Outcomes of Ending Self-Stigma, a Group Intervention to Reduce Internalized Stigma, Among Individuals With Serious Mental Illness

Participants were recruited from outpatient mental health programs and clinics at three U.S. Department of Veterans Affairs (VA) medical centers in the VA Capitol Health Care Integrated Service Network between October 2011 and May 2014. They were identified through clinician referrals, review of clinic rosters (via an approved partial HIPAA waiver), and recruitment flyers. Eligible veterans were ages 18–80 years; had a chart diagnosis of schizophrenia disorder, schizoaffective disorder, bipolar disorder, or major depression with psychotic features; and were receiving VA mental health services. Exclusion criteria included a documented history of severe or profound intellectual disability or previous receipt of the Ending Self-Stigma intervention. After being approached in person at a clinic appointment or via postal letter, interested and eligible individuals met with study staff, who explained all study details and obtained written informed consent.

Participants completed a 90-minute baseline assessment and comparable assessments at posttreatment and 6-month follow-up with a trained research assistant blinded to treatment assignment. Recruitment occurred in rounds of eight to 16, so each group contained four to eight participants after randomization immediately after the baseline assessment. Randomization occurred in blocks with random block sizes (1:1 allocation). The study statistician generated the assignment order within each block by using a random number generator. An a priori sample size calculation determined that the study needed to randomly assign 276 participants to detect a Cohen’s d effect size (treatment condition difference) of 0.43 (α=0.02; power=0.80), accounting for 25% attrition, group effect (intraclass correlation coefficient=0.05), and multiple testing. Toward the end of the study period, posttreatment assessment attrition was only 15%; thus, the necessary sample size was revised to N=235. The actual number of participants who were randomly assigned was N=249 (125 to the Ending Self-Stigma group and 124 to the Health and Wellness group). One participant assigned to Ending Self-Stigma was found to be ineligible after randomization and was withdrawn, leaving 124 participants in each group (total N=248) (see online online supplement). The study was approved by the institutional review boards of the University of Maryland School of Medicine and Washington D.C. VA Medical Center.

Ending Self-Stigma is a nine-session, manualized group intervention that teaches individuals tools and strategies they can use to address stigma and self-stigma in their everyday lives. Sessions were held weekly for 75–90 minutes and conducted as a closed group led by one to two trained group leaders. Each session teaches and practices a different strategy for addressing self-stigma and its effects (cognitive-behavioral strategies, strengthening positive aspects of the self, increasing belonging in community and with family and friends, and responding to stigma) and consists of a lecture, sharing of personal experiences, exercises, group support, and problem solving. Participants identify ways to practice skills between sessions to promote real-life, personalized experience with strategies.

The control condition, the Health and Wellness intervention, consisted of nine sessions focused on providing education and support regarding management of general physical well-being. Developed locally, this intervention was designed to be psychoeducational and focused on common general health concerns, with no content related to stigma or self-stigma. Sessions were also held once a week for 75–90 minutes, with each session reviewing previous session content, presenting new educational content, and allowing for discussion. Each session focused on one health-and-wellness issue (e.g., physical activity, healthy eating, sleep, relaxation skills, substance use, and medication management).

Facilitators for both groups were trained in the interventions and supported via biweekly supervision sessions by the principal investigator. Facilitators included a master’s-level clinician in psychology, two postdoctoral-level psychologists, one social worker, and a registered mental health nurse with previous experience conducting group therapy with individuals with serious mental illness.

Demographic information was obtained from participants at baseline. Psychiatric diagnosis was obtained from clinical charts. The total score of the Internalized Stigma of Mental Illness (ISMI) scale (15) and the four subscales (stereotype awareness, agreement, application to self, and harm to self-esteem) of the Self-Stigma of Mental Illness Scale (SSMIS) (24, 26) were used to measure internalized stigma. To assess other recovery-related outcomes, we used the Maryland Assessment of Recovery in Serious Mental Illness scale (27, 28), a 25-item self-report measure of recovery among people with serious mental illness; the two subscales of the Sense of Belonging Instrument (29, 30), psychological experiences of belonging and the antecedents to belonging, to measure aspects of perceived belongingness; and the General Self-Efficacy Scale (31), an eight-item measure of a person’s beliefs or expectations about his or her capabilities. Self-reported psychiatric symptoms were measured with the Brief Symptom Inventory (BSI) (32, 33), a 53-item multidimensional self-report inventory designed to assess nine symptom dimensions and three global indices of psychiatric distress. Discussions about stigma and its effects during clinical services were measured with a questionnaire developed for this study in which participants were asked to indicate how often (never, less than once a month, once or twice a month, three or four times a month, or more than four times a month) mental health stigma or its effects were discussed during any clinical services received (individual, group, or family services) in the previous 3 months at each time point.

Variable distributions were checked for outliers and potential errors and whether transformations were needed to meet model assumptions. To test whether Ending Self-Stigma reduced levels of internalized stigma from baseline to posttreatment and from baseline to the 6-month follow-up compared with Health and Wellness, we fit a repeated-measures, mixed-effects model to the total ISMI score (primary outcome). Intrasite correlation was modeled with a random site effect, and intraindividual correlations over repeated measures were estimated by specifying correlated error terms. Because of randomization, group means were specified to be equal at baseline (i.e., observed differences were assumed to be due to sampling error) (34). Analyses were performed with the intent-to-treat sample (i.e., all randomly assigned participants).

Next, changes in secondary outcomes (self-efficacy, recovery, and sense of belonging) from baseline to posttreatment, longer-term effects (6-month follow-up), changes in exploratory outcomes (SSMIS, depression, and psychotic symptoms), and the effects among participants who attended at least one class were examined with the same model as used for assessing the primary outcome. In exploratory analyses, we examined potential modifiers or subgroups that resulted in a differential intervention effect at posttreatment, including psychosis diagnosis (schizophrenia disorder, schizoaffective disorder, or major depression with psychotic features) vs. those without psychosis diagnosis (bipolar disorder), baseline internalized stigma (assessed with the ISMI scale) total score (≥2.0 vs. <2.0), and psychotic symptoms (BSI psychoticism subscale score ≥60 [“high”] vs. <60 [“not high”]) and whether attendance and average class size had effects on outcomes. To examine the effect or association of the above covariates with outcomes, we added to the main outcome analysis model the three-way interaction term (condition × time × modifier) and all two-way interaction terms with the modifier.

Finally, we analyzed variables regarding how much participants discussed stigma and internalized stigma in other therapies. First, we compared the Ending Self-Stigma group with the subgroup of Health and Wellness intervention participants who reported discussing internalized stigma less than once per month in individual, group, or family therapy sessions. Participants who did not receive one of these treatments were counted as not discussing internalized stigma for that treatment. Second, we used logistic mixed-effect models parallel to the main analysis model to examine whether treatment conditions had an effect on whether participants discussed internalized stigma once or more per month in individual and group therapy (as dichotomous outcomes) in the previous 3 months.

Most participants were men (87%); had a diagnosis of schizophrenia disorder (25%), schizoaffective disorder (26%), or bipolar disorder (42%); and had a mean±SD age of 53.4±9.2 years (Table 1). The participants’ characteristics were similar for both interventions.

Participants randomly assigned to Ending Self-Stigma attended 4.0±3.2 group sessions out of nine possible; 79% (N=98) attended at least one session, and 48% (N=59) attended five or more sessions. Those assigned to the Health and Wellness group attended 3.8±3.2 sessions each, and 43% (N=53) attended five or more sessions.

Twenty randomly sampled Ending Self-Stigma and 20 Health and Wellness sessions reflecting a variety of sessions during this study were rated by independent reviewers for facilitator adherence and competence. Each item was scored on a 3-point scale (0=unacceptable, 1=acceptable, 2=excellent). Scores were high for both the Ending Self-Stigma group, with a mean of 1.91±0.11 for adherence and 1.89±0.12 for competence, and the Health and Wellness group, with a mean of 1.92±0.11 for adherence and 1.94±0.11 for competence.

The Ending Self-Stigma and the Health and Wellness groups did not significantly differ in the mean change from baseline to posttreatment in the primary or in any of the secondary outcome variables (Table 2). These findings held even when comparing those who attended at least one class (N=98 [Ending Self-Stigma] vs. N=108 [Health and Wellness]). The two groups significantly differed in the SSMIS awareness subscale score, with those in the Health and Wellness group exhibiting significantly greater reductions in awareness of self-stigma from baseline to 6-month follow-up than those in the Ending Self-Stigma group. The two groups did not show any differences in the other exploratory outcome variables at posttreatment or 6-month follow-up.

An exploratory subgroup analyses revealed no significant main treatment effects among those with (or without) a psychotic diagnosis. However, high psychosis symptoms did significantly modify the treatment effect on the ISMI scale total score (t=−2.10, df=458, p=0.036). Among those with high baseline psychosis symptoms, the participants in the Ending Self-Stigma group experienced a greater reduction from baseline to posttreatment in the ISMI score than the participants in the Health and Wellness group. Average Ending Self-Stigma and Health and Wellness class size and changes in discussions of stigma were not associated with any treatment effects.

We found no significant differences between the two groups, but the combined sample exhibited reductions from baseline to posttreatment on the ISMI scale total score (t=−2.42, df=658, p=0.016) and the SSMIS agreement (t=−2.95, df=653, p=0.003) and harm–to–self-esteem (t=−2.39, df=655, p=0.017) subscales. The scores on the subscale psychological experience of belonging (t=2.00, df=655, p=0.045) of the Sense of Belonging Instrument increased overall, but no significant changes were detected on the Maryland Assessment of Recovery in Serious Mental Illness scale and the General Self-Efficacy Scale. Results were similar when data from those who did not attend any sessions were removed from the analyses. Only the Health and Wellness group showed a significant decline in depressive (t=−2.31, df=408, p=0.02) and psychotic (t=−2.24, df=408, p=0.03) symptoms at posttreatment.

In general, these improvements in internalized stigma were maintained over time. Significant reductions in the combined sample were found from baseline to 6-month follow-up on the ISMI scale total score (t=−2.75, df=658, p=0.006) as well as the agreement (t=−2.48, df=653, p=0.013), application-to-self (t=−2.24, df=653, p=0.25), and harm–to–self-esteem (t=−0.26, df=655, p=0.011) SSMIS subscales. We observed significant reductions depression among participants in the Health and Wellness group (t=−2.96, df=408, p<0.01); in addition, we noted declines in psychotic symptoms in both the Health and Wellness (t=−3.75, df=408, p<0.001) and Ending Self-Stigma (t=−2.55, df=408, p=0.01) groups.