treatment effect varied by age, race-ethnicity, gender, diagnosis (psychosis versus nonpsychosis), and results for the BSI, the BCI indices, and the R-BANS Immediate Memory Index. Due to multiple exploratory tests, we adjusted these p values via false discovery rate procedures (70). We explored the relationship between attendance and treatment effect by replacing the dichotomous ESS condition variable in the main model with a four-level ESS attendance variable indicating the number of classes attended (0, 1–3, 4–6, and 7–9). All analyses were performed by using SAS, version 9.3.

Results

Sample

Of 315 interested individuals, 16 were ineligible and 31 declined or were lost to contact before giving consent; 268 completed consent, baseline assessment, and randomization. Of these, 253 completed the postintervention assessment, and 230 completed the six-month follow-up (an 86% retention rate). Randomization assigned 131 participants to ESS and 137 to the control group; retention between the two groups at six-month follow-up did not differ (N=113, 86%, and N=117, 85%, respectively).

Demographic results, clinical characteristics, and stigma scores at baseline are shown in Table 1. Many participants were middle aged, African American (47%), and male (61%) with 12 or more years of education (69%). Primary psychiatric diagnosis was most often schizophrenia or schizoaffective disorder (53%) or bipolar disorder (27%). Most participants (N=209, 78%) reported having experienced disrespect or discrimination regarding their mental illness “sometimes” to “very often” at baseline. Baseline ISMI-29 total scores showed that 69 (26%) participants reported no to minimal self-stigma; 127 (48%) reported mild self-stigma; 62 (23%), moderate self-stigma; and 9 (3%), severe self-stigma (32). There are no published norms for SSMIS scores.

ESS Participation

Of 131 participants randomly assigned to ESS, 56 (43%) attended seven to nine of the nine sessions. Twenty-six (20%) attended four to six sessions; 28 (21%) attended one to three sessions; and 21 (16%) did not attend any sessions. Common reasons for absence were conflicting health appointments, transportation obstacles, not feeling well enough to attend, and forgetting.

Fidelity Results

All 13 ESS courses observed for fidelity were rated at or above 95%. No courses required the second fidelity check, and all data were retained for the study.

Primary Outcome: Internalized Stigma

ESS participants showed statistically larger reductions from baseline to postintervention compared with control-condition participants on the stereotype agreement (ES=–.31) and self-concurrence (ES=–.30) subscales of the SSMIS but not on the awareness or self-esteem decrement subscales (Table 2). On the ISMI-29, ESS participants showed significant but small improvements over the control group on alienation (ES=–.19) and stigma resistance (ES=–.27) but not on stereotype endorsement, discrimination, and social withdrawal nor on the total score. These findings reflect partial support of our hypotheses. No significant differences (ESS effects) persisted at six-month follow-up (Table 2), although gains made by ESS participants between baseline and postintervention on key subscales (SSMIS agreement, ISMI-29 alienation, and ISMI-29 stigma resistance) were largely maintained.

Secondary Outcomes: Recovery and Psychosocial Variables

There was a modest, statistically significant increase in MARS scores (ES=.18) from baseline to postintervention among ESS participants compared with control participants. There was no such effect on the SOBI or the General Self-Efficacy Scale (Table 2). MARS total scores continued to increase among ESS participants between postintervention and the six-month follow-up but not significantly more than among controls.

Exposure to ESS and Treatment Effect

Analyses comparing ESS participants who attended at least one session (N=110) with the control group yielded no meaningful changes from intent-to-treat results. Further, the treatment effects detected by the primary analysis were not modified by any baseline covariates. We also examined the associations between attendance and treatment effect by comparing mean changes in each outcome measure from baseline to postintervention among ESS participants grouped by number of sessions attended (0, 1–3, 4–6, and 7–9) and participants in the control group. Increased attendance was associated with the four subscales that showed significant reductions overall. Specifically, participants who attended seven to nine sessions showed statistically significant reductions on all four subscales compared with the control group: SSMIS stereotype agreement (t=−2.73, df=246, p=.050), SSMIS self-concurrence (t=−3.18, df=243, p=.002), ISMI-29 alienation (t=−2.11, df=252, p=.036), and ISMI-29 stigma resistance (t=−2.70, df=248, p=.007). However, there were no significant reductions in these subscales between the ESS participants who attended zero or one to three sessions and the control group. ESS participants who attended four to six sessions showed significant reductions compared with the control group on SSMIS stereotype agreement (t=−1.97, df=246, p=.050) and ISMI-29 stigma resistance (t=−3.06, df=251, p=.003).

Discussion and Conclusions

In this RCT, ESS helped participants improve key aspects of internalized stigma from baseline to postintervention compared with the control group. In addition, although baseline MARS recovery orientation scores were quite high, scores among ESS participants increased at postintervention, whereas scores among the participants in the control group decreased, and the difference was significant (Table 2). These findings reflect core elements of ESS and the dynamics of internalized stigma (2–12). ESS combines strategies, practice, and encouragement to question and resist stereotypes and draws on group camaraderie and personalized experiences to decrease alienation and demoralization. Furthermore, the stability of SSMIS awareness and the ISMI-29 discrimination scores shows that ESS did not iatrogenically increase participants’ sensitivity to societal stigma.

Most published self-stigma interventions have been tested only in limited studies, often showing pre-post improvements but lacking a control group, sufficient sample sizes, or follow-up time points (35,38,39,44–48,50,71). The exceptions are pre-post studies of NECT (49), an anti-stigma photovoice intervention (72), and an unusual HIV self-stigma–focused video intervention (73), each of which showed sustained improvement in self-stigma at follow-up for an intervention group versus a control group. To date, each is the sole rigorous study of the respective intervention to show positive results. This level of knowledge is insufficient for discerning what approaches might be used or combined to best effect. This study may begin to fill these gaps.

ESS attendance was variable, and attending more sessions predicted more benefit. Absences were for common reasons: conflicting appointments, transportation and logistical challenges, feeling ill, or forgetting. Going forward, interventions need to work around such obstacles to maximize effects. Diverse delivery formats and technologies, such as flexible session availability, one-on-one consultation, peer support, and Web-based or online video interaction, may be useful.

Notably, many of the 44 ESS participants who attended zero to two sessions said or implied that they never intended to attend because they were not interested in attending or because attendance was not associated with an honorarium, as were the research interviews. Our study also did not screen for a threshold of internalized stigma before enrollment. In “real-life” clinical use of ESS (or other interventions), people enroll out of concern about or interest in internalized stigma and its impacts. This gap between our research protocol and practical use of ESS was a study limitation.

It is notable that the pre-post benefits of ESS did not persist at six months in this RCT. Most relevant measures did not worsen among ESS participants over that time. Instead, for reasons unknown, the scores of participants in the control condition improved from postintervention to the six-month follow-up. However, ESS participants’ SSMIS self-concurrence and SSMIS self-esteem decrement scores did worsen slightly between postintervention and the six-month follow-up (Table 2). Self-stigma often fluctuates over time and circumstance, so fairly large ESs are necessary for an intervention to show effectiveness at follow-up. This suggests that ESS was insufficient to create lasting group effects in these core variables.

Unfortunately, that makes sense. Stigmatizing encounters are not uncommon (3,23), putting people at continuous longitudinal risk of internalized stigma and other stigma effects. Countering such ongoing exposure and risk likely requires longer-term resistance and resilience training and strategy reinforcement than would be possible in a nine-session course (74). In fact, numerous ESS study participants sought to take the class a second time after the study was over. Substantial, intermittent, and pervasive interventions may be needed to support protective strategies and counter-messaging in order to produce long-term effects. In addition, this study could not assess objective measures of recovery, such as employment, or other facets of stigma response, such as anticipated stigma, so it may have missed some of the effects of ESS.

Last, almost all study participants had contended with mental health problems for many years, and all were experiencing significant functional impairment, a requirement for eligibility for services at these settings. Some embraced the course avidly, commenting that it came at “the right time”—while they were already thinking about or struggling with internalized stigma. They described ESS as a source of validation, practical assistance, and inspiration. Another subset of participants described having had significant problems with internalized stigma when they were younger and wished they had had access to ESS then. We are investigating how to identify the “right time” to introduce self-stigma interventions and whether such interventions may have more impact earlier in a person’s illness trajectory.