Self-stigma in schizophrenia: a systematic review and meta-analysis of 37 studies from 25 high- and low-to-middle income countries

In schizophrenia, it is currently thought that stigma experience is increased by psychotic and depressive symptomatology, exposure to stigma at the workplace, and that self-stigma levels vary across countries without knowing the factors explaining these variations. The aim of the present meta-analysis was to synthetize the data of observational studies comprehensively exploring multiple self-stigma dimensions and associated factors. A systematic literature search without language or time restrictions was conducted in Medline, Google Scholar, and Web of Science for studies, last 09/2021. Eligible studies that included ≥80% of patients diagnosed with schizophrenia-spectrum disorders and used a validated scale measuring self-stigma dimensions were meta-analysed using random-effects models, followed by subgroup and meta-regression analyses. Study registration: PROSPERO CRD42020185030. Overall, 37 studies (n = 7717) from 25 countries (5 continents) published between 2007 and 2020 were included, with 20 studies conducted in high-income countries. These studies used two scales with total scores ranging 1–4. The mean estimate of perceived stigma was 2.76 [95% confidence interval (CI) = 2.60–2.94], experienced stigma 2.29 [95% CI = 2.18, 2.41], alienation 2.40 [95% CI = 2.29, 2.52], stereotype endorsement 2.14 [95% CI = 2.03, 2.27], social withdrawal 2.28 [95% CI = 2.17, 2.39] and stigma resistance 2.53 [95% CI = 2.43, 2.63]). Self-stigma levels did not reduce over time. Living outside urban areas, low-income, singleness, unemployment, high antipsychotic dose and low functioning were associated with different stigma dimensions. Some stigma dimensions were lower in studies carried out in Europe compared to other regions. Most studies published since 2007 report that self-stigma is a particular concern for a specific subgroup of patients. This subgroup is characterized by unemployment, high antipsychotic dose and low functioning. We identified important other missing factors that should be explored to improve the effectiveness of public policies and personalized interventions to reduce self-stigma. Importantly, classical illness severity indices (psychotic severity, age at illness onset, illness duration) and sociodemographic variables (age, sex and education) were not associated with self-stigma, moderating previous findings.


INTRODUCTION
Self-stigma is the status loss and discrimination triggered by negative stereotypes [1].The self-stigma associated with mental illness adds to the public health burden of mental illness itself by eroding individuals' social networks, self-esteem, and social status, all of which contribute to unemployment, isolation, delayed treatment-seeking, treatment-refractory symptoms, prolonged illness course, and avoidable hospitalizations [2].Anti-stigma initiatives have been established to decrease stigmatizing attitudes in the society as a whole [3].During the 2000s, a better understanding of the self-stigma or "internalized stigma" process -i.e., the internalization of public stigma [4]-has shifted the attention from public stigma to the subjective experience of stigmatized people.People with schizophrenia may report perceived stigma (thinking or being afraid of being the target of public stigma) or "experienced/objective" self-stigma.Experienced stigma has three dimensions (explored in three subscores): emotional (alienation /shame), cognitive (stereotype endorsement) and behavioral (social withdrawal) [2].In reaction to selfstigma, "stigma resistance", is defined as the ability to challenge or deflect self-stigmatizing beliefs [5].
Several individual and illness characteristics have been identified as factors influencing self-stigmatization [6].A systematic review carried out almost 10 years ago reported weighted percentage of approximately 49.2% of people with schizophrenia experiencing alienation and 26.8% stigma experience [7].This systematic review also reported that positive symptoms, depression, and general psychopathology were "associated with stigma in the majority of cases", however, we lack quantitative analyses to definitely conclude this [7].The authors also reported "ambiguous associations" with illness duration, negative symptoms, treatment compliance, and social functioning [7].Labor force has been associated with lower self-esteem in a recent large-scale cohort of outpatients with schizophrenia [8], which may play a role in stigma.We need to know if other factors, like the age at illness onset and antipsychotic treatments, may play a role in stigma experience.Stigma resistance has been shown to be associated with advanced stages of recovery in schizophrenia [5], but has been inconsistently associated with sociodemographic variables (age, sex, educational level, employment, marital status, income, and source of income) [5].
In summary, while self-stigma dimensions have been conceptualized fairly well in schizophrenia, to our knowledge, a quantitative synthesis of the data has not been conducted for a decade.Our objective was to synthetize in a quantitative metaanalysis the mean estimates of self-stigma dimensions and to analyse sociodemographic and illness-related associated factors.

Literature search strategy
This meta-analysis was conducted in line with the Preferred Reporting Items for Systematic reviews and Meta-Analysis guidelines [9].Systematic bibliographic searches were carried out according to the Cochrane methodology and in line with the PRESS 2015 Guideline treatment (presented in Supplementary Annex 1) [10].This project was registered in PROSPERO (reference number CRD42020185030).
The search paradigm was based on the Mesh terms of the Medline® database (using Pubmed®) and adapted for Google Scholar® and Web of Science®: ("schizophrenia" OR "schizoaffective disorder" OR "psychotic disorders") AND ("stigma" OR "stigmatization").The last search was carried out on September 30th, 2021.The reference lists of relevant reviews and articles were manually searched for additional eligible articles.If needed, the corresponding authors were asked to provide additional data not included in the original publications and any unpublished results.

Eligibility
The inclusion criteria were as follows: (1) Any language and date of publication; (2) Original research papers; (3) ≥80% of patients diagnosed with schizophrenia-spectrum disorders (schizophrenia, schizoaffective disorder, schizophreniform disorder); (4) Use of a validated scale that measures self-stigma and/or its domains, with or without cross-cultural validation.
The exclusion criteria were: use of non-validated scales (e.g., questionnaire created for the purpose of the study) or validated scales used in three studies or less, absence of results separately for participants with schizophrenia and other mental disorders, or crude scores reported without dividing these scores by the number of items (= absence of standardized score).
MV and MJ carried out the selection of included studies.In case of a non-consensus for the inclusion of a study, a third author (GF) made the final decision.

Self-stigma dimensions and associated factors
The following variables were extracted into a pre-defined form:  [11] positive factor), Negative symptoms (PANSS negative factor), Total psychotic symptoms (PANSS total score), depression (Calgary score) [12], insight into illness (Birchwood Insight Scale score) [13], Lifetime number of psychiatric hospitalizations, percentage of patients with tobacco/alcohol/cannabis/substance use disorder, mean antipsychotic daily dose (calculated in chlorpromazine equivalents [14], continuous), percentage of patients treated with second-generation or firstgeneration antipsychotics, Global Assessment of Functioning (GAF) [15]  Two researchers (MV and MJ) extracted data from the included studies in a systematic manner using a predesigned extraction form, which was based on the Joanna Briggs Institute Data Extraction Form for Prevalence and Incidence Studies [17].Additional items relevant to the current study were also added.GF examined each discrepancy in data extraction to reach consensus.

Quality assessment
Study quality analysis was performed using the NIH Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies [18].This tool was chosen to address the limits of the Newcastle-Ottawa scale [19].

Statistical analyses
Using the inverse-variance weighting method [20], a randomeffects model was used to calculate the pooled estimate of each self-stigma dimension and its 95% confidence interval (95% CI).Q and I 2 were calculated to assess heterogeneity across all studies and within subgroups [21], with I 2 ≥ 50% indicating significant heterogeneity [22].Sensitivity analyses were conducted using the leave-one-out method carried out on the total stigma score [23].Subgroup analyses for binary variables (Europa vs. other countries, Asia vs. other countries, High-income country vs Lower to middleincome countries, Small sample size (<100) vs. large sample size (>/=100), Multicentric vs. monocentric recruitment, Outpatient vs. mixed out-and/or inpatient recruitment), study published before 2016 (median year) vs. ≥2016) and univariable meta-regression analyses for continuous variables were carried out when they were reported in ≥4 independent samples (some studies reporting results for more than one sample).All binary variables were explored, as well as the following continuous variables: Mean sample age, percentage of males, percentage of highly educated participants, percentage of unemployed participants, percentage of participants living in an urban area, percentage of low-income participants, percentage of single participants, lifetime number of psychiatric hospitalizations, age at illness onset, illness duration, total psychotic symptoms (PANSS total score), positive symptoms (measured with PANSS [11] positive factor), negative symptoms (PANSS negative factor), daily antipsychotic dose (calculated by the chlorpromazine equivalent method), functioning measured by the Global Assessment of Functioning score (GAF) [15]).The other factors were reported in <4 independent samples and were therefore not analysed.Publication bias was assessed graphically with a funnel plot and statistically with Egger's test when ≥10 studies were included in the meta-analysis [24].All analyses and graphs were carried out using R software (version 3.1.1;R Foundation, Vienna, Austria) with the meta package.

Self-stigma dimensions' score estimates
The forest plots of the mean scores for the different self-stigma dimensions are presented in Fig. 2A-F

Source of heterogeneity
Meta regressions.All estimates were associated with high heterogeneity (I 2 > 96% for all self-stigma dimensions, Supplementary Fig. 1).The leave-one-out method did not identify a specific study that was responsible for heterogeneity (Supplementary Table 4).

DISCUSSION
This work adds important insight to a previous comprehensive systematic review published on self-stigma in schizophrenia in 2013 [7].First, we have reported for the first time quantitatively the mean self-stigma scores and its different dimensions that can be used to inform future evaluations of self-stigma in schizophrenia.Exploring the role of depression seems therefore a priority, as depression may impact both emotional and cognitive processes of self-stigma or be influenced negatively by selfstigma, but we lacked sufficient data to confirm that depressed patients reported higher self-stigma levels.The high heterogeneity (≥95% for all dimensions) also suggest that stigma is not a uniform phenomenon concerning all patients with schizophrenia to a similar degree.We found that the origin of study participants in urban areas and low-income countries was associated with the emotional dimension of self-stigma (alienation) and the cognitive dimension  of self-stigma (stereotype endorsement).This result suggests that these parameters may influence the emotional and cognitive dimensions of stigma more than behavioral dimension of stigma.
In HIV studies, stigma has been repeatedly shown to be lower in urban areas, due to better knowledge about the illness [63][64][65], and the same phenomenon has been shown in schizophrenia, for example in India [66] China [67], and Ethiopia [68,69].The percentage of participants living in urban vs. rural areas was reported in 12 studies for this analysis, which makes our result robust, while the percentage of participants living in low-income countries was only reported in four studies.However, only six studies included a mix of participants living in urban and rural areas.These studies were carried out only in Turkey (one study [53]) and India (five studies [28,29,38,55]), which limits the generalizability of this result.The association between lower selfstigma and lower income may also in part be due to the fact that the expectations of performance are lower in low-income areas and strata.The delta between the functioning of people with schizophrenia and other people would therefore be smaller, inducing lower increase of self-stigma.
Our results indicate that the participants of European studies reported lower rates of stigma compared to other regions, suggesting that the region of the study may also play a role.Most of these studies were carried out in Western Europe.Contrary to what could have been expected, our results suggest that the lower self-stigma level is not explained by the high-income country status.It may be hypothesized that high-income countries may have disseminated public health policies to lower the public stigma of mental illness, or that social interventions to protect people with mental disabilities may reduce self-stigma experience in those countries [8].It should be noted that we did not find a decrease in the level of stigma over time despite the implementation of public and individual interventions to reduce stigma over 20 years of highprofile national campaigns (e.g., Time to Change in England; Beyond Blue in Australia, Schizophrénie, bonheur et Cie in France, Stigma-Free in the US) [70].However, the oldest eligible study was published in 2007, and public online interventions are relatively recent.A recent systematic review of interventions carried out in low-to-middle-income countries has highlighted that most of interventions aimed at decreasing stigma were disseminated during the last 5 years [71].As two recent meta-analyses have confirmed their effectiveness, web-based interventions should be promoted in countries with high internet access [72,73].The evidence for the effectiveness of interventions aiming to reduce personal stigma was recently synthetized in a meta-analysis of 21 randomized clinical trials [74].The results were in favor of group self-assertiveness training and psychoeducation programs but with a low level of confidence in the quality of the trials [74].
We found that studies including higher rates of single participants with schizophrenia reported higher Perceived Stigma and lower Alienation scores.These results were provided by studies including middle-aged participants, and age and sex were not associated with personal stigma.However, this discrepancy should be interpreted with caveats, as the studies assessing perceived stigma (using the PDD scale) were not the same that those assessing alienation (using the ISMI scale), and were therefore not carried out in the same countries nor at the same time.Future studies should therefore administer both PDD and ISMI scales to explore the relationships between perceived stigma and other stigma dimensions.Although requiring replication, these results suggest that promoting intimate relationships may help people with schizophrenia coping with the negative emotions associated with stigma but that an intimate relationship may also play a role in the fear of stigma.Including spouses in interventions targeting stigma may therefore improve their effectiveness.Perceived stigma and stereotype endorsement were also higher in unemployed patients, which suggests that promoting work rehabilitation may improve these stigma dimensions in schizophrenia [75].
Moreover, we found that lower functioning and higher antipsychotic daily dose were associated with higher experienced stigma and social withdrawal, i.e., the external consequences of stigma.We may hypothesize that both poorer functioning and daily antipsychotic dose may result in more visible functional handicap and therefore in more objective/external consequences of stigma.However, these results should be interpreted with caution, as functioning was only explored in two Indian studies published by the same team [27,29] and one study carried out in China [30].Antipsychotic dose was calculated in two Indian studies [27,29] and one German study [47].Therefore, these results need replication to be confirmed.
Finally, the important aspect of stigma resistance has received insufficient attention.In this meta-analysis, stigma resistance increased with the percentage of low-income patients, the mean illness duration, and the mean positive symptoms score.These results should be explored with individual data to confirm these associations at the individual level.The multifaceted topic of resilience requires more research in severe mental disorders, including schizophrenia [76].A recent meta-analysis on resilience in schizophrenia identified that overall resilience was related to lower levels of self-stigma and higher self-esteem [77].
The results of this meta-analysis need to be interpreted within its limitations.The results of the meta regressions analyses may not be true at the individual patient level and meta-regression cannot substitute meta-analysis of individual participant data.Altogether, we have identified a strikingly low number of sociodemographic and illness factors explaining the heterogeneity identified in self-stigma and its dimensions.This finding suggests that most studies failed to assess in detail the factors influencing self-stigma.The absence of sufficient data to explore the role of depression is a major limitation in the field.Major depression is identified in about one-third of people with schizophrenia [78,79] and predicts multiple poor outcomes, including impaired quality of life [80].It is highly probable that depression would play a role in at least the emotional and cognitive dimensions of self-stigma [43].Better illness insight and better cognition have also been associated with a greater likelihood of depression and lower quality of life in schizophrenia [81][82][83] and may also play a role in self-stigma [43].However, these factors were insufficiently reported to be included in our analyses.The list of insufficiently explored factors is extremely long but includes for example immigrant status, history of incarceration or homelessness, parenting status, parental/caregiver insight, and stigma, selfperception of clinical severity, perceived and observed cognitive dysfunction, attributions of personal responsibility, shame proneness, avoidant or self-defeating personality traits, attachment style, self-compassion, all different quality of life dimensions, metacognitive abilities, social network/functioning/power, psychological flexibility, fear of negative evaluation, coping strategies, treatment alliance and adherence, mental health, and reading/writing literacy, and cultural factors (e.g., attributing mental illness to supernatural causes and loss of face in Eastern countries).We have excluded an important number of studies due to an inadequate calculation of stigma scores, which may have induced a bias in the results.To limit the selection bias of residual studies, we have included all studies regardless of their cross-cultural validation, which can be considered as a limit.Only two validated scales could be included due to their psychometric properties (i.e., use of a standardized global score).Different versions of the ISMI scale have been used in the studies with a variable number of items.This may have also contributed to the heterogeneity of the results.However, the leaveone-out method did not reveal that this study significantly influenced the stigma score.Moreover, we have assessed the stigma dimensions separately to address this point, given that some studies included stigma resistance (29 items) in their total score while some others did not (24 items, as recommended).While we have explored three large databases, the databases Embase® and PsychInfo® were not included in the search strategy.

CONCLUSION
This comprehensive meta-analysis found that most studies published since 2007 report a moderate perceived stigma and a mild overall self-stigma level in patients with schizophrenia, but with high heterogeneity across studies, suggesting that the level of self-stigma varies across specific subgroups of patients.The subgroup with the highest risk of self-stigma is characterized by unemployment, high antipsychotic daily dose, and low functioning.We need additional individual data meta-analyses to confirm the factors associated with increased self-stigma.We have identified important other missing factors that should be explored to improve the understanding of moderators of self-stigma and, thus, the effectiveness of more targeted public and personalized interventions.An important message is that classical illness severity indices (e.g., psychotic severity, age at illness onset, illness duration) and sociodemographic variables (age, sex, and education) were not significantly associated with stigma, moderating previous findings.However, since study of origin and other patient population and study design features can influence the findings, clearly, more studies are needed to better understand the frequency, severity, and determinants of self-stigma in patients with schizophrenia.

Fig. 1
Fig. 1 PRISMA flow chart.The search paradigm was based on the Mesh terms of the Medline® database (using Pubmed®) and adapted for Google Scholar® and Web of Science®.Overall, 37 studies, published 2007-2020, exploring self-stigma dimensions in schizophrenia (n = 7717 across 25 countries) were included in the random-effects meta-analysis.

Fig. 2
Fig. 2 Forest plot of studies exploring stigma dimensions (experienced stigma, alienation, stereotype endorsement, social withdrawal, and stigma resistance) in schizophrenia.A Perceived stigma B Experienced stigma C Alienation D Stereotype endorsement E Social withdrawal F Stigma resistance.
Sociodemographic factors (continuous): age, percentage of males, percentage of participants with high education level, percentage of single/unemployed/White or Caucasian/African-American/Black/Latino/ Asian/Urban/Living alone/low-income participants.Illness characteristics (continuous): mean age at illness onset (years), mean illness duration (years), mean duration of untreated psychosis (years), percentage of patients with lifetime psychiatric hospitalizations, percentage of participants with schizoaffective disorder, psychotic symptomatology measured by the Positive symptoms (measured with the Positive and Negative Syndrome Scale for schizophrenia (PANSS) Stigma scores (continuous): Perceived stigma, Experienced stigma, Alienation, Stereotype Endorsement, Social Withdrawal, Stigma Resistance scores.

Table 1 .
Characteristics of the included studies.

Table 2 .
Factors associated with self-stigma dimensions: meta-regression analyses.