Sustained Effect of a Brief Video in Reducing Public Stigma Toward Individuals With Psychosis: A Randomized Controlled Trial of Young Adults

More than two-thirds of people with psychosis worldwide receive no mental health care, at least in part because of stigma embedded in systemic or structural discrimination (1). Stigma stems from negative beliefs and attitudes that trigger people to avoid, fear, and reject individuals with mental illness (2). Young people with first-episode psychosis have identified stigma and how others perceive them as major barriers to care (3). Shorter duration of untreated psychosis among people with first-episode psychosis is associated with better short- and long-term prognosis (4). As public stigma toward individuals with first-episode psychosis has been identified as a barrier to help seeking, reducing stigma among young people may enhance their willingness to seek treatment, if needed (5–7). Moreover, interventions in the form of informational campaigns for the general public showed efficacy in increasing knowledge about mental illness and in reducing the duration of untreated psychosis of individuals with first-episode psychosis from 16 weeks to 5 weeks (8). To our knowledge, previous studies were aimed at reducing components of public stigma among the general population as opposed to young people.

Existing intervention studies on reducing stigma toward people with mental illness have shown promise. Social contact interventions, defined by interpersonal contact with a representative of the stigmatized group, are the most effective way to lower public stigma and prejudice (9–13). The mechanism is straightforward: an empowered protagonist with lived experience, who shares characteristics with the audience and is seen as a potential peer, shares his or her personal story, thus following the principle of moderately disconfirming stereotypes by balancing realities of the illness and messages of hope (14). Emphasizing recovery-oriented themes has lowered stigma more than symptom-based content (15). Social contact has shown similar efficacy with in-person and video interventions (16). Relative to in-person interventions, video-based interventions are less expensive, simpler to replicate, and easier to disseminate to large populations.

The current video-based literature has limitations, however. Interventions have focused on immediate change and have lacked long-term follow-up, and research has studied relatively small sample sizes among specific populations, limiting generalizability (17, 18). A recent review (19) found that contact-based interventions yielded small- to medium-sized stigma reductions immediately after the intervention. These effects did not vary by length of contact (10 to 65 minutes; average length, 24 minutes) or by type (in-person compared with video). Most (20 of 22; 91%) of these studies assessed college students, a nonrepresentative sample, as student research participation is often required for school credit. Sample sizes ranged from 20 to 782, with a median of 97 participants. Another meta-analysis found that despite the immediate benefits of antistigma interventions, 74% of the studies had no follow-up (20). Hence, studies utilizing rigorous longitudinal methodology are needed (21, 22).

We recently tested the efficacy of a 90-second social contact–based video intervention in reducing public stigma toward individuals with schizophrenia among 1,203 young adults (23). Our study focused on this age range for several reasons. First, youths are especially sensitive to stigma because of their stage of identity consolidation, characterized by a powerful need for a sense of competence, social acceptance, and autonomy (24). Second, young adults are very concerned with how their peers view them, and we wanted to intercede before attitudes become entrenched (25). Third, our brief online video intervention seemed a better fit for this age range, considering the high use of media among this population. Lastly, young adulthood overlaps with the onset of many mental disorders, including first-episode psychosis, and thus includes the potential peer group of individuals with first-episode psychosis who may hold stigmatizing attitudes toward individuals with the illness. The video presented schizophrenia with a human face rather than as a “brain disease.” As hypothesized, the video-based intervention group yielded lower rates of stigma than the written vignette and control groups, and the written vignette group yielded lower rates of stigma than the control group across all domains. Women showed lower rates of stigma than men solely in the video group, supporting our hypothesis that identification with the video protagonist reduces stigma. This finding coheres with previous studies showing that matching the gender of the respondent with the video protagonist significantly moderated changes from pre- to postintervention assessments across most of the stigma domains assessed (26). This proof-of-concept study was the first to demonstrate such an effect in a young adult sample and the first to employ so brief an antistigma intervention. After initial efforts to create the video, conducting such an intervention requires minimal resources, and it can be easily disseminated and reach a wide audience. However, the study lacked a baseline assessment, precluding direct measurement of actual change, and also lacked follow-up retesting to assess its durability.

To address these limitations, we designed a randomized controlled replication study with assessments at baseline, after the intervention, and at 30-day follow-up. Young participants were randomly assigned to a brief video-based intervention (“video”), to a written vignette intervention containing the same material (“vignette”), or to a nonintervention control condition (“control”). We hypothesized that we would replicate our previous findings by showing a greater reduction in stigma in the video group, with greater reduction among women, and that we would observe a sustained benefit in the video group relative to the other groups.

METHODS

Participants and Recruitment Procedure

We recruited participants using Amazon Mechanical Turk (MTurk) (27), a popular crowdsourcing platform widely used in medical and psychology research to assess treatment efficacy (28). We included only English-speaking U.S. residents ages 18–30 years. We used several methods to exclude invalid participants to ensure the validity of results. First, in designing the study, we added a timer to the “next” button to ensure that participants read instructions (5-second minimum), and we did the same for the written vignette (10 seconds) and for the watching of the video (90 seconds). This timer should have prevented most bots from continuing to the survey itself, as the HTML button did not appear on a page scan and the internal timer interval was too long for any timeout mechanism. Second, we used an open-ended question format requiring the participant’s age as a validity question and allowed only a two-digit number as a valid answer. We then compared that answer with the age data MTurk provided to disqualify fake participants. Third, we scanned the participants to exclude those who answered the survey more than once, who completed the assessment in less than the minimum expected time, who were from locations (coordinates) outside the United States, or who had suspicious IP addresses. Participants were compensated $1.10 for each step of the study, for an overall compensation of $2.20. The New York State Psychiatric Institute institutional review board approved the project. Before initiating the study, participants reviewed an online informed consent form. Those who agreed to participate were directed to complete the study procedures via Qualtrics.com, a secure online data collection platform.

Intervention

The goal of the study was to evaluate the acute change and longer-term efficacy of antistigma interventions using the same brief video as in our first study. A link to the Center for Practice Innovations’ resources is included in the online supplement. These resources include a portal for patients and their relatives that presents video clips of patients and families describing different aspects of coping with their everyday life difficulties. From those materials, we reduced to 90 seconds an approximately 11-minute video of an empowered young woman with schizophrenia who described her first psychotic episode and her struggles with medications and side effects. In the shortened video, she describes her symptoms during illness onset (e.g., “I went four days without sleeping.… I have seen things that weren’t there.”) and her current daily difficulties (e.g., “Every day is a battle.… Every day is hard.”), in which she continues to experience psychotic symptoms in attenuated form, thus still meeting some common psychosis stereotypes. Notably, even while experiencing these symptoms, she remained capable of working and having meaningful relationships. The video concluded with themes of hope and recovery (e.g., “It’s not the end of the world. It is not something that is impossible to live with.”). This highlighted her as a human being, presenting her continuing symptoms in the context of her personal human narrative. The video thus humanized the illness via social contact with an individual perceived at an equal power status (i.e., a peer) who still met some stereotypes of psychosis but who had meaningful relationships and functioned reasonably well in the roles in her life. The vignette segment of our study was a written description of the video content (both symptoms and recovery themes) but lacked the direct social contact with an individual with lived experience.

Instruments

The assessment measured public stigma across five domains, with 19 items derived from five scales that showed good initial reliability and validity in our first study. We analyzed each of the five subscales separately. The social distance domain had six items divided into three casual (e.g., “Would you be willing to have someone with schizophrenia as a neighbor?”) and three intimate types of social distance (e.g., “Would you be willing to allow a child of yours to date a person with schizophrenia?”), derived from Boyd et al. (29). The overall scale has good internal consistency (Cronbach’s alpha=0.92). Four items of the separateness (or “differentness”) domain came from Phelan (30), for example, “When you think of a person with schizophrenia, how different do you think they are from other people?” (alpha=0.69). For the stereotyping domain, we drew four items from the General Social Survey (31) to assess the participant’s perception of the ability of a person with schizophrenia to make decisions about treatment and money, as well as the likelihood that the person would be violent (alpha=0.80). Three social restriction items assessed the participant’s perception of whether a person with schizophrenia should babysit small children, marry, or have children (alpha=0.68). To measure public perception on recovery, we adapted two items from the 41-item Recovery Assessment Scale (32). This newly developed measure showed good reliability and construct validity in our previous study (i.e., the public perceived greater possibility of recovery for the video character after the intervention). One item assessed perceived recovery from a community member’s perspective, and another addressed the perception of the ability of a person with schizophrenia to meet current personal goals and have a plan for staying well. Responses range from 1 (“strongly agree”) to 4 (“strongly disagree”), and the alpha for this study was 0.77. Two items assessed whether participants had a friend or family member with serious mental illness (yes, no, and prefer not to answer), and if so, their level of intimacy with that individual (not intimate/close, somewhat intimate/close, very intimate/close, and prefer not to answer) (33).

Statistical Analysis

Data were analyzed using SPSS, version 26.0 (IBM, Armonk, N.Y.). We used Pearson’s chi-square and one-way analysis of variance (ANOVA) to compare demographic variables among groups. Repeated-measures ANOVAs and multivariate analyses of variance (MANOVA) compared mean stigma scores that were obtained as the sum of each stigma domain across the three groups and three time points; when between-group differences were found, post hoc tests were used to compare each group pair. We then used independent sample t tests to compare changes between time points across groups and gender differences within groups. A Bonferroni correction for multiple comparisons yielded a corrected p value significance threshold of 0.01.

RESULTS

Sample Characteristics

The sample of 1,055 participants was randomly and proportionally assigned to three study groups, after excluding 162 (13%) individuals who failed validity tests. Ninety-eight percent (N=1,029) of participants completed the postintervention assessment, and 78% (N=803) completed the 30-day follow-up (Figure 1). Demographic characteristics did not differ between those who did and did not complete the study. The study sample was evenly divided across gender (women, N=531, 50.3%), and the participants’ mean age was 26.7 years (SD=3.1). The majority (N=769, 73%) identified as White, while 123 (12%) identified as Asian, 107 (10%) as African American, 13 (1%) as Native American, and 43 (4%) as other. One hundred eleven (11%) reported Hispanic ethnicity. Study arms did not significantly differ by gender, age, race, or ethnicity (Table 1).

As hypothesized, study groups significantly differed in outcome. A three-by-three group-by-time MANOVA showed a significant group-by-time interaction for the total scores of all five stigma-related subscales (Wilks’s lambda F=12.6, p<0.001). A three-by-three group-by-time ANOVA also showed that mean stigma scores significantly decreased over time, predominantly in the video group, across all five stigma-related domains: social distance (F=38, p<0.001), stereotyping (F=31, p<0.001), separateness (F=6.9, p=0.001), social restriction (F=10, p<0.001), and perceived recovery (F=17, p<0.001). Figure 2 presents mean score curves of the study groups over time, across all stigma domains. No gender differences in changes between time points were found between study groups.

^a Higher scores indicate higher stigma.

To better understand the outcome differences across intervention arms, we compared changes from baseline to postintervention assessment and from baseline to 30-day follow-up between study groups (Table 2). Independent t tests showed a significant difference between the video and vignette groups in changes from baseline to postintervention assessment and from baseline to follow-up across all five stigma domains (social distance, stereotyping, separateness, social restriction, and perceived recovery). Similarly, the video and control groups significantly differed in changes from baseline to postintervention assessment and from baseline to follow-up across all stigma domains. However, while the vignette group differed from the control group in changes between baseline and postintervention assessment across all stigma domains, changes between baseline and 30-day follow-up were observed only in social distance (t=2.0, p=0.042) and stereotyping (t=2.2, p=0.030), whereas there were no differences in the domains of separateness, social restriction, and perceived recovery.

Three hundred seventy six (37%) respondents, evenly distributed across study arms, replied affirmatively to the question, “Do you have a friend or a family member diagnosed with serious mental illness?” (Table 1). Overall, participants of all study groups who knew people with serious mental illness had significantly lower baseline scores on the scales of social distance, stereotyping, social restriction, and perceived recovery. However, this difference in participants familiar with people with serious mental illness having lower stigma disappeared at the postintervention and 30-day follow-up assessments in the video group, whereas in the vignette and control groups, it persisted across all time points. Figure 3 presents the mean scores of the social distance domain, divided by close familiarity to a person diagnosed with serious mental illness, across all time points. We did not find gender differences within groups.

^a Scores ranged from 6 to 24, with higher scores indicating higher stigma. *p≤0.05. **p<0.01. ***p<0.001.

DISCUSSION

Our randomized controlled replication study tested the efficacy of a brief contact-based video intervention in reducing public stigma toward first-episode psychosis and schizophrenia among 1,055 young adults. In the 90-second video, a 22-year-old woman suffering from schizophrenia directly and candidly portrayed her personal story, describing how she is fulfilling her goal to live a meaningful and productive life. She humanized the illness. As hypothesized, this emotional video-based intervention had significantly greater potency than the vignette and control conditions in reducing stigma across all domains at the postintervention assessment. These findings replicate and extend our previous findings by demonstrating a pre- to postintervention reduction in stigma and not merely differences among study groups. These findings corroborate the literature emphasizing the importance of contact-based interventions (23, 34). Reducing public stigma toward psychosis has the potential to help young people with psychosis overcome barriers to care, enhance detection of psychosis, and shorten duration of untreated psychosis.

Our follow-up findings showed a continuing significant difference between study groups. Despite a slight rebound in stigma reduction at 30-day follow-up compared with postintervention assessment, the video-based intervention group remained significantly different from the vignette and control groups at 30-day follow-up. Differences between the vignette and control groups lost statistical significance by day 30. This finding emphasizes the core difference between social contact video-based interventions and written vignettes and perhaps other interventions—identification with the protagonist is viewed as creating emotional engagement, which in turn generates sustainable effect. Previous studies that tested longer-term effects of video interventions used videos far longer than ours (range=20–85 minutes) and/or included smaller samples of college students (range=43–243 participants), thus limiting generalizability (17–19, 35, 36). This is the first study to demonstrate a persisting effect in a relatively large sample of the young general population from such a brief video stimulus. Because of the slight rebound, it is reasonable to assume diluting of the intervention effect. Further studies should examine whether booster videos would enhance the durability of the antistigma effect.

Our finding that having a friend or family member with serious mental illness is associated with lower stigma is consistent with previous studies (33, 37). Although this difference was consistent across groups at baseline, it disappeared only in the video group after the intervention, across all stigma domains. This finding persisted at 30-day follow-up. The absence of such a difference in the video group suggests that identifying with the protagonist, even briefly, imitates close contact with a person with serious mental illness. Further studies should examine whether the video content should be adjusted to account for familiarity with a person with serious mental illness.

Is 90 seconds the ideal length for social contact–based video intervention? Longer videos have not yielded a greater beneficial effect than shorter videos (19), and brevity has advantages: lower cost, use of fewer resources, and greater ease of dissemination to large audiences. Traditionally, social contact interventions for young adults have occurred in educational environments like colleges and schools, and they involve in-person meetings with people with serious mental illness who share their recovery story or a longer workshop that includes formal educational formats like lectures or group discussions (19, 38). An emerging secular trend toward online platforms has been accelerated dramatically by the COVID-19 pandemic, which is changing dissemination of knowledge and increasing the use of remote learning, social media, and the Internet (39, 40). Finally, shorter videos better suit younger audiences who may have limited frustration tolerance or attention span. In an era of very brief knowledge bites, mostly consumed via social media (e.g., Instagram, Twitter, and TikTok), shorter interventions may be better suited to the attention span of the average young viewer. Although the 90-second video, the shortest yet tested, seemed to produce at least the benefit of longer videos, still shorter videos might prove even more suitable to young audiences. Additional research should attempt to define the optimal length for sustainable interventions for stigma reduction.

In our previous study, a secondary analysis found that the video had greater influence on women than did comparison conditions across all scales (23). We assumed that the gender effect reflected identification with the presenter: female viewers identifying with a female presenter. It is unclear why the present study failed to replicate this result, but it may be in part due to a smaller study sample of the video group (367 compared with 596). Previous research has shown that stigma levels differ across sociodemographic groups, but few studies have examined the influence of concordance between video presenter and viewer gender on stigma reduction (26). Other studies should examine whether matching sociodemographic characteristics enhances stigma reduction.

Limitations

Our study has several limitations. First, our findings may be limited to MTurk participants, who may not be fully representative of the general population (27). Seventy-three percent of participants described themselves as White, 10% as African American, and 12% as Asian, and 11% reported Hispanic ethnicity, percentages slightly divergent from the overall U.S. population reporting as 76% White, 13% African American, 6% Asian, and 18% Hispanic, based on the last U.S. census. In addition, MTurk participants may differ in education level. Second, the low number of African American participants gave us insufficient power to evaluate racial differences. Third, our study included only a single video, limiting the ability to test the influence of the gender, race, and ethnicity of multiple video intervention protagonists on stigma reduction. Fourth, we assessed attitudes only, the reporting of which is subject to social desirability bias. Future studies should measure behaviors and/or implicit attitudes. Lastly, as our study evaluated only a 30-day effect, future research should examine longer-term sustainability.

Conclusions

This randomized controlled study replicated and enhanced the findings of our first study, showing a longer-term effect of stigma reduction in the social contact–based video intervention arm. A 90-second video humanized schizophrenia and reduced stigma, suggesting its potential to increase the likelihood of seeking care and, ultimately, to improve access to treatment among young people with first-episode psychosis and schizophrenia. Future research should address gender, race, and ethnic differences and determine whether it is possible to further reduce video length without reducing efficacy.