Candidate Factors Maintaining Social Anxiety in the Context of Psychotic Experiences: A Systematic Review

Abstract

Social anxiety is common in psychosis and associated with impaired functioning, poorer quality of life, and higher symptom severity. This study systematically reviewed factors maintaining social anxiety in people with attenuated, transient, or persistent psychotic experiences. Other correlates of social anxiety were also examined. MEDLINE, Embase, CENTRAL, and PsycINFO were searched for relevant literature up to October 19, 2020. Forty-eight articles were eligible for narrative synthesis: 38 cross-sectional studies, 8 prospective studies, 1 uncontrolled trial, and 1 qualitative study. From 12060 participants, the majority was general population (n = 8771), followed by psychosis samples (n = 2532) and those at high risk of psychosis (n = 757). The methodological quality and risk of bias were assessed using the Mixed Methods Appraisal Tool. Ninety percent of studies were rated as high to very-high quality. Poorer quality studies typically failed to adequately control for confounds and provided insufficient information on the measurement validity and reliability. Prominent psychological factors maintaining social anxiety included self-perceptions of stigma and shame. Common correlates of social anxiety included poorer functioning and lower quality of life. In conclusion, stigma and shame could be targeted as a causal mechanism in future interventional studies. The integration of findings from this review lead us to propose a new theoretical model to guide future intervention research.

Introduction

Social anxiety disorder (SAD) is a common mental health problem for people at risk of psychosis (prevalence 6.1%–42.3%)^1–3 or with an established psychotic disorder (pooled prevalence 16%–26%).⁴ SAD is characterized by exaggerated fears of evaluation by others, leading to distress and/or avoidance of social interactions.⁵ It is a disabling disorder and a preceding cause of anxiety, affective, and substance dependence/abuse disorders.⁶ Many people with schizophrenia report having problems with social relationships and activities.⁷ With comorbid SAD, people with schizophrenia report significantly lower functioning, lower self-esteem, higher symptom severity,⁸ poorer quality-of-life (QoL),⁹ higher depression,⁴ and higher rates of suicide attempts.¹⁰ Despite SAD being a significant problem for people with psychosis,^4,11 there has been little treatment-relevant research.¹²

Cognitive behavioral therapy (CBT) is a recommended psychological intervention for people with schizophrenia,^13,14 effectively reducing psychotic symptoms in people with psychosis or those at risk of psychosis.^15–21 In addition to the evidence that CBT is the treatment of choice for a single diagnosis of SAD,^22,23 the mechanisms of therapeutic change are increasingly well understood. In particular, the use of experiential exercises to help people with SAD learn the adverse effects of self-focused attention and safety-seeking behaviors are core components of recommended treatments.²⁴ However, clinical guidelines are silent on treatment choice when SAD is a comorbid condition,¹² and it remains to be ascertained how CBT for SAD in people with psychosis may reduce social anxiety.^12,25,26 Hence, further examination of the therapeutic mechanisms underpinning CBT for SAD in psychosis require further investigation.¹²

To understand mechanisms underpinning SAD and psychosis, we adhered to 3 principles recommended in the interventionist-causal model approach²⁷ to identify candidate causal factors. These are: (1) a focus on a single factor that is measurable; (2) the putative causal process is amenable to change by the causal factor; and (3) the causal factor is integrated with a theoretical understanding to guide therapy. We set out to determine, integrate, and critically analyze the evidence for psychological factors in the maintenance of social anxiety in people with psychosis. Additionally, we explored other correlates of social anxiety.

Methods

Protocol and Registration

The present systematic review was reported according to the Preferred Reporting Items of Systematic Reviews and Meta-analyzes (PRISMA).²⁸ The protocol was registered on PROSPERO and can be accessed at www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42018117616.

Search Strategy and Information Sources

Four databases were searched on October 19, 2020: Cochrane Central Register of Controlled Trials (1996 to October 2020); Embase (1947 to October 2020); Ovid MEDLINE(R) (1946 to October 2020); and PsycINFO (1806 to October 2020).

Search terms used for population were ((psychosis) or (psychotic) or (schizophreni*) or (schizoaffective) or (delusion*) or (paranoi*) or (clinical high risk*) or (ultra high risk*) or (attenuated) or (at risk mental state*) or (recent onset) or (first episode psycho*) or (early psycho*)) and outcomes were ((social anxi*) or (social phob*)). Limits were applied for English language and human. Electronic search strategies for Embase and MEDLINE are shown in supplementary table 1. A manual search was completed for identified articles from the electronic search, and their reference lists, those articles meeting criteria for inclusion were subjected to forward and backward citation to identify further eligible papers. The journal Schizophrenia Bulletin was hand-searched. Authors were contacted when published studies had insufficient data or where there was a need for more data to clarify results. We also asked active researchers for unpublished or recently submitted studies. Ten percent of study selection, data extraction and quality assessment were independently performed by 2 researchers with excellent agreement, the rest was performed by one researcher (supplementary table 2). Due to difference in study designs, we used the Mixed Methods Appraisal Tool (MMAT)–version 2018²⁹ for critical appraisal. MMAT is widely used for evaluation of study strengths and weaknesses.³⁰

Eligibility Criteria

We examined studies involving people diagnosed with psychosis, those experiencing attenuated and milder forms of psychotic experiences (eg, schizotypy), since psychotic experiences are seen in the general population,^31,32 and distributed along a continuum.^31,33

Inclusion Criteria.

(1) study samples included people diagnosed with schizophrenia and psychosis spectrum disorders or people deemed to be at high risk of developing psychosis and psychotic experiences; (2) analog studies measuring psychotic-like experiences such as paranoia; and (3) measurement of any psychological factors linked to social anxiety and psychotic experiences.

Exclusion Criteria.

(1) literature reviews, single-case series, or case reports; (2) studies of mixed diagnostic samples that do not present data in sub-groups or only provide pooled or aggregated data.

Data Synthesis

We planned a narrative synthesis due to the anticipated high heterogeneity of populations, measurements, and outcomes. Psychological “maintenance factors” that lead to the persistence of social anxiety in psychotic experiences such as stigma, low self-esteem, and metacognition were considered. We also explored factors associated with social anxiety and referred to these as “correlated factors.”

Results

Identification of the Studies

Four thousand five hundred twenty-seven records were identified through database searching and 7 records from additional sources. After duplicates were removed, 3586 records were screened, resulting in 79 full-texts to be assessed against eligibility criteria. Excluded papers with reasons are presented in supplementary table 3. A total of 48 papers were included for narrative synthesis (figure 1).

Fig. 1.

Study selection process.

Study and Participant Characteristics

Included studies were cross-sectional (n = 38), prospective (n = 8), uncontrolled trial (n = 1) and qualitative (n = 1), published between 1992 and 2020, and originated from North America (n = 15), United Kingdom (n = 10), Asia (n = 10), Europe (n = 9), Australia (n = 3) and Africa (n = 1). The total number of participants across 48 studies was 12060, of which the majority were from the general population (n = 8771); followed by people with established psychosis (n = 2532) and high psychosis risk samples (n = 757), other participant details see supplementary table 4.

Assessment of Social Anxiety or Social Phobia and Psychosis

Table 1 shows the measures used to assess the level of social anxiety/social phobia and psychosis, including their brief details and evidence of psychometric properties. The Liebowitz Social Anxiety Scale,³⁴ the Social Interaction Anxiety Scale, and the Social Phobia Scale³⁵ were most frequently used for social anxiety or social phobia assessment. The Positive and Negative Syndrome Scale,³⁶ the Scale for the Assessment of Positive Symptoms³⁷ and the Scale for the Assessment of Negative Symptoms³⁸ were most commonly used to index psychosis.

Table 1.

Measurements Used to Assess Level of Social Anxiety or Social Phobia and Psychosis

Measurements Used for Social Anxiety or Social Phobia	Frequency of Use	Measures	Items	Evidence of Reliability/Validity
Liebowitz Social Anxiety Scale (LSAS), LSAS self- rating (LSAS-SR)	18	Fear and avoidance of social situations and used mostly in the social anxiety research34 and in schizophrenia10	24	Good reliability and validity in social anxiety,110,111 and good reliability in schizophrenia.10
Social Interaction Anxiety Scale (SIAS)	14	Anxiety in interpersonal encounters, used alongside with SPS and mostly in the social anxiety research35	20	Good reliability and validity,35 good discriminant validity with SPS and SPAI.112
Social Phobia Scale (SPS)	6	Performance anxiety in situations where the individual fears being observed and scrutinized by others, used alongside with SIAS and mostly in the social phobia research35	20	Good reliability and validity,35 good discriminant validity with SIAS and SPAI112
Fear of Negative Evaluation (FNE)	3	Anxiety about being negatively evaluated by others and mostly in the social phobia research113	12	Good reliability and validity113
State trait anxiety inventory (STAI)	3	Various experiences of anxiety including social anxiety. Trait anxiety refers to persistent anxiety, while State anxiety reflects momentary anxiety114	40	Good reliability115 and validity116
Multidimensional Anxiety Questionnaire (MAQ)	3	Various experiences of anxiety including social anxiety, assessing worries about social embarrassment and social avoidance,117 used in schizophrenia118	40	Good reliability and validity in people with mental illness, 117 and good validity in schizophrenia.118
Brief Social Phobia Scale (BSPS)	1	Fear, avoidance and physiological symptoms associated with common social situations119	11	Acceptable reliability and validity119
Social Avoidance and Distress Scale (SADS)	1	Fear, discomfort, subjective distress and the avoidance of social situations and used mostly in social anxiety113	28	Good reliability and validity113
Social Phobia and Anxiety Inventory (SPAI)	1	Somatic, cognitive, and behavioral aspects of social phobia across a wide range of social situations and settings120	45	Good reliability and validity,120,121 good discriminant validity with SIAS and SPS112
Interaction Anxiousness Scale (IAS)	1	Subjective experience of anxiety associate with social interactions122	15	Good reliability and validity122
Unsicherheits-Fragebogen (U-scale)a	1	Experiences of social anxiety123	65	The scale was proved to be valid and transferable across samples124
Simulated social interaction test (SSIT)	1	Social skills responded to 8 social interactions (eg, disapproval/ criticism, social visibility/assertiveness)125	8	Good reliability and validity in schizophrenia126
Measurements Used for Characterizing Psychosis	Frequency of Use	Measures	Items	Evidence of Reliability/Validity
Positive and Negative Syndrome Scale (PANSS)	28	Psychopathology (positive, negative and emotional discomfort) in schizophrenia36	30	Good to excellent reliability127
Scale for the Assessment of Positive Symptoms (SAPS)	5	Positive symptoms of schizophrenia, used alongside with SANS37	34	Good validity and reliability37
Scale for the Assessment of Negative Symptoms (SANS)	5	Negative symptoms of schizophrenia, used alongside with SAPS38	25	Good validity and reliability128
Brief Psychiatric Rating Scale (BPRS)	3	Psychopathology during the week prior to the assessment129	18	Good validity and reliability130
Clinical Global Impression (CGI)	2	All symptomatology together (psychotic symptoms, anxiety, and depressive) in one number (CGI-severity subscale)131	1	Strong validity and good reliability, but lack of correlation coefficient with depression132
Green Paranoid Thoughts Scale–Persecutory Paranoia Subscale (GPTS)	1	Two specific subtypes of paranoia: social reference paranoia and persecutory paranoia133	16	Good validity and reliability133
Details of Threat questionnaire (DoT)	1	Nature of the perceived threat arising from persecutory delusions: the power of persecutor, the strength of delusional conviction, the perceived impact or awfulness of threat and perceived controllability of the threat134	4	NA
Community Assessment of Psychic Experiences (CAPE)	1	Lifetime prevalence of positive, negative and depressive symptoms on scales regarding frequency and distress in general population135	42	May overestimate the prevalence of positive symptoms, psychiatrists required to validate patient’s self-report136
Paranoid checklist	1	A multidimensional representation of paranoid ideation rating on frequency, conviction and distress associated with paranoia32	18	Good validity and excellent reliability32
Inventory of hostility and suspiciousness	1	Paranoia and related concepts: Interpersonal Suspiciousness/ Hostility, Negative Mood/Withdrawal, Anger/Impulsiveness, Mistrust/Wariness and Perceived Hardship/Resentment137	47	Satisfactory validity and reliability137

Note: ^aUnsicherheits-Fragebogen scale assessing for social anxiety.

Quality Assessment

Using MMAT, methodological quality of included studies ranged from 2** to 5***** quality criteria met, of which 43 studies (89.6%) were met at least 4**** quality criteria (tables 2 and 3). The most frequent limitations were the absence of expected confounding or appropriate methods to control for confounders^{10,11,25,39–45} and failure to use measures with established validity and reliability.^{39,40,44,46–52} Other reasons for lower quality were the high risk of non-response bias,^53–55 insufficient representativeness of the study population^53,56,57 and incomplete outcome data,^40,46 which decreased the generalizability of the results (supplementary table 5).

Table 2.

Studies Addressing Psychological Maintenance Factors of Social Anxiety in Psychotic Experiences Contexts

Citation	Design	Sample Characteristic (N)	Measurements 1. Diagnostic Criteria 2. Symptom Scales		Maintenance Factors	Findings	Quality Criteria Meta
			Psychosis	Social anxiety
Michail et al25	Cross-sectional	Total 135 FEP (60) FEP+SAD (20) SAD (31) NC (24)	1. ICD-10 2. PANSS	1. ICD-10 2. SIAS, SPS	Stigma - PBIQ Shame - OAS Social rank - SCS	FEP+SAD reported higher levels of PBIQ: entrapment, loss of social goals, poorer illness control and lower perceived social status (F1,79 = 14.5, F1,79 = 12 and F1,79 = 13.1 and F1,79 = 12 respectively) than FEP. Plus, FEP+SAD reported higher level of OAS (F1,135 = 123.1) and lower level of SCS (F1,135 = 49.6) than SAD. All ps < 0.01.	4****
Gumley et al58	Cross-sectional	Total 38 SZ (19) SZ+SAD (19)	1. DSM-IV 2. PANSS	1. DSM-IV	Stigma - PBIQ Self-esteem - RSES	SZ+SAD reported higher levels of PBIQ: self vs illness (F1,36 = 5.0, P < .05); entrapment (F1,36 = 12.7, P < .01); and shame (F1,36 = 10.6, P < .01)) and lower level of RSES (F1,36 = 10.2, P < .01) than SZ.	5*****
Birchwood et al59	Cross-sectional	Total 79 SZ (56) SZ+SAD (23)	1. ICD-10 2. PANSS, IS	2. SIAS, FNE	Stigma - PBIQ Social rank - SCS Shame - OAS	SZ+SAD reported less controllable of being psychosis and more entrapping (multivariate F = 15.6, P < .001), and more SCS (F = 27.4, P < .001) compared to SZ. Regarding regression analysis, the PBIQ shame (OR = 1.4, P = .038), PBIQ group fit (OR = 1.3, P = .018) and OAS (OR = 1.1, P = .039) were associated with the presence of SAD, after controlling depression.	5*****
Lysaker et al60	Prospective	SZ (78)	1. DSM-IV 2. PANSS	2. MAQ social anxiety	Stigma - ISMIS	Regarding stepwise regression, baseline ISMIS discrimination experience and PANSS negative symptoms significantly predicted MAQ social anxiety at 5 months, after controlling social anxiety at baseline (R2 = 0.45, P < .001).	5*****
Pyle et al61	Prospective	CAARMS (288)	1. CAARMS 2. GPTS-PP	2. SIAS	Stigma - PBEQ	Based on hierarchical regression, SIAS at baseline predicted SIAS at 6 months (B = 0.218, partial r = .205, t = 2.347, P = <.05). Plus, internalized stigma: negative appraisal and social acceptance experiences, did not predict SIAS at follow-up.	5*****
Vrbova et al62	Cross-sectional	Total 61 SZ (42) SZ+SAD (19)	1. ICD-10 2. PANSS, CGI	2. LSAS	Stigma - ISMIS	SZ+SAD reported higher level of ISMIS (t = 4.251, P < .0001).	5*****
Sutliff et al75	Cross-sectional	Total 42 SZ (24) SZ+SAD (18)	1. DSM-IV 2. PANSS	2. LSAS	Social rank - SCS	SZ+SAD reported lower level of SCS than SZ (t = 2.90, P = .006).	5*****
Aherne et al63	Cross-sectional	FEP (45)	1. ICD-10 2. Paranoid checklist	2. SIAS, SPS	Shame - TADS - CES - IES-R - ISS - OAS	Regression model showed TADS, CES, IES-R, ISS and OAS were associated with SPS (R2 = 0.299, F1,37 = 6.587, P < .000) and SIAS (R2 = 0.242, F1,37 = 7.134, P < .000). TADS, CES, IES-R, ISS and OAS was associated with paranoia (R2 = 0.092, F4,37 = 3.007, P = .032).	5*****
Romm et al70	Cross-sectional	FEP (144)	1. DSM-IV 2. PANSS	2. LSAS-SR	Self-esteem - RSES	Regarding regression analysis, RSES and PANSS suspiciousness were associated with LSAS-SR (B = -0.04, P = .000 and B = 0.07, P = .047, adjusted R2 = 0.46).	5*****
Romm et al71	Cross-sectional	Total 144 FEP (30) FEP+NonGSAD (46) FEP+GSAD (68)	1. DSM-IV 2. PANSS, IS	2. LSAS-SR	Self-esteem - RSES	FEP+GSAD reported lower level of RSES than FEP+NonGSAD and FEP alone (F40.39, P < .001).	5*****
Lysaker et al72	Prospective	SZ (39)	2. PANSS	2. LSAS	Self-esteem - MSEI	Regarding regression analysis, baseline MSEI predicted LSAS at 6 months (R2 = 0.06, P < .05), after controlling for baseline LSAS (F2,36 = 17.93, P < .001).	5*****
Lecomte et al73	Cross-sectional	Total 47 SZ (25) SZ+SAD (22)	1. DSM-IV-TR	2. BSPS, SIAS	Self-esteem - SERS-SF ToM - FEIT - FEDT - METT - Emotional recognition (a real- life situation)	SZ+SAD reported lower level of SERS-SF compared to SZ (P < .01). There were no significant differences between SZ and SZ+SAD for any of the total scores for emotional recognition.	5*****
Katherine et al43	Cross-sectional	Total 48 SZ (13) SAD (13) Panic (10) NC (12)	1. DSM-IV-TR 2. PS	1. DSM-IV-TR 2. SIAS	Negative self- referent appraisals - SCQ - SAQ-R - EBS	There were no significant differences of automatic thought (SCQ), underlying assumptions (SAQ-R) and schema (EBS) between people with SZ (with persecutory delusions) and social phobia.	4****
Voges et al76	Cross-sectional	SZ (60)	1. DSM-IV 2. PANSS	1. DSM-IV 2. SPAI	Negative self- referent appraisals - SISST	Patients reported SISST negative self- statement subscale positively correlated with lower level of SPAI (r = .74, P < .004).	5*****
Wong77	Cross-sectional	SZ (137)	1. DSM-IV 2. SAPS, SANS	1. LSAS-SR	Negative self- referent appraisals - SUMD - IRIS - SPQ - SAPS	After removing all non-significant paths in the hypothetical model, the final model suggested only 2 direct paths to social anxiety: ideas of reference (standardized path coefficient β = 0.26, P = .002) and negative symptoms (β = 0.29, P < .001)	5*****
Stopa et al78	Cross-sectional (Qualitative study)	Total 18 SZ (9) Social phobia (9)	1. DSM-IV-TR	1. DSM-IV-TR	Negative self- referent appraisals (interview)	Three common themes of interpersonal threat experiences were found in both groups: participants’ experience of threat, reactions while under threat, and subsequent reflections on threat situations, as well as the superordinate theme of narrative coherence. Key differences emerged between the groups in their perceptual experiences, ability to stand back from the threat following the event, and narrative coherence.	5*****
Piccirillo et al53	Cross-sectional	General population (179)	2. GPTS	2. SIAS	Post-event processing - PEP questionnaire	Higher SIAS and higher GPTS persecutory paranoia subscale (GPTS-PP) were significantly associated with higher levels of PEP at post-social exclusion intervention (SIAS: B = 0.36, P < .001 and GPTS-PP: B = 0.16, P < .05) and 1 week later (SIAS: B = 0.09, P < .05 and GPTS-PP: B = 0.09, P < .05).	3***
Achim et al82	Cross-sectional	Total 140 SZ (29) SZ+SAD (26) NC (84)	1. DSM-IV 2. PANSS	2. LSAS	Mentalization - BICS	Across all SZ patients or when assessed separately for the SZ− or the SZ+ groups, there were no significant correlations between level of LSAS and BICS. All ps > 0.26.	5*****
Lysaker et al79	Cross-sectional	Total 88 (all SZ) Paranoia+/ Poorest ToM (14) Paranoia-/Low- middle ToM (29) Paranoia+/High- middle ToM (23) Paranoia-/Highest ToM (22)	1. DSM-IV 2. PANSS	2. LSAS	Theory of Mind - ToM test batteryb	Paranoia+/high-middleToM group reported higher levels of LSAS than other groups: paranoia+/poorestToM; paranoia-/ highestToM and paranoia-/low-middleToM (LSAS avoidance: F = 5.03, P < .01; and LSAS fear: F = 3.31, P < .05), where paranoia+ refers to significantly higher paranoia than paranoia-.	5*****
Pepper et al80	Cross-sectional	Total 199 ASD (53) EP (51) SAD (64) NC (31)	1. DSM-IV	1. ADIS-IV/V	Theory of Mind - FPRT - FBPST - FEEST - EQ - RMET - Movie Stills task (with and without face condition)	SAD reported higher score of RMET (P < .01) and Movie Still with (P < .001) and without face (P < .01) than EP. There were no significant differences of ToM (FPRT, FBPST, FEEST and EQ) between SAD and EP.	5*****
Lysaker et al81	Cross-sectional	Total 98 (All SZ) Low mastery (33) Intermediate- mastery (52) High mastery (13)	1. DSM-IV	2. MAQ social anxiety	Metacognitive mastery - MAS	Intermediate-mastery group reported more MAQ social anxiety (F = 3.48, P < .05).	5*****
Achim et al46	Cross-sectional	Total 82 SZ (29) SZ+SAD (12) NC (41)	2. PANSS	1. DSM-IV 2. LSAS	Reasoning bias - brief-IPSAQ	SZ+SAD reported significantly lower level of brief-IPSAQ externalizing bias subscale than controls. There were no significant differences of brief-IPSAQ personalizing bias subscale amongst the 3 groups (F2,79 = 0.39, P = .68).	4****
Rus-Calafell et al55	Non- randomized controlled trials	SZ (12) Intervention: avatars for social skills enhancement	1. DSM-IV-TR 2. PANSS	2. SADS, AI, SSIT	Social avoidance - SADS	When compared between pre- and post-treatment, and post-treatment and follow-up, patient reported significantly improvement of levels of social anxiety: SSIT anxiety subscale (F2,22 = 39.76, P < .05, Cohen’s d = 0.48); and SADS avoidance (F2,22 = 14.80, P < .05, Cohen’s d = 0.58).	4****
Gajwani et al85	Cross-sectional	UHR (51)	2. SIPS	2. SIAS, SPS	Attachment - RAAS	RAAS was associated SIAS and SPS (β = 0.47, P < .001, R2 = 0.22 and 0.39, P < .01, R2 = 0.15). A significant relationship between SIAS and RAAS was mediated by BDI (F 2,49 = 14.66, P < .001, R2 = 0.38).	5*****
Michail et al86	Cross-sectional	Total 135 FEP (60) FEP+SAD (20) SAD (31) NC (24)	1. ICD-10 2. PANSS	1. ICD-10 2. SIAS, SPS	Attachment - RAAS	FEP+SAD and SAD reported higher level of insecure adult attachment than FEP and NC (x21 = 38.5, P < .01).	5*****
Russo et al45	Cross-sectional	Total 120 UHR (60) NC (60)	1. CAARMS	2. SSI social anxiety subscale	Attachment - PAM anxiety and avoidance subscale	Amongst UHR, there were no significant correlations between SSI social anxiety and insecure anxiety (r = .36, P = .07), and SSI social anxiety and avoidant attachment (r = .28, P = .14).	3***
Achim et al87	Cross-sectional	Total 62 FEP (31) NC (31)	1. DSM-IV 2. PANSS	2. LSAS	Empathy - IRI	Amongst FEP, there was significant correlations between LSAS and IRI perspective taking subscale (r = −.51, P = .004).	5*****
Armando et al88	Cross-sectional	Total 169 PLEs+SAD (32) SAD (96) Control Group (41)	1. CAPE	1. DSM-IV	Intolerance of uncertainty - IUS	PLEs+SAD reported higher levels of IUS and BDI-II, BAI and CAPE negative than those SAD alone (P < .0001).	5*****

Note: ADIS, Anxiety Disorders Interview Schedule for DSM-IV or V; AI, Assertion Inventory; ASD, Autism Spectrum Disorder; BAI, Beck Anxiety Inventory; BDI, Beck Depression Inventory; BDI-II, BDI 2^nd edition; BFNE, Brief Fear of Negative Evaluation scale; BICS, Batterie Intégrée de Cognition Sociale; CAARMS, Comprehensive Assessment of At Risk Mental State; CAPE, Community Assessment of Psychic Experiences; CES, Centrality of events Scale; CGI, Clinical Global Impression; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders 4^th edition; DSM-IV-TR, DSM-IV Total Revision; EBS, Evaluative Beliefs Scale; EP, Early Psychosis; EQ, Empathy Quotient of Cambridge Behaviour Scale; ES, Effect Size; FBPST, False Belief Picture Sequencing Task; FEDT, Facial Emotion Discrimination Test; FEEST, Facial Expressions of Emotions: Stimuli and Tests; FEIT, Facial Emotion Identification Test; FEP, First Episode Psychosis; FNE, Fear of Negative Evaluation scale; FPRT, Faux Pas Recognition Task; GPTS, Green Paranoid Thoughts Scale; GSAD, Generalized SAD; IAS, Interaction Anxiousness Scale; ICD-10, International Classification of Diseases 10^th edition; IES-R, Impact of Event scale-Revised; IHS, Inventory of Hostility and Suspiciousness; IPSAQ, Internal Personal and Situational Attributions Questionnaire; IRI, Interpersonal Reactivity Index; IRIS, Ideas and Delusions of Reference Interview Scale; IS, Insight Scale; ISMIS, Internalised Stigma of Mental Illness Scale; ISS, Internalised Shame Scale; IUS, Intolerance of Uncertainty Scale; LSAS, Liebowitz Social Anxiety Scale; LSAS-SR, LSAS Self Rating version; MAQ, Multidimensional Anxiety Questionnaire; MAS, Metacognition Assessment Scale; METT, Ekman’s Micro-Expression Training Tool; MSEI, Multidimensional Self-Esteem Inventory; NC, Normal Control; OAS, Other as Shame Scale; PAM, Psychosis Attachment Measure; PANSS, Positive and Negative Syndrome Scale; PBEQ, Personal Beliefs about Experiences Questionnaire; PBIQ, Personal Beliefs about Illness Questionnaire; PEP, Post-Event Processing; PLE, Psychotic-Like Experiences; PS, Paranoia Scale; RAAS, Revised Adult Attachment Scale; RMET, Reading the Mind in the Eyes; RSES, Rosenberg Self-Esteem Scale; SAD, Social Anxiety Disorder; SADS, Social Avoidance and Distress Scale; SANS, Scale for the Assessment of Negative Symptoms; SAPS, Scale for the Assessment of Positive Symptoms; SAQ-R, Social Attitudes Questionnaire Revised; SCQ, Social Cognitions Questionnaire; SCS, Social Comparison Scale; SERS-SF, Self-Esteem Rating Scale-Short Form; SIAS, Social Interaction Anxiety Scale; SIPS, Structured Interview for Prodromal Syndromes; SISST, Social Interaction Self-Statement Test; SPAI, Social Phobia and Anxiety Inventory; SPS, Social Phobia Scale; SPQ, Schizotypal Personality Questionnaire; SSI, Schizotypal Symptoms Inventory Brief Version; SSIT, Simulated Social Interaction Test; SUMD, Scale to Assess Unawareness of Mental Disorder; SZ, SchiZophrenia spectrum disorder; TADS, Trauma And Distress Scale; ToM, Theory of Mind; UHR, Ultra High Risk; VR-CBT, Virtual-Reality-based Cognitive Behavioural Therapy

^aScoring as number of quality criteria met; for example, 4**** means 4 criteria (of totally 5) of a study design were met.

^bToM test battery includes the Hinting Test, the Bell-Lysaker Emotional Recognition Task, the eyes test, and the Picture arrangement subtest of Wechsler Adult Intelligence Scale III.

Table 3.

Studies Addressing Correlated Factors of Social Anxiety in Psychotic Experiences Contexts

Citation	Design	Sample Characteristic (N)	Measurements 1. Diagnostic Criteria 2. Symptom Scales		Correlated Factors	Findings	Quality Criteria Meta
			Psychosis	Social Anxiety
Nemoto et al52	Prospective	SZ (118)	1. DSM-IV 2. PANSS, CGI-severity scale	2. LSAS	Quality of life - WHO-QOL26 Functioning - GAF - SFS Well-being - SWNS	Regarding a stepwise regression adjusted with demographic data, change in LSAS was significantly associated with change of the outcome models in predicting WHO- QOL26 (β = -0.01, P = .005, adjusted R2 = 0.167), SFS (β = -0.33, P < .001, adjusted R2 = 0.212) and SWNS (β = -0.25, P < .001, adjusted R2 = 0.234).	4****
Kumazaki et al47	Prospective	Total 36 SZ+Worsenedb LSAS (12) SZ+Stable LSAS (24)	1. ICD-10 2. PANSS	2. LSAS	Quality of life - WHO-QOL26 Functioning - GAF - SFS	WHO-QOL26 significantly predicted level of LSAS at follow-up (adjusted 0.85, P < .05, respectively) after controlling baseline of LSAS. PANSS, SFS and GAF were not significantly associated with development of social anxiety.	4****
Vrbova et al62	Cross-sectional	Total 61 SZ (42) SZ+SAD (19)	1. ICD-10 2. PANSS, CGI	2. LSAS	Quality of life - Q-LES-Q Personality factors - TCI-R Hopelessness - ADHS	SZ+SAD reported lower level of Q-LES-Q (t = 4.863, P < .0001) and ADHS (t = 2.710, P < .01) than SZ. SZ+SAD revealed higher level of TCI-R harm avoidance and lower self-directed subscales (t = 4.203, P < .0001 and t = 4.447, P < .0001) than SZ.	5*****
Kwong et al92	Cross-sectional	SZ (159)	1. DSM-IV 2. PANSS	2. LSAS	Quality of life - SF-36 MCS and PCS subscales	Total score of LSAS significantly correlated with SF-36 MCS (r/tc = −0.484, P < .001) and PCS (r/t = −0.302, P < .001).	5*****
Lowengrub et al48	Cross-sectional	Total 50 SZ (31) SZ+SAD (19)	1. ICD-10 2. PANSS, CGI	2. LSAS	Quality of life - SQLS	Total score of LSAS significantly correlated with SQLS (r = −.47, P < .01).	4****
Huppert et al49	Cross-sectional	SZ (32)	2. PANSS, SAPS, SANS, IHS	1. DSM-IV, ADIS 2. SIAS, SPS	Quality of life - QOLI	Levels of QOLI significantly correlated with level of SPS (r = −.48, P < .01), SIAS (r = −.48, P < .01) and ADIS social phobia (r = −.42, P < .05).	4****
Blanchard et al41	Prospective	Total 52 SZ (37) NC (15)	1. DSM- III-R 2. BPRS	2. BFNE, IAS	Well-being - WB Social anhedonia - SAS	Amongst SZ, level of SAS positively correlated with level of IAS and BFNE (r = .64 and .48), while WB negatively correlated with level of IAS and BFNE (r = −.52 and −.48), all Ps < .005.	4****
Bipeta et al50	Cross-sectional	Total 64 SZ (47) SZ+SAD (17)	1. ICD-10 2. PANSS	2. SIAS	Well-being - WHO-5 Functioning - GAF	SZ+SAD reported lower level of WHO-5 (t = 2.66, P = .01) and GAF (t = 2.1437, P = .036) than SZ.	4****
Romm et al71	Cross-sectional	Total 144 FEP (30) FEP+NonGSAD (46) FEP+GSAD (68)	1. DSM-IV 2. PANSS, IS	2. LSAS-SR	Quality of life - QOLI Functioning - GAF - Premorbid adjustment scale	FEP+GSAD reported lower level of premorbid social functioning, academic functioning, GAF and QOLI (F = 7.62 and 15.13, 12.51 and 10.91, all ps < 0.001) than FEP and FEP+NonGSAD.	5*****
El-Masry et al44	Cross-sectional	Total 107 SZ (67) SZ+SAD (19) SAD (21)	1. DSM-IV 2. SAPS, SANS	2. LSAS	Quality of life - SF-36	SZ+SAD reported lower levels of SF-36 subscales: general health, vitality, social function, role-emotional and mental health than SZ, all ps < 0.05.	3***
Chudleigh et al42	Cross-sectional	Total 60 FEP (20) At risk of psychosis (20) NC (20)	1. CAARMS 2. BPRS	2. BSPS	Functioning - SFS - WHODAS	Amongst FEP, level of SFS: performance and competence of independence subscales correlated with BSPS (r = −.52 and r = −.58), plus level of WHODAS: self-care and getting along with people subscales correlated with level of BSPS (r = .71 and r = .53). All all Ps < .01.	4****
Voges et al76	Cross-sectional	SZ (60)	1. DSM-IV 2. PANSS	1. DSM-IV 2. SPAI	Functioning - SFS	SPAI significantly correlated with SFS (r = −.32, P < .001).	5*****
Pallanti et al10	Cross-sectional	Total 107 SZ (51) SZ+SAD (29) SAD (27)	1. DSM-IV 2. SAPS, SANS	1. DSM-IV 2. LSAS	Quality of life - SF-36 Functioning - SAS* Suicidality - Suicide behavior (by interview) and the number of lifetime suicide	SZ+SAD reported lower level of SAS* (F4.85, P < .04), higher number of suicide attempts (F5.19, P < .03) and lethality of suicide attempts (F34.14, P < .001) than SZ. SZ+SAD reported lower level of SF-36: general health, vitality, social functioning, role-emotional and mental health subscales (F1,78 = 8.71, 4.79, 25.41, 9.94 and 8.96; P < .01, P < .05, P < .001, P < .01 and P < .01, respectively) than SZ.	4****
Aikawa et al90	Cross-sectional	Total 207 SZ (177) SZ+SAD (30)	1. DSM-IV 2. PANSS	1. MINI 2. LSAS	Functioning - SFS	Lower level of SFS, female, younger age of onset and longer untreated duration were associated with LSAS (β = -0.42, P < .001, adjusted R2 = 0.255).	5*****
Lecomte et al73	Cross-sectional	Total 47 SZ (25) SZ+SAD (22)	1. DSM- IV-TR	2. BSPS, SIAS	Functioning - SFS	SIAS was associated with SFS engaging in conversations subscales (β = -0.61, P < .001, adjusted R2 = 0.35).	5*****
Cacciotti-Saija et al51	Cross-sectional	SZ (51)	1. DSM- IV-TR 2. SAPS, SANS	2. SIAS	Functioning - SFS	SIAS (β = -0.56, P < .001) and SANS (-0.37, P < .01) were associated with SFS (adjusted R2 = 0.66).	4****
Khaliln et al54	Cross-sectional	SZ (53)	1. ICD-9	2. U-Scale	Family factors - EMBU - AfS	Level of social anxiety (U-scale) positively correlated with a paternal rejection (P < .005), but not correlated with mother. Those with SZ who scored their key relatives as more critical and hostile attributed to themselves (Afs) reported higher scores on social anxiety dimensions: fear of failure and criticism (P < .01), social contact anxiety (P < .001), inability to refuse (P < .001) and decency (P < .01).	4****
Michail et al86	Cross-sectional	Total 135 FEP (60) FEP+SAD (20) SAD (31) NC (24)	1. ICD-10 2. PANSS	1. ICD-10 2. SIAS, SPS	Family factors - MOPS Traumatic experiences - CTQ	FEP+SAD and SAD reported higher level of traumatic experiences (CTQ: emotional abuse (F1,97 = 4.8, P < .05) and sexual abuse (F1,97 = 3.7, P < .05)) and dysfunctional parental behaviors (MOPS: paternal indifference (F1,97 = 5.6, P < .05) and paternal abuse (F1,97 = 6.1, P < .05)) than FEP and NC.	5*****
Schutters et al56	Prospective	General population (2548)	1. DIA-X⁄M- CIDI	1. DSM-IV	Personality factors - RSRI - TPQ	Regarding multinomial logistic regression analysis, people having comorbid paranoid with social phobia associated with RSRI behavioral inhibition and TPQ harm/avoidance (Relative Risk = 26.22 and 1.12, all ps < 0.001), when compared to those without a history of social phobia or paranoid symptoms.	4****
Park et al40	Cross-sectional	Total 54 SZ (27) NC (27)	1. DSM- IV-TR 2. PANSS, SAS**	2. STAI trait anxiety	Anomalous experiences - SAS	SZ reported higher level of STAI than NC in happy condition (t = -5.00, df = 42.7, P < .01). Amongst SZ, STAI correlated with SAS** in happy (r = .56, P < .01) and angry conditions (r = .54, P < .01), and with SAS in happy condition (r = .38, P < .05).	2**
Jang et al39	Cross-sectional	Total 30 SZ (15) NC (15)	2. PANSS	2. STAI state anxiety	Anomalous experiences	Virtual avatar could evoke level of STAI, showing positive correlation between the STAI and PANSS negative subscales: blunted affect (evoked by happy avatar: r = .549, P = .034; and neutral avatar: r = .536, P = .039); and passive/apathetic social withdrawal (happy avatar: r = .536, P = .039; and neutral avatar: r = .658, P = .008).	3***
Lysaker et al95	Cross-sectional	Total 71 (All SZ) WCST impaired+no delusions (39) WCST impaired+delusions (11) WCST not impaired+no delusions (15) WCST not impaired+delusions (6)	1. DSM- III-R 2. PANSS	2. LSAS, STAI	Executive functionings - WCST	Patients having impaired cognitive flexibility with significant delusion group reported higher level of LSAS (F = 4.12, P < .05) than all other groups. Subgroup analysis showed this group reporting higher on LSAS particularly fear subscale (Fisher LSD P < .05).	5*****
Rietdijk et al57	Prospective	General population (7076)	1. DSM- III-R 2. CIDI Psychosis section	2. CIDI Social anxiety section	Subclinical paranoia - CIDI Psychosis section	Of 489 subjects who did have lifetime subclinical paranoid symptoms but no lifetime social phobia at baseline, 23 subjects (4.7%) developed social phobia (OR = 4.07; 95% CI = 2.50–6.63; P < .001). The OR remained significant after controlling for neuroticism (OR = 2.62; 95% CI = 1.57–4.36; P < .001).	4****
Michail et al11	Cross-sectional	Total 111 FEP (60) FEP+SAD (20) SAD (31)	1. ICD-10 2. PANSS, DoT	1. ICD-10 2. SIAS, SPS, BFNE	Persecutory threat - DoT	FEP+SAD (n = 9/20 (45%)) had higher number of express persecutory threat (DoT) than FEP alone (n = 7/60 (11.6%)), x21 = 10.4, P < .01.	4****
Lysaker et al94	Cross-sectional	SZ (143)	1. DSM-IV 2. PANSS	2. MAQ social anxiety	Hopelessness - BHS	MAQ social anxiety significantly correlated with BHS hope (r = −.44, P < .001).	5*****

Note: ADHS, Adult Dispositional Hope Scale; ADIS, Anxiety Disorders Interview Schedule for DSM-IV; AfS, Angehbrigen-Fragebogen fur Schizophrene patienten (assessing for patient’s attitude towards him); BHS, Beck Hopelessness Scale; BFNE, Brief Fear of Negative Evaluation scale; BPRS, Brief Psychiatric Rating Scale; BSPS, Brief Social Phobia Scale; CAARMS, Comprehensive Assessment of At Risk Mental State; CGI, Clinical Global Impression; CIDI, Composite International Diagnostic Interview; CTQ, Childhood Trauma Questionnaire; DIA-X/M-CIDI, Munich-CIDI (a modified CIDI version 1.2); DoT, Details of Threat questionnaire; DSM-III-R, Diagnostic and Statistical Manual of Mental Disorders 3rd edition Revision; DSM-IV, DSM 4th edition; DSM-IV-TR, DSM-IV Total Revision; EMBU, Egna Minnen av Barndoms Uppfostran (assessing for memories of parental behavior); FEP, First Episode Psychosis; GAF, Global Assessment of Functioning scale; GSAD, Generalized SAD; IAS, Interaction Anxiousness Scale; ICD-9, International Classification of Diseases 9th edition; ICD-10, ICD 10th edition; LSAS, Liebowitz Social Anxiety Scale; LSAS-SR, LSAS Self-Rating version; MAQ, Multidimensional Anxiety Questionnaire; MINI, Mini International Neuropsychiatric Interview; MOPS, Measure Of Parental Style; NC, Normal Control; PANSS, Positive and Negative Syndrome Scale; Q-LES-Q, Quality of Life Enjoyment and Satisfaction Questionnaire; QoL, Quality of Life; QOLI, Lehman Quality Of Life Interview; RSRI, Retrospective Self-Report of Inhibition; SAD, Social Anxiety Disorder; SANS, Scale for the Assessment of Negative Symptoms; SAPS, Scale for the Assessment of Positive Symptoms; SAS, Social Anhedonia Scale; SAS*, Social Adjustment Scale score; SAS**, Schizotypal Ambivalence Scale; SF-36, 36-tem Short Form health survey (Mental and Physical Component Summary (MCS and PCS)); SFS, Social Functioning Scale; SIAS, Social Interaction Anxiety Scale; SPAI, Social Phobia and Anxiety Inventory; SPS, Social Phobia Scale; SQLS, Schizophrenia Quality of Life Scale; STAI, State Trait Anxiety Inventory; SWNS, Subjective Well-being under Neuroleptic drug treatment Short form; SZ, SchiZophrenia spectrum disorder; TCI-R, Temperament and Character Inventory-Revised; TPQ, Tridimensional Personality Questionnaire; U-Scale, Unsicherheits-Fragebogen scale (assessing for social anxiety); WB, Well-Being scale; WCST, Wisconsin Card Sorting Test; WHO-5, World Health Organisation-5 Well-Being Index; WHODAS, WHO Disability Assessment Scale II; WHO-QOL26, WHO-Quality of Life 26

^a coring as number of quality criteria met; for example, 4**** means 4 criteria (of totally 5) of a study design were met.

^bworsened means an LSAS total score a ≥30% increase from baseline.

^cr/t means Pearson’s product-mean correlation analyzes and independent t-tests were performed to examine the relationships of SF-36 scores with continuous and categorical variables.

Psychological Factors Maintaining Social Anxiety in the Context of Psychotic Experiences

Psychological factors maintaining social anxiety in people with psychotic experiences contexts were extracted and described (table 2). We divided these factors into 4 main categories: Cognitive, Metacognitive, Behavioral, and Other (supplementary table 6). Generally, the studies related to metacognitive factors revealed inconsistent patterns with social anxiety outcomes, while other factors appeared more consistent.

Cognitive Factors.

The most frequently reported factors were cognitive variables, with the most common being stigma and shame, followed by self-esteem, social rank, and negative self-referent appraisals.

Stigma and Shame

Seven studies focused on stigma and shame.^25,58–63 The presence of SAD was significantly associated with higher stigma and external shame amongst patients with First Episode Psychosis (FEP),^25,59 and schizophrenia spectrum disorders (SZ).^58,62 Amongst FEP, stigma (OR = 1.3, P = .018) and external shame (OR = 1.1, P = .039) were associated with social anxiety after controlling for depression.⁵⁹ Severity of social anxiety in FEP was significantly associated with childhood trauma; shame memories; traumatic impact from memories; and internal and external shame,⁶³ using the Trauma and Distress Scale; ⁶⁴ Centrality of Event Scale; ⁶⁵ Impact of Event Scale-Revised; ⁶⁶ Internal Shame Scale; ⁶⁷ and Other as Shamer Scale,⁶⁸ respectively. A 5-month follow-up study of SZ found that SAD at follow-up was predicted by the Discriminative Experiences of Stigma Scale⁶⁹ at baseline and negative symptoms (total R² = 0.46 and 0.42, P < .001).⁶⁰ Amongst those at risk of psychosis internalized stigma did not predict social anxiety at 6-month follow-up once baseline social anxiety was controlled for.⁶¹

Self-esteem

Five studies investigated low self-esteem in people with SAD and paranoia.^58,70–73 SZ with SAD was associated with poorer self-esteem than those without SAD.^58,73 Amongst FEP with generalized SAD, self-esteem was lower compared to FEP with non-generalized SAD and FEP without SAD.⁷¹ Generalized SAD is characterized by a more pervasive fear of most social situations, whereas non-generalized SAD is restricted to more specific situations (eg, a fear of public speaking but no experience of anxiety in casual social gatherings), according to DSM-IV.⁷⁴ Amongst FEP, SAD was associated with low self-esteem (β = −0.04, P < .001, adjusted R² = 0.46).⁷⁰ A prospective study of SZ, SAD at 6-month follow-up was predicted (P < .001) by the level of self-esteem (R² = 0.06, P < .05) after controlling SAD at baseline.⁷²

Social rank

Three studies investigated how people compare themselves to others focusing on appraisals of social rank.^25,59,75 FEP plus SAD^25,59 and SZ plus SAD⁷⁵ reported seeing themselves as having lower social rank compared to people with psychosis alone. Furthermore, FEP plus SAD reported lower social rank than those with only SAD.²⁵

Negative Self-Referent Appraisals

Negative self-referent appraisals were investigated in 4 studies^43,76,77 including one qualitative study.⁷⁸ SZ who had higher social anxiety rated themselves more negatively (r = .74, P < .001), while those with lower social anxiety rated themselves more positively (r = −.37, P < .004).⁷⁶ SZ (persecutory delusions) and social phobia showed no significant differences in automatic thoughts, underlying assumptions and core beliefs.⁴³ In people with early operationalized psychosis, ideas of reference was found directly related to social anxiety (standardized path coefficient β = 0.26, P = .002), using path analysis.⁷⁷

A qualitative approach was used to examine interpersonal threat experiences in people with SZ (persecutory delusions) and SAD, between the 2 groups there were 3 major processes including “experience of threat,” “reactions” while under threat, and subsequent “reflections” on threat situations. There were differences found only in the SZ group, which were poor metacognitive awareness in perceptual experiences, inability to stand back from the threat following the event and lack of narrative coherence.⁷⁸

Metacognitive Factors.

Six studies examined metacognitive factors in social anxiety amongst patients with psychosis. Metacognitive factors included Theory of Mind (ToM); ^73,79,80 metacognitive mastery; ⁸¹ mentalization; ⁸² or reasoning biases.⁴⁶

Starting with ToM findings, compared to FEP, people with SAD alone had higher scores for emotional recognition tasks.⁸⁰ In another study comparing those with FEP and SAD, there were no significant differences in emotional recognition.⁸⁰ Comparing SZ and SZ plus SAD there were no differences in emotional recognition.⁷³ In SZ, the level of ToM and paranoia were combined for cluster analysis. Those with SZ in the high-middleToM/paranoia+ group (where paranoia+ refers to significantly higher paranoia than paranoia-) reported greater social anxiety level than other groups (poorestToM/paranoia+; highestToM/paranoia- and low-middleToM/paranoia-).⁷⁹

With regards to metacognitive mastery,⁸³ (the ability to utilize knowledge of mental states to intentionally manage conflicts and subjective distress), SZ with intermediate levels of mastery reported higher social anxiety than those with low and high mastery group (P < .05).⁸¹ When assessed separately for the SZ with or without SAD groups, there were no significant correlations between mentalization and social anxiety.⁸²

Lastly, reasoning biases, including personalizing and externalizing biases were measured amongst 3 groups: SZ, SZ with SAD and normal control. Compared across 3 groups, there were no significant differences levels of personalizing bias. SZ with SAD reported a significantly lower level of externalizing bias than control.⁴⁶

Behavioral Factors.

Social avoidance has been examined in a single-arm trial using Virtual Reality to deliver a treatment to enhance social skills in SZ. This found improved social anxiety (effect size = 0.48, P < .05) and reduced social avoidance (effect size = 0.58, P < .05) at post-treatment and follow-up, respectively.⁵⁵

One study investigated post-event processing (PEP)⁵³–a ruminative process occurring after a distressing social event and attempts to reduce the likelihood of negative social consequences.⁸⁴ In other words, PEP is a covert behavior that functions as a safety behavior preventing disconfirmation of negative social anxiety beliefs. This study included undergraduate students in the game to assess the perception of exclusion, and 2 confederates as additional participants to act and lead participants believing that they were excluded.⁵³ The game was preset so that in the first 5 passes the participant received the ball twice, then the 2 confederates chose to toss the ball to each other for the duration of the game–the participant was socially excluded. PEP, SIAS and GPTS-PP were measured at pre- and post-social exclusion intervention, and 24-hour and 1-week followed-up. It was found that higher levels of social anxiety and paranoia predicted the higher PEP after the intervention (SIAS: B = 0.36, P < .001 and GPTS-PP: B = 0.16, P < .05) and 1-week later (SIAS: B = 0.09, P < .05 and GPTS-PP: B = 0.09, P < .05).

Other Maintenance Factors.

Other factors maintaining social anxiety in psychosis were examined, including attachment,^45,85,86 empathy⁸⁷ and intolerance of uncertainty.⁸⁸

Three studies examined self-reported attachment. FEP plus SAD or SAD alone reported better adult attachment than those with FEP and normal controls.⁸⁶ Amongst ultra-high risk (UHR) participants, an insecure adult attachment was associated with social anxiety using SIAS (β = 0.47, P < .001, R² = 0.22) and SPS (β = 0.39, P < .01, R² = 0.15) and the relationship between adult attachment and SIAS was mediated by depression.⁸⁵ However, amongst people with UHR, there were no significant correlations between social anxiety and insecure anxious attachment, or avoidant attachment.⁴⁵

Empathy was reported using the Interpersonal Reactivity Index,⁸⁹ consisting of cognitive (perspective taking and fantasy scales) and affective components (empathic concern and personal distress scales). Amongst FEP, the lower perspective-taking of empathy scale was associated with higher social anxiety (r = −.51, P = .004). Other empathy scales were not associated with social anxiety.⁸⁷

People with Psychotic-Like Experiences (PLE) with SAD reported higher Intolerance of Uncertainty (IU) than those with SAD alone and healthy controls (P < .001).⁸⁸

Correlates of Social Anxiety

Correlates were categorized into 7 groups: functioning, QoL, well-being, family factors, personality factors, anomalous experiences and others (subclinical paranoia, persecutory threat, traumatic experiences, suicidality and hopelessness, social anhedonia and executive functioning; see table 3). Evidence related to correlates of social anxiety generally showed consistent findings, associations with functioning and QoL/well-being were commonly investigated compared to others.

Functioning.

Ten studies reported on SAD and functioning across psychosis groups.^{10,42,47,50–52,71,73,76,90} FEP plus generalized SAD reported a lower level of premorbid social functioning and daily functioning compared to FEP plus non-generalized SAD or FEP alone.⁷¹ In SZ, those with SAD returned lower functioning scores than SZ alone.^10,50 The lower level of Social Functioning Scale (SFS)⁹¹ was related to the greater social anxiety amongst FEP,⁴² and SZ.⁷⁶ Furthermore, in SZ the lower SFS was associated with the higher social anxiety in a cross-sectional study (β = -0.42, P < .001, adjusted R² = 0.255)⁹⁰ and a longitudinal study (β = -0.33, P < .001, adjusted R² = 0.212).⁵² Lower social anxiety was associated with the higher SFS (β = -0.56, P < .001, adjusted R² = 0.66)⁵¹ and SFS Engaging in conversations subscale (β = -0.61, P < .001, adjusted R² = 0.35).⁷³ A prospective study of SZ reported that higher level of SFS was not associated with worsening social anxiety at 5-year follow-up, the development of worsened of social anxiety defined by an LSAS score ≥30% from the baseline value.⁴⁷

Quality of Life and Well-Being.

Nine studies examined QoL related to SAD with psychosis.^{10,44,47–49,52,62,71,92} In FEP, those with generalized SAD reported lower QoL than FEP with nongeneralized SAD or FEP alone.⁷¹ Amongst SZ, those with SAD significantly reported lower QoL than those with SZ alone.^10,44,62 Additionally, a higher severity of social anxiety was associated with a lower level of QoL.^48,49,92 In prospective studies of SZ, higher social anxiety was associated with lower QoL (β = -0.01, P = .005, adjusted R² = 0.167),⁵² and lower QoL predicted increased social anxiety at 5-year follow-up (adjusted OR 0.85, P < .05) after adjusting the baseline social anxiety.⁴⁷

When QOL is conceptualized as the broader notion of well-being, those with SZ and SAD had significantly lower well-being compared to those without SAD.⁵⁰ Also, amongst SZ higher SAD was associated with lower well-being,⁴¹ and the higher social anxiety was prospectively associated with the lower patients’ well-being (β = -0.25, P < .001, adjusted R² = 0.234).⁵²

Family Factors.

A study of FEP found that parental rearing style reported by those with SAD (FEP plus SAD or SAD alone) revealed higher dysfunctional paternal indifference (F1,97 = 5.6, P < .05) and abuse (F1,97 = 6.1, P < .05) than those without SAD (FEP alone and normal control).⁸⁶ Furthermore, higher social anxiety in SZ was significantly associated with the higher paternal rejection, but not maternal rejection. Those with SZ who scored their key relatives (eg, spouse, father or mother) as more critical and hostile towards themselves reported higher scores on social anxiety.⁵⁴

Temperament and Personality Factors.

In an analogue study, general population were interviewed using Munich-Composite International Diagnostic Interview (M-CIDI)⁹³ to define any/subclinical/clinical paranoid or phobia symptoms. They were then observed prospectively including completed temperamental and personality measurements. Lifetime comorbid condition (paranoia and social phobia symptoms) was associated with behavioral inhibition temperament (Relative Risk = 26.22, P < .001) and harm avoidance personality (Relative Risk = 1.12, P < .001) compared to individuals without a history of social phobia or paranoid symptoms.⁵⁶ In SZ, those with SAD had higher harm avoidance and lowered self-directed personality than those without SAD (t = 4.203, P < .0001 and t = 4.447, P < .0001).⁶²

Anomalous Experiences.

Two virtual reality studies examined perceptual disturbances in SZ.^39,40 The first study provided avatars with happy and neutral face conditions to evoke patients’ social anxiety. Amongst SZ higher social anxiety was correlated with higher PANSS negative subscales: blunted affect and passive/apathetic social withdrawal, when evoked by happy (r = .55, P = .034 and r = .54, P = .039) or neutral faces (r = .54, P = .039 and r = .66, P = .008), respectively.³⁹ Another avatar study in SZ reported higher social anxiety in the happy condition, compared to normal controls (t = −5.00, P < .01). In SZ group, the higher social anxiety was related to the higher schizotypal ambivalence (r = .56, P < .01) and social anhedonia scores (r = .38, P < .05) when evoked by happy conditions, and related to the higher schizotypal ambivalence scores (r = .54, P < .01) when evoked by angry conditions.⁴⁰

Other Factors.

In a general population prospective study, subclinical paranoid symptoms were a predictor of the development of social phobia, controlling for neuroticism (OR = 2.62; 95%CI = 1.57–4.36; P < .001).⁵⁷Amongst FEP, those with SAD expressed more persecutory threat than those with FEP alone.¹¹ Considering reported traumatic experiences, people with SAD (FEP plus SAD or SAD alone) reported higher emotional abuse (F1,97 = 4.8, P < .05) and sexual abuse (F1,97 = 3.7, P < .05) than those without SAD (FEP alone and normal controls).⁸⁶

Regarding suicidality and hopelessness, those with SZ and SAD reported a higher number of suicide attempts (F5.19, P < .03) and lethality of suicide attempts (F34.14, P < .001) compared to SZ alone.¹⁰ SZ with SAD reported lower hope than those without SAD (t = 2.710, P < .01),⁶² and the lower hope was associated with higher social anxiety (r = −.44, P < .001).⁹⁴Social anhedonia was investigated in SZ, where greater social anhedonia correlated with higher social anxiety.⁴¹

SZ were investigated for executive functioning, delusion severity and social anxiety. SZ who had impaired cognitive flexibility plus a significant delusion (PANSS Delusions Score≥5) reported higher social anxiety (LSAS: F = 4.12, P < .05) than other groups (impaired cognitive flexibility plus no delusion, not impaired plus no delusion, not impaired plus delusion).⁹⁵

Discussion

This review sought to identify, describe, and critically analyze candidate factors that maintain social anxiety in people experiencing psychosis. We synthesized the data using interventionist-casual model criteria that stipulate the candidate factors should be: (1) measurable; (2) amenable to change in a putative casual process; and (3) theoretically relevant. We also justified the factors and developed an integrated-theoretical model for improvement of targeted treatment of SAD with psychosis.

Psychological Maintenance Factors

We identified a number of factors from the eligible studies included in the current review. We clustered the findings according to Cognitive, Metacognitive and Behavioral factors. Amongst people with psychosis or schizophrenia who had an additional diagnosis of SAD, there were higher levels of perceived stigma and shame, lower levels of self-esteem and social rank and more negative self-appraisals. These findings were derived from high quality studies.

Although there were identified metacognitive factors including ToM, metacognitive mastery, mentalization and reasoning biases, not all relationships between social anxiety and metacognition were linear. This is perhaps because those people with a lower level of metacognition might not be aware of a socially feared event, while those with higher level might have a better adaptation to deal with problems with social anxiety, resulting in reduced severity, when compared to those with a moderate level.^79,81 It was evidenced that metacognitive beliefs were found to empirically contribute to social anxiety,⁹⁶ and metacognitive processes of people with psychosis can be changed in an experimental study.⁹⁷ Though there is promising evidence, findings on metacognitive factors were mixed and synthesis of these findings is made difficult by different approaches to the definition and measurement of metacognition.

We found limited evidence that behavioral factors have been systematically investigated. This is a neglected area of research and our findings show promise in delineating the role of social avoidance and other defensive behaviors (ie, PEP) in the maintenance of social anxiety. Because safety behaviors, such as social avoidance play a role in maintaining social anxiety,⁸⁴ then intervention on these factors should reduce social anxiety experiences in psychotic contexts.

Importantly, although largely findings were drawn from cross-sectional studies, we found consistent evidence for the potential role of cognitive factors, which the candidate factors can be the stigma and shame. Because they fit with the substantial characteristics of potential mechanism in the interventionist-causal approach,²⁷ which the stigma and shame were measurable^67,68,98 and can be developed in the theoretical understanding to guide therapy.⁵⁹ Furthermore, these factors are likely to be amenable to change with psychological interventions targeting these factors as a causal mechanism. Therefore, cognitive factors such as appraisals of stigma and shame may be amenable for the development of interventionist-causal approaches to SAD in psychosis.

Correlates of Social Anxiety in Psychosis

Social anxiety was frequently associated with 2 correlates including poorer functioning and lower QoL, followed by lower well-being, family factors and personality factors, anomalous experiences, and other correlates presented in the result section. From our review, lower functioning was consistently associated with higher social anxiety amongst people with psychosis. The poor functioning also influences the defeatist performance belief (DPB),⁹⁹ which is overgeneralized negative thoughts about one’s ability to successfully perform tasks. This DPB is important because it can lead to preventing the initiation of goal-directed behaviors and engagement in social interactions.⁹⁹ We also commonly found that higher social anxiety was related to poorer QoL and well-being. It is evident that lower QoL and lower well-being was associated with higher symptoms of psychosis.^100,101 Notably we found consistent evidence that social anxiety was correlated with poorer functioning and QoL. It is important that functioning and QoL should be included as outcomes in future intervention studies targeting SAD in psychosis.^52,102

Integration of Theoretical Model and its Implication

Based on our findings we propose a theoretical integration as shown in figure 2, based on previous work on social anxiety; ^84,103 paranoia; ³² and stigma.⁵⁹

Fig. 2.

The cognition model processing of social anxiety to severe threat. Note: the gradient shading box shows the intensity continuum of social anxiety to paranoia; the darker color the more paranoia, as followed: Perceived self as ridiculed/embarrassment (eg, I look awkward/sick); Ideas of reference (eg, people talking about me); Mild threat and harm from others (eg, people trying to cause minor distress, such as irritation); Moderate threat and harm (eg, people deliberately trying to approach me, such as being hostile towards me); and Severe threat and harm (eg, people trying to cause significant physical psychological or social harm).

Three major processes of the model were constructed. We will use stigma and shame to explain the model. With the proximal social assumptions, individuals with bio-psycho-social vulnerabilities are, firstly, aware that other people are critical when encountering feared social situation. Due to negative processing the self as a social object, individuals may feel different, vulnerable or stigmatized, the internalized negative self-representation is formed.

Secondly, activation of the internalized self-representations, are then subject to metacognitive processing. The individual with internal stigma- and shame-based representations may perceive their self as ridiculed (eg, I look awkward), or at risk of social harm from others (eg, others are threatening). One can perceive threat at different level consistent with the hierarchy model (see the shading box in figure 2).³² Then, their perceived assumptions will be assessed relating to: social attitudes, called other-to-self focus (eg, neighbors disgust people like me); and self-image, called self-to-self focus (eg, I am indeed despicable).

Lastly, negative appraisals about stigma and shame result in defensive behaviors (eg, avoiding eye contact), and cognitive (eg, hypervigilance due to anticipating other attitudes) and physical symptoms (eg, sweating, tremor) of anxiety. These symptoms interact in vicious circles via PEP. These defensive behaviors also maintain and prevent disconfirmation of the negative belief of social anxiety in psychosis. Negative consequences may appear as poorer daily functioning, QoL, well-being, and increased hopelessness and suicidality. Additionally, although negative affect can be a negative consequence, nonetheless, it was not included in the model, because negative affect also increases accessibility of negative appraisals and feelings of stigma/shame that, in turn, increases social anxiety.

Our model aims to help people suffering from social anxiety in the context of psychosis. Although the psychological factors related to experiencing discrimination (stigma/shame) are not unique to SAD in psychosis, these factors are very relevant in SAD in psychosis compared to the established cognitive model of social anxiety.⁸⁴ Due to our findings being drawn mostly from cross-sectional studies with limited evidence of experimental and manipulationist tests, additional studies will be needed to develop better effective treatment of SAD in psychosis. Stigma and shame should be tested in interventionist-causal manipulation designs, using social anxiety as the dependent variable and stigma/shame as the mediator variables.

Strengths and Limitations

This review has a number of strengths. The factors that maintain significant social anxiety problems in psychosis and other relevant correlates were thoroughly examined. We used rigorous methods (eg, independent study selection), took a broad and inclusive approach, and assessed the quality of the literature. But there are also limitations to be considered. We did not include non-English-language studies and unpublished gray literature which may have resulted in publication bias and exclusion of some relevant evidence. However, we believe this limitation is minimal as we utilized a comprehensive literature searching and covered studies from diverse geographical regions (Africa, Asia-Pacific, North America, and Europe). Secondly, the quality assessment, indicates that many studies did not address confounding factors and may not have proven the validity/reliability of study measures. This could lead to erroneous conclusions¹⁰⁴ and minimize trustworthiness¹⁰⁵ of a study. Majority of studies, nonetheless, were met at least 4**** (of 5*****) quality criteria. We observed a gender disparity across studies with men over-represented in the psychosis samples. In contrast, the general population and high-risk samples showed comparable proportions of male and female participants. Lastly, the heterogeneity of data prevented us from applying meta-analysis.

Directions for Future Research

Most studies were cross-sectional and conducted with Westernized English-speaking populations. Cross-cultural studies are required to improve understanding of the role that culture plays in the experience of stigma and the expression of psychopathology.¹⁰⁶ It is already known that the content of persecutory delusion is likely to depend on culturally prevalent threats or beliefs about malevolent influence^107,108 and so it is relevant to examine whether these effects extend to social anxiety related beliefs and appraisals. The development of experimental designs using interventionist-causal methods with targeted factor and focus whether modifying safety behaviors associated with reducing social anxiety in psychosis should be tested. Moreover, due to lack of evidence on other psychotic experiences, given the potential impact of psychotic experiences; for example, voices in social interactions,¹⁰⁹ this also seems to be an important topic for exploration.

Conclusion

Our analysis of the literature suggest that stigma and shame are key candidate psychological mechanisms with a strong role in maintaining social anxiety in the context of psychosis. Given the generally strong methodological quality of the included studies we can be reasonably confident that these cognitive factors warrant further investigation. For example, further studies using psychometrically robust methods and applying mediation analyses will help disentangle the different factors involved the spectrum of problems from social anxiety to paranoia. Both stigma and shame meet the criteria for being treated as relevant factors in an interventionist-causal model that we offer. This clinical model could be used as a basis for treatment development. Given that social anxiety was reliably associated with poorer functioning and QoL there is an important clinical need to improve targeted treatments for these problems.

Supplementary Material

Supplementary material is available at Schizophrenia Bulletin online.

References 138 to 167 are cited in supplementary table 3.

Funding

This study was conducted primarily by W.A. as part of a PhD program supervised by H.M. and A.G., under funding from Faculty of Medicine, Prince of Songkla University, Thailand.