Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Paranoia in patients attending child and adolescent mental health services:

Paranoia in patients attending child and adolescent mental health services: Objective: Paranoia may be particularly prevalent during adolescence, building on the heightened social vulnerabilities at this age. Excessive mistrust may be corrosive for adolescent social relationships, especially in the context of mental health disorders. We set out to examine the prevalence, symptom associations, and persistence of paranoia in a cohort of young people attending child and adolescent mental health services. Method: A total of 301 patients (11–17 years old) completed measures of paranoia, affect, peer difficulties and behav- ioural problems. Clinicians also rated each participant’s psychiatric symptoms. Patterns of association were examined using linear regressions and network analyses. In total, 105 patients repeated the measures several months later. Results: Most of the adolescents had affective disorders (n = 195), self-harm/suicidality (n = 82), or neurodevelopmental conditions (n = 125). Few had suspected psychosis (n = 7). Rates of paranoia were approximately double compared with previous reports from the general population. In this patient sample, 35% had at least elevated paranoia, 15% had at least moderate paranoia, and 6% had high paranoia. Paranoia had moderate associations with clinician-rated peer difficulties, self-harm, and trauma, and small associations with clinician-rated social anxiety, depression, generalised anxiety, and edu- cational problems. Network analyses showed paranoia had the strongest unique relationship with peer difficulties. Paths from peer difficulties to anxiety, self-harm, post-traumatic stress disorder symptoms, and behavioural problems were all via paranoia. Both self-harm and post-traumatic stress disorder were solely associated with paranoia in the network. Paranoia remained persistent for three-quarters and was associated with greater psychological problems over time. Conclusion: Paranoia is relatively common and persistent across a range of clinical presentations in youth. When para- noia occurs alongside emotional problems, important peer interactions may be adversely affected. Wider consideration of paranoia in adolescent patients is needed. Keywords Youth mental health, psychotic experiences, delusions, emotional disorders, network analysis Introduction Paranoia – the unfounded idea that others deliberately intend harm – is one of the most prominent symptoms of psychotic disorders. Yet the clinical reality is that paranoia Oxford Cognitive Approaches to Psychosis, Department of Psychiatry, is rarely specific to psychosis, with evidence it occurs University of Oxford, Oxford, UK across a range of disorders (D’Agostino et al., 2019; Oxford Health NHS Foundation Trust, Oxford, UK Freeman et al., 2019a). Indeed, there is growing evidence Corresponding author: that paranoia builds upon concerns about the self (e.g. Jessica C Bird, Oxford Cognitive Approaches to Psychosis, Department social vulnerability, low self-esteem) and psychological of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 processes (e.g. threat anticipation, worry) central to many 7JX, UK. emotional disorders (Freeman, 2016). In adolescence, an Email: Jessica.bird@psych.ox.ac.uk Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1167 age when feelings of social vulnerability are typically Health Services (CAMHS). We had three objectives. The heightened, paranoia may be especially prevalent (Bird first objective was to describe the prevalence of paranoia in et al., 2019; Ronald et al., 2014). Paranoia in adolescents is this cohort using a measure specifically validated for ado- associated with a range of psychological difficulties includ- lescents and compare these rates to previous reports from ing affective symptoms, peer difficulties, behavioural prob- the general population (Bird et al., 2019). The second lems, and poor sleep (Bird et al., 2019; Taylor et al., 2015; objective was to examine the patterns of association Wigman et al., 2011; Zavos et al., 2014). Persistent para- between paranoia, psychiatric symptoms, and social func- noia has the potential to leave young people feeling unsafe tioning. To do this, the bivariate associations between para- in their daily lives, mistrustful in relationships, and iso- noia and the presence of clinician-rated symptoms were lated. The resulting effects on social relationships during first examined; then, network analysis was used to examine this sensitive period for social interaction (Orben et al., the unique relations with self-report and selected clinician- 2020) could have far-reaching impact, with evidence that rated symptoms. Network approaches can statistically esti- poor social functioning predicts the long-term persistence mate complex systems of interaction (Borsboom and of psychiatric disorders in adolescence (Ford et al., 2017). Cramer, 2013), therefore providing potential insights into To date, however, there has been extremely little detailed the mechanisms linking paranoia with other difficulties. research on paranoia in clinical populations of youth. The final objective was to examine the persistence of para- There is a substantial literature showing psychotic expe- noia in a subgroup of the cohort and its relationship with riences in general are common in adolescents accessing other difficulties over time. services, and, although transient for a number, the presence of such symptoms indicates a pluripotent risk for multiple Method psychiatric disorders and poor outcomes (Kelleher et al., 2012; McGorry et al., 2018). However, individual psy- Participants chotic experiences such as paranoia, hallucinations, grandi- Over 15 months, adolescents (11–17 years) were recruited osity, and cognitive disorganisation are separable during routine clinical appointments at a Tier 3 outpatient phenomenon (found to be distinct in factor analytic studies) CAMHS team and a Tier 4 adolescent inpatient unit based (e.g. Peralta and Cuesta, 1999; Ronald et al., 2014) that can in Oxfordshire, UK. Both services were part of Oxford occur independently of each other (e.g. Hermans et al., Health National Health Service (NHS) Foundation Trust. In 2020) and that have a degree of aetiological difference (e.g. the United Kingdom, Tier 3 CAMHS provide specialist Garety et al., 2013; Zavos et al., 2014). Individual psy- multidisciplinary assessment and treatment for adolescents chotic experiences will require a degree of difference in under 18 years with complex mental health problems and explanation and tailoring of treatment. The effects on day- Tier 4 units provide highly specialist care for under 18s to-day life may also vary – social relationships, for exam- requiring admission for severe psychiatric problems and ple, may be especially affected by paranoia due to the high levels of risk. Participants were invited to take part mistrust of others inherent in the cognitions. regardless of their reason for accessing services, clinical In recent years, significant advances have been made in diagnosis, or current treatment. The only exclusions were a the treatment of persecutory delusions in adults by adopting moderate/severe learning disability or inability to complete a targeted focus on paranoia and its contributory causal fac- questionnaires in English. Informed parental consent and tors (Freeman, 2016). Yet much of the adolescent literature child assent (11–15 years) or consent (16–17 years) was has conceptualised psychotic experiences as a single con- obtained prior to taking part. The study received approval struct, with individual symptoms primarily viewed as inter- by an NHS Research Ethics Committee (Ref: 17/SC/0539). changeable indicators of psychosis risk. As a result, studies typically include measures that sum together a broad range of psychotic experiences into a total score, with individual Measures domains often assessed to unequal degrees. Indeed, these measures typically include only one or two items for each The Bird Checklist of Adolescent Paranoia. The Bird Checklist psychotic experience, and, so, may have limited precision of Adolescent Paranoia (B-CAP; Bird et al., 2019, 2020) is for detecting (and understanding) those symptoms. Much an 18-item self-report scale for adolescents that assesses the of the adolescent literature is also biased towards the assess- frequency of paranoid thoughts in the past fortnight. Items ment of hallucinations, which is often the only consistently are rated on a 6-point scale (0 = never, 5 = all the time) with defined construct across different measurement tools, and higher scores indicating higher paranoia. Three subtypes of in many instances is used as a proxy for all psychotic expe- paranoia are assessed within an overarching single con- riences (e.g. Kelleher et al., 2017). struct: social harm, conspiracy ideas, and physical threat. Here, we adopt a targeted approach: systematically The B-CAP has very good psychometric properties includ- assessing paranoia and potential correlates in a cohort of ing strong reliability across the severity spectrum and mea- adolescents accessing UK Child and Adolescent Mental surement invariance for both age and gender in adolescents Australian & New Zealand Journal of Psychiatry, 55(12) 1168 ANZJP Articles (Bird et al., 2020). The B-CAP also demonstrates good con- coordinator or psychiatrist) completed a routine measure of current validity with other measures of paranoia and adoles- current difficulties (i.e. Current View). All three routine cent’s reports that their fears of others are excessive (Bird measures were completed as part of participant’s standard et al., 2019). We recently validated score ranges for the care. Case note diagnoses/presenting problems were B-CAP where a score of 23+ indicates mildly elevated obtained from the diagnosis section of participant’s elec- paranoia, 40+ indicates moderate paranoia, 54+ indicates tronic records, recent clinical assessment/review letters, high paranoia, and 71+ indicates severe paranoia (Bird and discussion with care coordinators. The study involved et al., 2020). an optional follow-up where the self-report questionnaires were repeated after at least 3 months for a subsample of par- The Revised Child Anxiety and Depression Scale. The Revised ticipants who were contactable and agreed to do so. Child Anxiety and Depression Scale (RCADS; Chorpita Follow-up questionnaires were completed at the clinic or et al., 2000) is a 47-item self-report questionnaire examin- online via a Qualtrics survey. ing anxiety and depression in 8- to 17-year olds. Items are rated on a 4-point scale (0 = never, 3 = always) with higher Statistical analysis scores indicating higher severity. Six subscales are pro- duced: depression, panic, obsessive compulsiveness, gen- All analyses were conducted in R, version 3.6.1 (R Core eralised anxiety, social anxiety, and separation anxiety. Team, 2013). For each questionnaire, missing values were imputed using predictive mean matching for individuals The Strengths and Difficulties Questionnaire. The Strengths with missing data for less than 20% of items. As the Current and Difficulties Questionnaire (SDQ; Goodman, 2001) is a View items were examined individually as distinct varia- 25-item mental health screening questionnaire for adoles- bles, missing values were not imputed. cents aged 11–17 years. Items are rated on a 3-point scale (0 = not true, 2 = certainly true), with higher scores indicat- Prevalence. Paranoia prevalence was assessed with mean ing greater difficulties. Four problem subscales are derived scores, item endorsement defined as a score of 2+ (i.e. comprising emotional symptoms, conduct problems, hyper- ‘couple of times’ in past 2 weeks), and the proportion scor- activity/inattention, and peer difficulties. An additional ing above validated B-CAP thresholds (Bird et al., 2020). ‘impact’ score is derived from items concerning overall Paranoia scores were compared between genders using a distress and social impairment (Goodman, 1999). The emo- t-test and the correlation between paranoia and age was tional symptoms domain was not included in the analysis examined. due to the conceptual overlap with the RCADS. Prevalence rates of paranoia in this sample were pre- sented alongside previously reported mean scores and item The Current View. The Current View (Jones et al., 2013) is endorsements on the B-CAP from a representative dataset a practitioner-completed tool assessing a wide range of of 801 adolescents aged 11–15 years (mean age = 13.3, clinical difficulties. Here, we examined clinician ratings of standard deviation [SD] = 1.16, girls = 410, boys = 382, the following psychiatric symptoms and indicators of social other gender = 9) from a secondary school in the United functioning: anxiety (separation, social, generalised, obses- Kingdom (Bird et al., 2019). Here, we report the proportion sive-compulsive disorder [OCD], panic, and agoraphobia), of adolescents from this school cohort who scored above depression, deliberate self-harm, fluctuations in mood recently validated B-CAP score ranges (Bird et al., 2020) to (bipolar), hallucinations/delusions (psychosis), post-trau- enable direct comparison with the clinical sample. matic stress disorder symptoms, substance abuse, conduct problems, emerging personality disorder, attention-deficit Clinical associations. The bivariate relationships between hyperactivity disorder (ADHD), autism spectrum disorder paranoia and the presence of clinician-rated difficulties (ASD), history of abuse/neglect, peer relationship prob- were assessed using a series of linear regressions. We did lems, persistent family relationship problems, and current not correct for non-normality in the residuals as linear educational problems. All items were coded to indicate regression models without normally distributed errors pro- presence/absence of that problem, except for educational duce valid estimates in large samples (Schmidt and Finan, difficulties where the sum of two items rating severity of 2018). For eight variables, however, weighted least squares attendance and attainment problems on a 3-point scale was (WLS) regression was used to account for heteroscedastic- used. ity in the residuals (Romano and Wolf, 2017). Standardised beta (β) estimates are presented with 95% confidence inter- vals (CIs). Procedure Network analysis was used to estimate the unique pat- Participants completed the paranoia questionnaire along- terns of association between paranoia, self-report psycho- side the routinely administered RCADS and SDQ. logical problems, and the clinician-rated presence of two Clinicians involved in each participant’s care (i.e. care distinct symptoms with clinical relevance to paranoia: Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1169 deliberate self-harm and post-traumatic stress. In a network network. Calculated using Dijkstra’s (1959) algorithm, the model, individual variables are represented by nodes, and shortest path represents the fastest route to get from one pairs of nodes may be connected by an edge that indicates node to another, taking the strength of edge weights along the presence of an association after conditioning on all different possible routes into account. Edges not required other variables (Borsboom and Cramer, 2013). for the shortest paths are suppressed, allowing a clear visu- Consequently, the lack of an edge between two variables alisation of the direct and indirect pathways between indicates an absence of a relationship once all other varia- selected variables. bles are known. For all edges, 95% CIs were constructed using a non- Due to the mixture of continuous and binary variables in parametric bootstrap with 1000 iterations in the package our data, we estimated a Mixed Graphical Model (MGM) ‘bootnet’ (Epskamp et al., 2018). The bootstrap difference using the package ‘mgm’ (Haslbeck and Waldorp, 2020). test was used to compare edge weights. Due to the regulari- Missing data was handled using listwise deletion, resulting sation, edge weights are biased towards zero and thus CIs in a sample of 218 participants with complete data on all 13 cannot be interpreted as a significance test against zero variables. To overcome potential sampling variation and (Epskamp et al., 2018). limit the estimation of spurious edges, we used a regularisa- tion technique with the Least Absolute Shrinkage and Paranoia persistence. Follow-up data were collected for Selection Operator (LASSO; Tibshirani, 1996). The paranoia and the two other self-report measures in a sub- LASSO regularisation employs a penalty by limiting the group of participants. Change in paranoia over time was sum of the partial correlation coefficients, leading to a examined using the effect size (ES) formula = M -M / pre post shrinking of estimates with some becoming exactly zero SD and a Wilcoxon signed-rank test. Individual change pre (Epskamp and Fried, 2018). The degree of regularisation is in paranoia was examined using the reliable change index controlled by the tuning parameter λ, selected using the (RCI; Jacobson and Truax, 1991) where an RCI of ±1.96 extended Bayesian information criterion (EBIC). The EBIC indicates significant change. For the RCI calculation, the hyperparameter is set between 0 and 0.5 to determine the B-CAP Cronbach’s α of 0.94 from the current sample was extent to which a parsimonious model is preferred (Foygel used. To examine the relationship between paranoia persis- and Drton, 2010), with higher values producing more cau- tence and symptoms over time, participants were split into tious estimations. We used an EBIC hyperparameter of 0.3. a persistent/increasing paranoia group (⩾23 at both times, Node predictability was also estimated to show the extent or ⩾23 at either time point with non-significant RCI) and a to which each node is predicted by its neighbouring nodes low/transient paranoia group (⩽22 at both times, or signifi- (i.e. those it shares an edge with); this represents the pro- cant decreases to ⩽22 at follow-up). Using the package portion of variance explained (R ) for continuous variables ‘lme4’ (Bates et al., 2015), linear mixed-effects models and the proportion of correct classification (CC ), or were conducted for each symptom domain with fixed total accuracy, for binary variables (Haslbeck and Waldorp, effects for paranoia group, time, and a group by time inter- 2020). We also calculated the normalised accuracy (nCC) action, and a random effect for participants. for binary variables which break down the CC to repre- total sent the additional contribution of connected nodes beyond what can be trivially predicted from the marginal intercept Results model (CC ) (Haslbeck and Waldorp, 2018). marg Participant characteristics Once estimated, the unique relations among the variables were visualised using the package ‘qgraph’ (Epskamp et al., A total of 301 adolescents took part (mean age = 15.1, 2012) in a weighted network model where the thickness and SD = 1.75). There was a higher proportion of girls (n = 184, saturation of the edge colour represents the size of the rela- 61%) than boys (n = 117, 39%) and most were White British tionship. Blue edges represent positive conditional depend- (n = 240, 80%). Participants included 271 community ence associations while red edges represent negative CAMHS patients (mean age = 15.0, SD = 1.80, girls = 164, associations. The node predictability values are visualised boys = 107) and 30 inpatients (mean age = 16.0, SD = 0.81, by a shaded ring around each node. For the binary variables, girls: n = 20, boys: n = 10). Adolescents were accessing ser- these rings are split to represent the accuracy of the intercept vices with a range of problems, although the most common model and the additional contribution of connected nodes. were affective disturbances and neurodevelopmental con- No minimum edge weight was set in the visualisation. The ditions (Table 1). Seven participants had suspected psycho- network layout was determined by the Fruchterman and sis and an additional four were noted to experience Reingold (1991) algorithm, positioning the most strongly hallucinations alongside other difficulties. Beyond those connected nodes in the centre. In a separate graph, the short- who had suspected psychosis, paranoia was recorded as a est paths between paranoia and every other variable were presenting problem in the clinical records of only one computed to highlight potential mediation pathways in the participant. Australian & New Zealand Journal of Psychiatry, 55(12) 1170 ANZJP Articles Table 1. Primary presenting problem(s) for accessing CAMHS as recorded by participant’s care team and mean paranoia scores for each problem. n Percentage Paranoia (SD) Anxiety/depression 195 65 22.0 (19.8) Emotion dysregulation, self-harm and suicidality 82 27 27.4 (19.5) Autism spectrum disorder 79 26 21.4 (21.2) Attention-deficit hyperactivity disorder 41 14 12.7 (13.2) Anger/conduct problems 30 10 17.3 (16.7) Disordered eating 24 8.0 21.2 (18.6) Trauma 23 7.6 25.5 (19.7) Sleep problems 20 6.6 21.6 (16.3) Gender identity issues 8 2.7 19.2 (18.7) Family relationship issues 8 2.7 17.8 (13.5) Psychosis 7 2.3 26.1 (23.9) Substance misuse 7 2.3 23.9 (17.4) Tic disorders 5 1.7 19.8 (30.1) Hallucinations 4 1.3 23.8 (22.6) Paranoia 1 0.3 32.0 (NA) SD: standard deviation; NA: not applicable; CAMHS: Child and Adolescent Mental Health Services. Occurring alongside other difficulties in participants without suspected psychosis. not differ between those with and without Current View rat- Prevalence ings (t = 0.20, df = 35.3, p = 0.84). A total of 275 participants Paranoid thoughts were common in this clinical sample, completed either the RCADS or the SDQ (mean age = 15.1, with item endorsement ranging from 14% to 54% (Table 2). SD = 1.75, girls: n = 171, boys: n = 104, outpatient: n = 250, The mean number of suspicions endorsed was 5.85 inpatient: n = 25). Paranoia was slightly higher in those that (SD = 5.17). Out of the 301 patients, 35% had at least mildly completed either measure (mean = 20.3, SD = 18.5) than elevated paranoia, 15% had at least moderate paranoia, 6% those who did neither (mean = 15.7, SD = 14.7), although this had at least high paranoia, and 3% had severe levels of par- difference was not significant (t = 1.60, df = 41.5, p = 0.12). anoia (Table 3). As shown in Tables 2 and 3, the rates of paranoia were approximately double those previously Clinician-rated problems. Bivariate associations between reported in a general population sample of adolescents. paranoia and the presence of each clinician-rated problem Paranoia in the patient sample was significantly higher in are shown in Table 4. The presence of peer relationship prob- girls than boys (t = 4.08, df = 288.2, p < 0.001), with 41% of lems had the strongest association with paranoia (β = 0.64, girls reporting at least mildly elevated levels compared to p < 0.001) and explained 11% of the variance in paranoia 24% of boys. There was no relationship between age and scores. The second largest association was for self-harm paranoia (r = 0.08, p = 0.16). The 30 inpatients had somewhat (β = 0.55, p < 0.001) which accounted for 7% of the variance higher paranoia scores overall (mean = 27.1, SD = 21.5) than in paranoia. Similar sized medium associations were also the community patients (mean = 19.2, SD = 17.7), although observed for post-traumatic stress symptoms (β = 0.54, this was not significant (t = 1.93, df = 33.5, p = 0.062). p = 0.001) and a history of abuse/neglect (β = 0.50, p = 0.013), although only 4% and 2% of the variance in paranoia was explained by these factors, respectively. It was notable that Clinical associations of the 104 patients with at least elevated paranoia, 38 (37%) had clinician-rated trauma (post-traumatic stress or history The clinician-rated Current View was completed for 272 par- of abuse/neglect). Depression and social anxiety showed ticipants (mean age = 15.0, SD = 1.77, girls: n = 166, boys: small but significant associations with paranoia that each n = 106, outpatient: n = 248, inpatient: n = 24). Paranoia did Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1171 Table 2. B-CAP item endorsement in CAMHS sample (n = 301) and previously reported weekly rates from the general population (n = 801). Non- CAMHS clinical Item 0 1 2 3 4 5 Weekly+ Weekly+ 1. People at school are trying to 135 33 68 37 17 11 44% 25% make me feel unwanted 2. I’m sure people are gossiping 120 39 76 31 12 23 47% 21% about me on social media 3. I am being pushed out of 124 54 63 28 22 10 41% 22% conversations on purpose 4. My friends or partner are 177 49 32 21 12 10 25% 10% ignoring my messages to upset me 5. People are trying to embarrass 185 39 31 24 9 13 26% 20% me in class on purpose 6. People are making sly 118 58 60 36 14 15 42% 16% comments to upset me 7. I think people are lying to me 93 44 74 47 19 24 54% 30% on purpose 8. People say things under their 143 43 48 33 18 16 38% 24% breath to wind me up 9. Nasty tricks are being played 216 32 30 14 1 8 18% 8% on me 10. People are trying to confuse 164 40 48 19 15 15 32% 17% me on purpose 11. Groups of people are planning 197 35 31 17 11 10 23% 10% against me 12. People are collecting my 237 21 23 7 4 9 14% 7% information or photos to use against me 13. I’m sure people are seeking 201 36 30 17 8 9 21% 11% revenge on me 14. I feel like I am being followed 212 23 26 18 9 13 22% 12% or stalked 15. I am scared of what strangers 124 50 45 35 22 25 42% 32% will do to me 16. People will try to kidnap me 193 42 26 22 11 7 22% 14% 17. I could be attacked at any time 132 54 43 21 25 26 38% 23% 18. I feel unsafe around people 149 46 37 20 23 26 35% 19% everywhere I go CAMHS: Child and Adolescent Mental Health Services; B-CAP: Bird Checklist of Adolescent Paranoia. Endorsement rates as reported in Bird et al. (2019). explained 6% of the variance. Small significant associations anxiety, respectively. The presence of ADHD symptoms accounting for only 4% and 2% of the variance in paranoia showed a small negative association that explained 2% of the were observed for educational difficulties and generalised variance in paranoia scores. Australian & New Zealand Journal of Psychiatry, 55(12) 1172 ANZJP Articles Table 3. Mean scores and proportions of CAMHS patients (n = 301) scoring above validated score thresholds compared to previously collected data from the adolescent general population (n = 801; Bird et al., 2019). CAMHS General population All Girls Boys All Girls Boys Mean score (SD) 20.0 (18.2) 23.1 (19.4) 15.0 (14.9) 12.5 (14.0) 15.8 (15.0) 8.2 (10.8) ⩽22 (average range) 197 (65%) 108 (59%) 89 (76%) 667 (83%) 314 (77%) 351 (92%) 23+ (mildly elevated+) 104 (35%) 76 (41%) 28 (24%) 134 (17%) 96 (23%) 31 (8%) 40+ (moderate+) 46 (15%) 34 (18%) 12 (10%) 52 (7%) 40 (10%) 8 (2%) 54+ (high+) 18 (6%) 15 (8%) 3 (3%) 16 (2%) 11 (3%) 3 (0.8%) 71+ (severe+) 10 (3%) 9 (5%) 1 (0.9%) 4 (0.5%) 2 (0.5%) 2 (0.5%) CAMHS: Child and Adolescent Mental Health Services; SD: standard deviation. In the general population sample, 9/801 participants identified as ‘other gender’. These participants were not included in the gender group comparison due to the limited sample size. There was a small-medium association between the pres- (p < 0.05) but not self-harm or post-traumatic stress. None ence of clinician-rated psychosis (hallucinations/delusions) of the other edges with paranoia were significantly different and higher paranoia (β = 0.47, p = 0.061, R = 0.01). This was in size (p > 0.05; supplementary Table S2). A total of 56% not statistically significant, most likely due to limited power of the variance in paranoia was explained by the direct edges with only 17 patients rated as having these symptoms; nota- with these seven variables (see supplementary Table S2 for bly, nine of these (53%) had at least mildly elevated para- predictability values of all nodes). The absence of edges in noia. Small associations that were not significant (p > 0.05) Figure 1(a) shows that paranoia was conditionally inde- and each accounted for only 1% of the variance in paranoia pendent from depression, distress/social impairment, hyper- were observed for substance abuse, emerging personality activity, generalised anxiety, and social anxiety, indicating disorder, separation anxiety, family relationship problems, primarily indirect relationships through other variables in panic, conduct problems and OCD (Table 4). The associa- the network. tions between paranoia and agoraphobia, extremes of mood, The shortest paths from paranoia to all other variables eating problems, and ASD were of a negligible size in Figure 1(b) shows the direct relationship was the domi- (β < 0.20) and non-significant (p > 0.05). nant pathway between paranoia and all seven variables for which an edge was present. The shortest path network then shows that the fastest route from paranoia to distress/social Network analysis. The fully estimated network between para- impairment was via peer difficulties, indicating a mediat- noia, self-report psychological problems and selected clini- ing role of peer difficulties in this relationship. Potential cian-rated symptoms is shown in Figure 1(a) (see supplement mediation pathways are also highlighted from paranoia to for 95% CIs of all edges). Once the contribution of all other hyperactivity via conduct problems, and to depression, variables was controlled, paranoia demonstrated the largest generalised anxiety, and social anxiety via panic. unique relationship with peer difficulties (edge weight = 0.35, Notably, paranoia was the only variable that both self- 95% CI = [0.22, 0.47]). Figure 1(a) shows paranoia also had harm and post-traumatic stress had a unique association with a key role in connecting peer difficulties with the rest of the once all other variables were controlled (Figure 1(a)). The network, with the paths from peer difficulties to four of the normalised accuracy (i.e. predictability) values suggested the anxiety domains, behavioural problems, self-harm, and post- single edge with paranoia accounted for 22% of the remain- traumatic stress all occurring via paranoia. ing accuracy of self-harm beyond what was predicted by the Paranoia also demonstrated direct edges with self-harm intercept model (nCC = 0.22; CC = 0.51; CC = 0.62). (edge weight = 0.17, 95% CI = [−0.05,0.38]), conduct prob- marg total Conversely, the edge with paranoia did not lead to any lems (edge weight = 0.17, 95% CI = [0.02, 0.31]), panic increase in accuracy beyond the intercept model for post- (edge weight = 0.14, 95% CI = [−0.01, 0.28]), post-traumatic traumatic stress (nCC = 0.00; CC = 0.75; CC = 0.75). stress (edge weight = 0.14, 95% CI = [−0.07, 0.36]), obses- marg total sive compulsiveness (edge weight = 0.11, 95% CI = [−0.03, 0.26]), and separation anxiety (edge weight = 0.08, 95% Paranoia persistence CI = [−0.05, 0.21]). The edge with peer difficulties was sig- A total of 105 participants (mean age = 15.1, SD = 1.71, girls: nificantly larger than the edges with conduct problems, n = 75, boys: n = 30) agreed to repeat the questionnaires several panic, obsessive-compulsiveness, and separation anxiety Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1173 Table 4. Associations between paranoia severity and the presence of clinician-rated problems. Problem type Clinician rating Linear regressions Absent Present β 95% CI p R n Mean n Mean Social anxiety 79 14.2 (13.7) 189 22.5 (19.4) 0.45 [0.23, 0.67] <0.001 0.06 Separation anxiety 172 18.5 (17.0) 97 23.1 (20.5) 0.25 [0.00, 0.50] 0.050 0.01 Generalised anxiety 97 16.8 (17.4) 171 21.6 (18.7) 0.26 [0.01, 0.51] 0.042 0.02 OCD 220 20.8 (19.1) 49 17.1 (14.9) –0.20 [–0.51, 0.11] 0.21 0.01 Panic 187 18.7 (18.2) 84 23.0 (18.7) 0.23 [–0.03, 0.49] 0.078 0.01 Agoraphobia 217 19.6 (18.1) 52 21.8 (19.3) 0.12 [–0.18, 0.43] 0.43 0.00 Depression 75 13.9 (13.2) 197 22.4 (19.5) 0.46 [0.25, 0.68] <0.001 0.06 Self-harm 143 15.3 (14.3) 129 25.4 (20.8) 0.55 [0.31, 0.79] <0.001 0.07 Eating problems 222 19.4 (18.6) 50 23.0 (17.3) 0.19 [–0.11, 0.50] 0.21 0.01 Psychosis 254 19.6 (18.1) 17 28.2 (21.6) 0.47 [–0.02, 0.96] 0.061 0.01 Bipolar 246 19.8 (18.6) 26 22.5 (16.4) 0.15 [–0.26, 0.55] 0.48 0.00 PTSD 199 17.7 (16.9) 63 27.5 (20.6) 0.54 [0.22, 0.85] 0.001 0.04 Abuse or neglect 221 18.6 (17.6) 43 27.9 (21.2) 0.50 [0.11, 0.89] 0.013 0.02 Conduct problems 218 19.3 (18.7) 52 23.2 (16.9) 0.22 [–0.09, 0.52] 0.16 0.01 Substance abuse 242 19.4 (18.5) 30 25.4 (16.4) 0.33 [–0.05, 0.71] 0.089 0.01 Emerging PD 208 18.9 (17.8) 62 23.6 (19.5) 0.25 [–0.03, 0.54] 0.080 0.01 Peer difficulties 98 12.5 (13.3) 173 24.2 (19.5) 0.64 [0.42, 0.85] <0.001 0.11 Family difficulties 111 17.6 (18.4) 157 22.0 (18.3) 0.24 [0.00, 0.48] 0.054 0.01 ADHD 196 21.6 (19.2) 74 15.9 (15.4) –0.31 [–0.55, –0.07] 0.010 0.02 ASD 172 19.6 (17.6) 93 21.0 (20.2) 0.08 [–0.18, 0.33] 0.56 0.00 Education problems – – – – 0.22 [0.08, 0.36] 0.002 0.04 β: standardised beta; CI: confidence interval; OCD: obsessive-compulsive disorder; PTSD: post-traumatic stress disorder; PD: personality disorder; ADHD: attention-deficit hyperactivity disorder; ASD: autism spectrum disorder. Mean paranoia scores shown with standard deviations in parentheses for those with and without each problem. Significant results highlighted in bold. Weighted least squares regression used due to heteroscedasticity in residuals. months later (mean = 21.3 weeks, SD = 6.52). The difference in least mildly elevated baseline paranoia, 30 had consistently baseline paranoia between those with follow-up data elevated or increasing scores, 5 showed significant reduc- (mean = 22.6, SD = 19.6) and those without (mean = 18.6, tions that remained in the elevated range, and 11 had sig- SD = 17.3) was small and not significant (t = 1.78, df = 190.9, nificant reductions into the average range. p = 0.077). Linear mixed-effects models showed that, compared to There was no overall difference in paranoia between those with low/transient paranoia (n = 55), across the two baseline (mean = 22.6, SD = 19.6) and follow-up (mean = time points, participants with persistent/increasing para- 23.7, SD = 19.4; V = 2296, p = 0.73, ES = 0.06). On an indi- noia (n = 50) had consistently higher levels of depression vidual basis, however, 18/105 participants had significant (β = 0.81, 95% CI = [0.45, 1.18], p < 0.001), panic (β = 0.75, increases (RCI > 1.96) in paranoia and 16/105 had signifi- 95% CI = [0.38, 1.12], p < 0.001), social anxiety (β = 0.75, cant decreases (RCI < −1.96). Of the 46 participants with at 95% CI = [0.38, 1.11], p < 0.001), generalised anxiety Australian & New Zealand Journal of Psychiatry, 55(12) 1174 ANZJP Articles Figure 1. (a) Network analysis of paranoia and other symptoms. Edges indicate positive associations and rings represent node predictability based on neighbouring nodes. Pink, blue, and orange rings (i.e. continuous variables) indicate R values. For binary (i.e. purple) variables, the shaded rings represent the proportion of correct classification, split into the accuracy of the intercept model (purple section) and the additional contribution of connected nodes (dark blue section). (b) Shortest paths from paranoia to all other variables, with dashed lines representing suppressed edges. (a) (b) Discussion (β = 0.74, 95% CI = [0.38, 1.10], p < 0.001), separation anx- iety (β = 0.64, 95% CI = [0.26, 1.02], p = 0.001), peer diffi- The adolescents attending CAMHS were primarily doing culties (β = 0.63, 95% CI = [0.24, 1.01], p = 0.002), conduct so because they had emotional disorders such as anxiety problems (β = 0.50, 95% CI = [0.11, 0.90], p = 0.014), and depression. This was to be expected. However, para- hyperactivity (β = 0.44, 95% CI = [0.04, 0.84], p = 0.032), noia was common in these young patients, with several sus- and distress/social impairment (β = 0.62, 95% CI = [0.23, picious thoughts occurring in one-third to one-half of the 1.01], p = 0.0026), but not OCD (β = 0.22, 95% CI = [−0.18, clinical cohort. Over half of patients regularly thought peo- 0.63], p = 0.28). ple were lying to them on purpose, over 40% felt scared of There were small paranoia group by time interactions at what strangers would do to them, and 35% felt unsafe eve- the threshold for significance for generalised anxiety rywhere around people. Overall, 35% reported at least (β = 0.38, 95% CI = [0.02, 0.74], p = 0.043) and social anxi- mildly elevated paranoia and 15% reported at least moder- ety (β = 0.34, 95% CI = [0.00, 0.68], p = 0.052), indicating ate paranoia. Rates of paranoia were approximately double those with persistent paranoia had somewhat less improve- those observed in adolescents from the general population ment in these symptoms compared to those with low/tran- sient paranoia. Group by time interactions were negligible (Bird et al., 2019). Previous findings that adolescent girls, and non-significant for all other domains (p > 0.05; sup- compared to boys, may be especially likely to report suspi- plementary Table S4). cious thinking were replicated (Bird et al., 2019; Ronald Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1175 et al., 2014). Although traditionally conceptualised as a to a dangerous world, though this does not mean it is inevita- symptom of psychotic disorders, paranoia in this adoles- ble or that it is without negative consequences. But our find- cent sample primarily occurred alongside common mental ings also show paranoia is certainly not confined to health problems and only a minority had suspected psycho- traumatised youth: the trauma variables only accounted for a sis. Although limited in size, the available follow-up data very small amount of the variance in paranoia and almost indicated that the paranoia was often persistent. Yet para- two-thirds of patients with paranoia did not have a (clinician- noia may well be overlooked: only one participant had the rated) history of trauma. presence of paranoia recorded in their clinical notes. Arguably one of the most important findings from the Paranoid thinking in the adolescent patients was associ- study is a close relationship between paranoia and peer ated with a wide range of clinician-rated problems includ- relationship difficulties. This association was the strongest ing anxiety, depression, trauma, self-harm, peer relationship, of all those assessed from both clinicians and patients, even and educational difficulties. Paranoia in CAMHS patients after controlling for the influence of all other variables in may therefore be expected to present in the context of emo- the network. Although the relationship will undoubtedly be tional problems, adverse life experiences, and impaired bidirectional to a degree, our previous work using a social functioning. It may also be particularly common in Bayesian approach to causal discovery found adolescent young people who self-harm: elevated paranoia was pre- peer difficulties are more likely to be influenced by para- sent in almost half of patients for whom emotion dysregula- noia than vice versa (Bird et al., 2019). This pathway is tion, self-harm, or suicidality was a primary reason for plausible, as the ability to trust is necessary for relation- accessing services. Network analysis also showed that once ships, whereas fear of other people will make it difficult to all other variables were controlled, the presence of self- socialise and make friends. We also found the most com- harm was solely associated with paranoia, with this edge mon pathway from emotional and behavioural problems to contributing to 22% of the predictability of self-harm peer difficulties occurred via paranoia, suggesting paranoia (beyond the intercept model). This relationship is consist- may be a common route to impairments in adolescent peer ent with findings from the adult literature (Freeman et al., relationships. At an age when peer acceptance is most 2019b) and evidence that self-harm is associated with psy- highly valued (Somerville, 2013), the potential impact on chotic experiences in general in adolescents (Hielscher friendships is likely to be a substantial cause of distress for et al., 2019; Martin et al., 2015). The co-occurrence of para- young people. In line with this, peer difficulties were the noia with so many different psychiatric symptoms could mediating link connecting paranoia and the overall distress also be an indicator of more severe presentations, with ado- and functional impact of young people’s problems. lescents who report persistent paranoia having greater lev- els of symptoms and social impairments over time. Limitations Consistent with a cognitive conceptualisation of paranoia as an unfounded threat belief (Freeman, 2016), network The study has several limitations. First, the sample was not analyses showed paranoia had unique associations with anxi- a fully representative cohort. It was not possible to invite all ety symptoms, especially panic. The network analysis fur- patients accessing participating services to take part, since ther demonstrated a relationship between paranoia and services could not be covered by the research team all the post-traumatic stress symptoms. Once all other variables time. However, attempts were made to minimise sampling were controlled, the presence of post-traumatic stress symp- bias by inviting patients to take part regardless of their rea- toms was solely related to paranoia. This relationship is con- son for accessing services or clinical diagnosis. The cohort sistent with evidence that negative interpersonal experiences also included a higher proportion of girls than boys, contribute to the development of paranoia (Freeman et al., although this may be representative of CAMHS given the 2013; Shevlin et al., 2015). It is important to emphasise, higher rates of common mental health problems in adoles- however, that justified fears of harm in relation to ongoing cent girls (NHS Digital, 2018). Nevertheless, the pattern of bullying or abuse is not paranoia (a term that only applies to associations between paranoia and other variables could be unfounded ideas). Paranoia in those with adversity occurs influenced by gender, and, as a result, the network structure when their concerns generalise excessively beyond specific may have biased understanding towards girls. However, experiences to the point they become clearly unfounded (e.g. there is a lack of clear evidence showing the relationships when an individual with past bullying develops a persistent between paranoia and causal factors differ by gender. concern that people are conspiring to humiliate them and Another notable source of sampling bias was the primarily interprets friendliness from others as a trick). Although sev- affluent catchment areas for the services included with a eral mechanisms driving this generalisation are likely, one local demographic of mostly White British individuals. As proposal is that negative experiences lead to learned beliefs experiences such as racism and child adversity are likely to about other people (i.e. as threatening) and the self (i.e. as contribute to the development of paranoia (Bentall et al., vulnerable) upon which paranoia flourishes (Freeman, 2012; Shaikh et al., 2016), clinical levels of paranoia in 2016). Paranoia can be an understandable protective response youth may differ by locality. Australian & New Zealand Journal of Psychiatry, 55(12) 1176 ANZJP Articles A strength of this study was the ability to compare the C.S., A.-L.T., L.C., H.J.S. and A.C.J. contributed to data collec- tion and management. D.F. and F.W. supervised the work and con- prevalence of paranoia in CAMHS patients with a repre- tributed to the design, theoretical interpretation, and writing. All sentative general population sample of adolescents using authors contributed to the final version of the manuscript prior to the same measure. This was not a perfect comparison, how- submission. ever, as the general population sample were slightly younger than the patients in this study. But as age was not Declaration of Conflicting Interests associated with paranoia in either sample, the effect of this The author(s) declared no potential conflicts of interest with difference on the comparison is likely to be minimal. respect to the research, authorship, and/or publication of this Another limitation was that aside from the B-CAP, our article. other measures were missing for approximately 10% of participants. This reflected the reality of routine measure- Funding ment in CAMHS where clinical pressures could prevent clinicians from completing the Current View and patients The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The sometimes left before completing all questionnaires. study was funded by an NIHR Research Professorship awarded to Notably, as follow-up questionnaires were collected as an D.F. (NIHR-RP-2014-05-003). The work was also supported by optional second stage of the study, only a third of the sam- the NIHR Oxford Health Biomedical Research Centre (BRC- ple provided longitudinal data. Planned prospective studies 1215-20005). This paper presents independent research funded by examining paranoia in representative clinical samples will the National Institute for Health Research (NIHR). The views be needed to understand fully the relationship over time expressed are those of the authors and not necessarily those of the with other mental health problems. NHS, the NIHR, or the Department of Health. It must also be acknowledged that a degree of measure- ment error is likely in self-report measures of paranoia. It is ORCID iD not possible in self-report questionnaires to determine if Jessica C Bird https://orcid.org/0000-0001-9457-1506 concerns about intended harm from others are unfounded. However, the B-CAP has shown good construct validity as Supplemental Material a measure of unfounded ideation, with evidence that scores Supplemental material for this article is available online. are distinct from bullying and are associated with adoles- cents’ ratings that their fears of others are excessive (Bird et al., 2019, 2020). Evidence also shows that, in general, References self-report paranoia questionnaires predict genuine para- Bates D, Mächler M, Bolker B, et al. (2015). Fitting linear mixed-effects noid ideation in controlled virtual reality settings (Freeman models using lme4. Journal of Statistical Software, 67(1), 1–48. Bentall RP, Wickham S, Shevlin M, et al. (2012) Do specific early-life et al., 2010, 2014). Nevertheless, clinical interview valida- adversities lead to specific symptoms of psychosis? A study from the tion of paranoia in young people accessing CAMHS would 2007 the Adult Psychiatric Morbidity Survey. Schizophrenia Bulletin be beneficial. 38: 734–740. Overall, this study highlights paranoia as a common, Bird JC, Evans R, Waite F, et al. (2019) Adolescent paranoia: Prevalence, potentially clinically important, and overlooked problem in structure, and causal mechanisms. Schizophrenia Bulletin 45: 1134–1142. young people who are accessing mental health services. Bird JC, Loe BS, Kirkham M, et al. (2020) The assessment of paranoia Greater awareness by clinicians of paranoia in patients in young people: Item and test properties of the Bird Checklist of attending CAMHS may be required. The use of validated Adolescent Paranoia. Schizophrenia Research 220: 116–122. tools such as the B-CAP may be helpful for clinicians to Borsboom D and Cramer AOJ (2013) Network analysis: An integrative identify paranoia within young people’s broader clinical approach to the structure of psychopathology. Annual Review of Clinical Psychology 9: 91–121. presentation and monitor change. Targeted intervention for Chorpita BF, Yim L, Moffitt C, et al. (2000) Assessment of symptoms of paranoia, suitably adapted for this age group, may then be DSM-IV anxiety and depression in children: A revised child anxiety appropriate to help young people feel safer in their daily and depression scale. Behaviour Research and Therapy 38: 835–855. lives. D’Agostino A, Rossi Monti M and Starcevic V (2019) Psychotic symp- toms in borderline personality disorder: An update. Current Opinion in Psychiatry 32: 22–26. Acknowledgements Dijkstra EW (1959) A note on two problems in connexion with graphs. We would like to thank all staff members at the South Oxfordshire Numerische Mathematik 1: 269–271. Child and Adolescent Mental Health Service and the Highfield Epskamp S and Fried EI (2018) A tutorial on regularized partial correla- Adolescent Unit, Oxford for supporting this research, and all the tion networks. Psychological Methods 23: 617–634. young people who gave their valuable time to take part. Epskamp S, Borsboom D and Fried EI (2018) Estimating psychological networks and their accuracy: A tutorial paper. Behavior Research Methods 50: 195–212. Author Contributions Epskamp S, Cramer AOJ, Waldorp LJ, et al. (2012) qgraph: Network visu- J.C.B. led the research design, data collection, data management, alizations of relationships in psychometric data. Journal of Statistical statistical analysis, and manuscript preparation. E.C.F., M.K., Software 48: 11–18. Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1177 Ford T, Macdiarmid F, Russell AE, et al. (2017) The predictors of per- Kelleher I, Keeley H, Corcoran P, et al. (2012) Clinicopathological sig- sistent DSM-IV disorders in 3-year follow-ups of the British Child nificance of psychotic experiences in non-psychotic young people: and Adolescent Mental Health Surveys 1999 and 2004. Psychological Evidence from four population-based studies. The British Journal of Medicine 47: 1126–1137. Psychiatry 201: 26–32. Foygel R and Drton M (2010) Extended Bayesian information criteria Kelleher I, Ramsay H and DeVylder J (2017) Psychotic experiences and for Gaussian graphical models. Advances in Neural Information suicide attempt risk in common mental disorders and borderline per- Processing Systems 23: 604–612. sonality disorder. Acta Psychiatrica Scandinavica 135: 212–218. Freeman D (2016) Persecutory delusions: A cognitive perspective on McGorry PD, Hartmann JA, Spooner R, et al. (2018) Beyond the ‘at risk understanding and treatment. The Lancet Psychiatry 3: 685–692. mental state’ concept: Transitioning to transdiagnostic psychiatry. Freeman D, Antley A, Ehlers A, et al. (2014) The use of immersive vir- World Psychiatry 17: 133–142. tual reality (VR) to predict the occurrence 6 months later of paranoid Martin G, Thomas H, Andrews T, et al. (2015) Psychotic experiences thinking and posttraumatic stress symptoms assessed by self-report and psychological distress predict contemporaneous and future non- and interviewer methods: A study of individuals who have been phys- suicidal self-injury and suicide attempts in a sample of Australian ically assaulted. Psychological Assessment 26: 841–847. school-based adolescents. Psychological Medicine 45: 429–437. Freeman D, Loe BS, Kingdon D, et al. (2019a) The revised Green et al., NHS Digital (2018) Mental Health of Children and Young People in Paranoid Thoughts Scale (R-GPTS): Psychometric properties, sever- England 2017. London: NHS Digital. ity ranges, and clinical cut-offs. Psychological Medicine. Epub ahead Orben A, Tomova L and Blakemore S-J (2020) The effects of social dep- of print 20 November. DOI: 10.1017/S0033291719003155. rivation on adolescent development and mental health. The Lancet Freeman D, Bold E, Chadwick E, et al. (2019b) Suicidal ideation and behav- Child & Adolescent Health 4: 634–640. iour in patients with persecutory delusions: Prevalence, symptom associa- Peralta V and Cuesta MJ (1999) Dimensional structure of psychotic symp- tions, and psychological correlates. Comprehensive Psychiatry 93: 41–47. toms: An item-level analysis of SAPS and SANS symptoms in psy- Freeman D, Pugh K, Vorontsova N, et al. (2010) Testing the continuum chotic disorders. Schizophrenia Research 38: 13–26. of delusional beliefs: An experimental study using virtual reality. R Core Team (2013) R: A Language and Environment for Statistical Journal of Abnormal Psychology 119: 83–92. Computing. R Foundation for Statistical Computing. Available at: Freeman D, Thompson C, Vorontsova N, et al. (2013) Paranoia and www.R-project.org/2013 (accessed 9 March 2020). post-traumatic stress disorder in the months after a physical assault: Romano JP and Wolf M (2017) Resurrecting weighted least squares. A longitudinal study examining shared and differential predictors. Journal of Econometrics 197: 1–19. Psychological Medicine 43: 2673–2684. Ronald A, Sieradzka D, Cardno AG, et al. (2014) Characterization of Fruchterman TMJ and Reingold EM (1991) Graph drawing by force- psychotic experiences in adolescence using the Specific Psychotic directed placement. Software: Practice and Experience 21: 1129–1164. Experiences Questionnaire: Findings from a study of 5000 16-year- Garety PA, Gittins M, Jolley S, et al. (2013) Differences in cognitive and old twins. Schizophrenia Bulletin 40: 868–877. emotional processes between persecutory and grandiose delusions. Schmidt AF and Finan C (2018) Linear regression and the normality Schizophrenia Bulletin 39: 629–639. assumption. Journal of Clinical Epidemiology 98: 146–151. Goodman R (1999) The extended version of the Strengths and Difficulties Shaikh M, Ellett L, Dutt A, et al. (2016) Perceived ethnic discrimination Questionnaire as a guide to child psychiatric caseness and consequent and persecutory paranoia in individuals at ultra-high risk for psycho- burden. Journal of Child Psychology and Psychiatry 40: 791–799. sis. Psychiatry Research 241: 309–314. Goodman R (2001) Psychometric properties of the Strengths and Shevlin M, McAnee G, Bentall RP, et al. (2015) Specificity of association Difficulties Questionnaire. Journal of the American Academy of Child between adversities and the occurrence and co-occurrence paranoia & Adolescent Psychiatry 40: 1337–1345. and hallucinations: Evaluating the stability of childhood risk in an Haslbeck JMB and Waldorp LJ (2018) How well do network models adverse adult environment. Psychosis 7: 206–216. predict observations? On the importance of predictability in network Somerville LH (2013) The teenage brain: Sensitivity to social evaluation. models. Behavior Research Methods 50: 853–861. Current Directions in Psychological Science 22: 121–127. Haslbeck JMB and Waldorp LJ (2020) mgm: Estimating time-vary- Taylor MJ, Gregory AM, Freeman D, et al. (2015) Do sleep disturbances ing mixed graphical models in high-dimensional data. Journal of and psychotic-like experiences in adolescence share genetic and Statistical Software 93: 1–46. environmental influences? Journal of Abnormal Psychology 124: Hermans K, van der Steen Y, Kasanova Z, et al. (2020) Temporal dynam- 674–684. ics of suspiciousness and hallucinations in clinical high risk and first Tibshirani R (1996) Regression shrinkage and selection via the Lasso. episode psychosis. Psychiatry Research 290: 113039. Journal of the Royal Statistical Society: Series B (Methodological) Hielscher E, Connell M, Lawrence D, et al. (2019) Association between 58: 267–288. psychotic experiences and non-accidental self-injury: Results from a Wigman JTW, Vollebergh WAM, Raaijmakers QAW, et al. (2011) nationally representative survey of adolescents. Social Psychiatry and The structure of the extended psychosis phenotype in early ado- Psychiatric Epidemiology 54: 321–330. lescence: A cross-sample replication. Schizophrenia Bulletin 37: Jacobson NS and Truax P (1991) Clinical significance: A statistical 850–860. approach to defining meaningful change in psychotherapy research. Zavos HMS, Freeman D, Haworth CMA, et al. (2014) Consistent etiology Journal of Consulting and Clinical Psychology 59: 12–19. of severe, frequent psychotic experiences and milder, less frequent Jones M, Hopkins K, Kyrke-Smithm R, et al. (2013) Current View Tool manifestations: A twin study of specific psychotic experiences in ado- Completion Guide. London: CAMHS Press. lescence. JAMA Psychiatry 71: 1049–1057. Australian & New Zealand Journal of Psychiatry, 55(12) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Australian & New Zealand Journal of Psychiatry SAGE

Paranoia in patients attending child and adolescent mental health services:

Download PDF
12 pages

Loading next page...
 
/lp/sage/paranoia-in-patients-attending-child-and-adolescent-mental-health-G5FXCDqpIS

References (81)

Publisher
SAGE
Copyright
Copyright © 2022 by The Royal Australian and New Zealand College of Psychiatrists
ISSN
0004-8674
eISSN
1440-1614
DOI
10.1177/0004867420981416
Publisher site
See Article on Publisher Site

Abstract

Objective: Paranoia may be particularly prevalent during adolescence, building on the heightened social vulnerabilities at this age. Excessive mistrust may be corrosive for adolescent social relationships, especially in the context of mental health disorders. We set out to examine the prevalence, symptom associations, and persistence of paranoia in a cohort of young people attending child and adolescent mental health services. Method: A total of 301 patients (11–17 years old) completed measures of paranoia, affect, peer difficulties and behav- ioural problems. Clinicians also rated each participant’s psychiatric symptoms. Patterns of association were examined using linear regressions and network analyses. In total, 105 patients repeated the measures several months later. Results: Most of the adolescents had affective disorders (n = 195), self-harm/suicidality (n = 82), or neurodevelopmental conditions (n = 125). Few had suspected psychosis (n = 7). Rates of paranoia were approximately double compared with previous reports from the general population. In this patient sample, 35% had at least elevated paranoia, 15% had at least moderate paranoia, and 6% had high paranoia. Paranoia had moderate associations with clinician-rated peer difficulties, self-harm, and trauma, and small associations with clinician-rated social anxiety, depression, generalised anxiety, and edu- cational problems. Network analyses showed paranoia had the strongest unique relationship with peer difficulties. Paths from peer difficulties to anxiety, self-harm, post-traumatic stress disorder symptoms, and behavioural problems were all via paranoia. Both self-harm and post-traumatic stress disorder were solely associated with paranoia in the network. Paranoia remained persistent for three-quarters and was associated with greater psychological problems over time. Conclusion: Paranoia is relatively common and persistent across a range of clinical presentations in youth. When para- noia occurs alongside emotional problems, important peer interactions may be adversely affected. Wider consideration of paranoia in adolescent patients is needed. Keywords Youth mental health, psychotic experiences, delusions, emotional disorders, network analysis Introduction Paranoia – the unfounded idea that others deliberately intend harm – is one of the most prominent symptoms of psychotic disorders. Yet the clinical reality is that paranoia Oxford Cognitive Approaches to Psychosis, Department of Psychiatry, is rarely specific to psychosis, with evidence it occurs University of Oxford, Oxford, UK across a range of disorders (D’Agostino et al., 2019; Oxford Health NHS Foundation Trust, Oxford, UK Freeman et al., 2019a). Indeed, there is growing evidence Corresponding author: that paranoia builds upon concerns about the self (e.g. Jessica C Bird, Oxford Cognitive Approaches to Psychosis, Department social vulnerability, low self-esteem) and psychological of Psychiatry, University of Oxford, Warneford Hospital, Oxford OX3 processes (e.g. threat anticipation, worry) central to many 7JX, UK. emotional disorders (Freeman, 2016). In adolescence, an Email: Jessica.bird@psych.ox.ac.uk Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1167 age when feelings of social vulnerability are typically Health Services (CAMHS). We had three objectives. The heightened, paranoia may be especially prevalent (Bird first objective was to describe the prevalence of paranoia in et al., 2019; Ronald et al., 2014). Paranoia in adolescents is this cohort using a measure specifically validated for ado- associated with a range of psychological difficulties includ- lescents and compare these rates to previous reports from ing affective symptoms, peer difficulties, behavioural prob- the general population (Bird et al., 2019). The second lems, and poor sleep (Bird et al., 2019; Taylor et al., 2015; objective was to examine the patterns of association Wigman et al., 2011; Zavos et al., 2014). Persistent para- between paranoia, psychiatric symptoms, and social func- noia has the potential to leave young people feeling unsafe tioning. To do this, the bivariate associations between para- in their daily lives, mistrustful in relationships, and iso- noia and the presence of clinician-rated symptoms were lated. The resulting effects on social relationships during first examined; then, network analysis was used to examine this sensitive period for social interaction (Orben et al., the unique relations with self-report and selected clinician- 2020) could have far-reaching impact, with evidence that rated symptoms. Network approaches can statistically esti- poor social functioning predicts the long-term persistence mate complex systems of interaction (Borsboom and of psychiatric disorders in adolescence (Ford et al., 2017). Cramer, 2013), therefore providing potential insights into To date, however, there has been extremely little detailed the mechanisms linking paranoia with other difficulties. research on paranoia in clinical populations of youth. The final objective was to examine the persistence of para- There is a substantial literature showing psychotic expe- noia in a subgroup of the cohort and its relationship with riences in general are common in adolescents accessing other difficulties over time. services, and, although transient for a number, the presence of such symptoms indicates a pluripotent risk for multiple Method psychiatric disorders and poor outcomes (Kelleher et al., 2012; McGorry et al., 2018). However, individual psy- Participants chotic experiences such as paranoia, hallucinations, grandi- Over 15 months, adolescents (11–17 years) were recruited osity, and cognitive disorganisation are separable during routine clinical appointments at a Tier 3 outpatient phenomenon (found to be distinct in factor analytic studies) CAMHS team and a Tier 4 adolescent inpatient unit based (e.g. Peralta and Cuesta, 1999; Ronald et al., 2014) that can in Oxfordshire, UK. Both services were part of Oxford occur independently of each other (e.g. Hermans et al., Health National Health Service (NHS) Foundation Trust. In 2020) and that have a degree of aetiological difference (e.g. the United Kingdom, Tier 3 CAMHS provide specialist Garety et al., 2013; Zavos et al., 2014). Individual psy- multidisciplinary assessment and treatment for adolescents chotic experiences will require a degree of difference in under 18 years with complex mental health problems and explanation and tailoring of treatment. The effects on day- Tier 4 units provide highly specialist care for under 18s to-day life may also vary – social relationships, for exam- requiring admission for severe psychiatric problems and ple, may be especially affected by paranoia due to the high levels of risk. Participants were invited to take part mistrust of others inherent in the cognitions. regardless of their reason for accessing services, clinical In recent years, significant advances have been made in diagnosis, or current treatment. The only exclusions were a the treatment of persecutory delusions in adults by adopting moderate/severe learning disability or inability to complete a targeted focus on paranoia and its contributory causal fac- questionnaires in English. Informed parental consent and tors (Freeman, 2016). Yet much of the adolescent literature child assent (11–15 years) or consent (16–17 years) was has conceptualised psychotic experiences as a single con- obtained prior to taking part. The study received approval struct, with individual symptoms primarily viewed as inter- by an NHS Research Ethics Committee (Ref: 17/SC/0539). changeable indicators of psychosis risk. As a result, studies typically include measures that sum together a broad range of psychotic experiences into a total score, with individual Measures domains often assessed to unequal degrees. Indeed, these measures typically include only one or two items for each The Bird Checklist of Adolescent Paranoia. The Bird Checklist psychotic experience, and, so, may have limited precision of Adolescent Paranoia (B-CAP; Bird et al., 2019, 2020) is for detecting (and understanding) those symptoms. Much an 18-item self-report scale for adolescents that assesses the of the adolescent literature is also biased towards the assess- frequency of paranoid thoughts in the past fortnight. Items ment of hallucinations, which is often the only consistently are rated on a 6-point scale (0 = never, 5 = all the time) with defined construct across different measurement tools, and higher scores indicating higher paranoia. Three subtypes of in many instances is used as a proxy for all psychotic expe- paranoia are assessed within an overarching single con- riences (e.g. Kelleher et al., 2017). struct: social harm, conspiracy ideas, and physical threat. Here, we adopt a targeted approach: systematically The B-CAP has very good psychometric properties includ- assessing paranoia and potential correlates in a cohort of ing strong reliability across the severity spectrum and mea- adolescents accessing UK Child and Adolescent Mental surement invariance for both age and gender in adolescents Australian & New Zealand Journal of Psychiatry, 55(12) 1168 ANZJP Articles (Bird et al., 2020). The B-CAP also demonstrates good con- coordinator or psychiatrist) completed a routine measure of current validity with other measures of paranoia and adoles- current difficulties (i.e. Current View). All three routine cent’s reports that their fears of others are excessive (Bird measures were completed as part of participant’s standard et al., 2019). We recently validated score ranges for the care. Case note diagnoses/presenting problems were B-CAP where a score of 23+ indicates mildly elevated obtained from the diagnosis section of participant’s elec- paranoia, 40+ indicates moderate paranoia, 54+ indicates tronic records, recent clinical assessment/review letters, high paranoia, and 71+ indicates severe paranoia (Bird and discussion with care coordinators. The study involved et al., 2020). an optional follow-up where the self-report questionnaires were repeated after at least 3 months for a subsample of par- The Revised Child Anxiety and Depression Scale. The Revised ticipants who were contactable and agreed to do so. Child Anxiety and Depression Scale (RCADS; Chorpita Follow-up questionnaires were completed at the clinic or et al., 2000) is a 47-item self-report questionnaire examin- online via a Qualtrics survey. ing anxiety and depression in 8- to 17-year olds. Items are rated on a 4-point scale (0 = never, 3 = always) with higher Statistical analysis scores indicating higher severity. Six subscales are pro- duced: depression, panic, obsessive compulsiveness, gen- All analyses were conducted in R, version 3.6.1 (R Core eralised anxiety, social anxiety, and separation anxiety. Team, 2013). For each questionnaire, missing values were imputed using predictive mean matching for individuals The Strengths and Difficulties Questionnaire. The Strengths with missing data for less than 20% of items. As the Current and Difficulties Questionnaire (SDQ; Goodman, 2001) is a View items were examined individually as distinct varia- 25-item mental health screening questionnaire for adoles- bles, missing values were not imputed. cents aged 11–17 years. Items are rated on a 3-point scale (0 = not true, 2 = certainly true), with higher scores indicat- Prevalence. Paranoia prevalence was assessed with mean ing greater difficulties. Four problem subscales are derived scores, item endorsement defined as a score of 2+ (i.e. comprising emotional symptoms, conduct problems, hyper- ‘couple of times’ in past 2 weeks), and the proportion scor- activity/inattention, and peer difficulties. An additional ing above validated B-CAP thresholds (Bird et al., 2020). ‘impact’ score is derived from items concerning overall Paranoia scores were compared between genders using a distress and social impairment (Goodman, 1999). The emo- t-test and the correlation between paranoia and age was tional symptoms domain was not included in the analysis examined. due to the conceptual overlap with the RCADS. Prevalence rates of paranoia in this sample were pre- sented alongside previously reported mean scores and item The Current View. The Current View (Jones et al., 2013) is endorsements on the B-CAP from a representative dataset a practitioner-completed tool assessing a wide range of of 801 adolescents aged 11–15 years (mean age = 13.3, clinical difficulties. Here, we examined clinician ratings of standard deviation [SD] = 1.16, girls = 410, boys = 382, the following psychiatric symptoms and indicators of social other gender = 9) from a secondary school in the United functioning: anxiety (separation, social, generalised, obses- Kingdom (Bird et al., 2019). Here, we report the proportion sive-compulsive disorder [OCD], panic, and agoraphobia), of adolescents from this school cohort who scored above depression, deliberate self-harm, fluctuations in mood recently validated B-CAP score ranges (Bird et al., 2020) to (bipolar), hallucinations/delusions (psychosis), post-trau- enable direct comparison with the clinical sample. matic stress disorder symptoms, substance abuse, conduct problems, emerging personality disorder, attention-deficit Clinical associations. The bivariate relationships between hyperactivity disorder (ADHD), autism spectrum disorder paranoia and the presence of clinician-rated difficulties (ASD), history of abuse/neglect, peer relationship prob- were assessed using a series of linear regressions. We did lems, persistent family relationship problems, and current not correct for non-normality in the residuals as linear educational problems. All items were coded to indicate regression models without normally distributed errors pro- presence/absence of that problem, except for educational duce valid estimates in large samples (Schmidt and Finan, difficulties where the sum of two items rating severity of 2018). For eight variables, however, weighted least squares attendance and attainment problems on a 3-point scale was (WLS) regression was used to account for heteroscedastic- used. ity in the residuals (Romano and Wolf, 2017). Standardised beta (β) estimates are presented with 95% confidence inter- vals (CIs). Procedure Network analysis was used to estimate the unique pat- Participants completed the paranoia questionnaire along- terns of association between paranoia, self-report psycho- side the routinely administered RCADS and SDQ. logical problems, and the clinician-rated presence of two Clinicians involved in each participant’s care (i.e. care distinct symptoms with clinical relevance to paranoia: Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1169 deliberate self-harm and post-traumatic stress. In a network network. Calculated using Dijkstra’s (1959) algorithm, the model, individual variables are represented by nodes, and shortest path represents the fastest route to get from one pairs of nodes may be connected by an edge that indicates node to another, taking the strength of edge weights along the presence of an association after conditioning on all different possible routes into account. Edges not required other variables (Borsboom and Cramer, 2013). for the shortest paths are suppressed, allowing a clear visu- Consequently, the lack of an edge between two variables alisation of the direct and indirect pathways between indicates an absence of a relationship once all other varia- selected variables. bles are known. For all edges, 95% CIs were constructed using a non- Due to the mixture of continuous and binary variables in parametric bootstrap with 1000 iterations in the package our data, we estimated a Mixed Graphical Model (MGM) ‘bootnet’ (Epskamp et al., 2018). The bootstrap difference using the package ‘mgm’ (Haslbeck and Waldorp, 2020). test was used to compare edge weights. Due to the regulari- Missing data was handled using listwise deletion, resulting sation, edge weights are biased towards zero and thus CIs in a sample of 218 participants with complete data on all 13 cannot be interpreted as a significance test against zero variables. To overcome potential sampling variation and (Epskamp et al., 2018). limit the estimation of spurious edges, we used a regularisa- tion technique with the Least Absolute Shrinkage and Paranoia persistence. Follow-up data were collected for Selection Operator (LASSO; Tibshirani, 1996). The paranoia and the two other self-report measures in a sub- LASSO regularisation employs a penalty by limiting the group of participants. Change in paranoia over time was sum of the partial correlation coefficients, leading to a examined using the effect size (ES) formula = M -M / pre post shrinking of estimates with some becoming exactly zero SD and a Wilcoxon signed-rank test. Individual change pre (Epskamp and Fried, 2018). The degree of regularisation is in paranoia was examined using the reliable change index controlled by the tuning parameter λ, selected using the (RCI; Jacobson and Truax, 1991) where an RCI of ±1.96 extended Bayesian information criterion (EBIC). The EBIC indicates significant change. For the RCI calculation, the hyperparameter is set between 0 and 0.5 to determine the B-CAP Cronbach’s α of 0.94 from the current sample was extent to which a parsimonious model is preferred (Foygel used. To examine the relationship between paranoia persis- and Drton, 2010), with higher values producing more cau- tence and symptoms over time, participants were split into tious estimations. We used an EBIC hyperparameter of 0.3. a persistent/increasing paranoia group (⩾23 at both times, Node predictability was also estimated to show the extent or ⩾23 at either time point with non-significant RCI) and a to which each node is predicted by its neighbouring nodes low/transient paranoia group (⩽22 at both times, or signifi- (i.e. those it shares an edge with); this represents the pro- cant decreases to ⩽22 at follow-up). Using the package portion of variance explained (R ) for continuous variables ‘lme4’ (Bates et al., 2015), linear mixed-effects models and the proportion of correct classification (CC ), or were conducted for each symptom domain with fixed total accuracy, for binary variables (Haslbeck and Waldorp, effects for paranoia group, time, and a group by time inter- 2020). We also calculated the normalised accuracy (nCC) action, and a random effect for participants. for binary variables which break down the CC to repre- total sent the additional contribution of connected nodes beyond what can be trivially predicted from the marginal intercept Results model (CC ) (Haslbeck and Waldorp, 2018). marg Participant characteristics Once estimated, the unique relations among the variables were visualised using the package ‘qgraph’ (Epskamp et al., A total of 301 adolescents took part (mean age = 15.1, 2012) in a weighted network model where the thickness and SD = 1.75). There was a higher proportion of girls (n = 184, saturation of the edge colour represents the size of the rela- 61%) than boys (n = 117, 39%) and most were White British tionship. Blue edges represent positive conditional depend- (n = 240, 80%). Participants included 271 community ence associations while red edges represent negative CAMHS patients (mean age = 15.0, SD = 1.80, girls = 164, associations. The node predictability values are visualised boys = 107) and 30 inpatients (mean age = 16.0, SD = 0.81, by a shaded ring around each node. For the binary variables, girls: n = 20, boys: n = 10). Adolescents were accessing ser- these rings are split to represent the accuracy of the intercept vices with a range of problems, although the most common model and the additional contribution of connected nodes. were affective disturbances and neurodevelopmental con- No minimum edge weight was set in the visualisation. The ditions (Table 1). Seven participants had suspected psycho- network layout was determined by the Fruchterman and sis and an additional four were noted to experience Reingold (1991) algorithm, positioning the most strongly hallucinations alongside other difficulties. Beyond those connected nodes in the centre. In a separate graph, the short- who had suspected psychosis, paranoia was recorded as a est paths between paranoia and every other variable were presenting problem in the clinical records of only one computed to highlight potential mediation pathways in the participant. Australian & New Zealand Journal of Psychiatry, 55(12) 1170 ANZJP Articles Table 1. Primary presenting problem(s) for accessing CAMHS as recorded by participant’s care team and mean paranoia scores for each problem. n Percentage Paranoia (SD) Anxiety/depression 195 65 22.0 (19.8) Emotion dysregulation, self-harm and suicidality 82 27 27.4 (19.5) Autism spectrum disorder 79 26 21.4 (21.2) Attention-deficit hyperactivity disorder 41 14 12.7 (13.2) Anger/conduct problems 30 10 17.3 (16.7) Disordered eating 24 8.0 21.2 (18.6) Trauma 23 7.6 25.5 (19.7) Sleep problems 20 6.6 21.6 (16.3) Gender identity issues 8 2.7 19.2 (18.7) Family relationship issues 8 2.7 17.8 (13.5) Psychosis 7 2.3 26.1 (23.9) Substance misuse 7 2.3 23.9 (17.4) Tic disorders 5 1.7 19.8 (30.1) Hallucinations 4 1.3 23.8 (22.6) Paranoia 1 0.3 32.0 (NA) SD: standard deviation; NA: not applicable; CAMHS: Child and Adolescent Mental Health Services. Occurring alongside other difficulties in participants without suspected psychosis. not differ between those with and without Current View rat- Prevalence ings (t = 0.20, df = 35.3, p = 0.84). A total of 275 participants Paranoid thoughts were common in this clinical sample, completed either the RCADS or the SDQ (mean age = 15.1, with item endorsement ranging from 14% to 54% (Table 2). SD = 1.75, girls: n = 171, boys: n = 104, outpatient: n = 250, The mean number of suspicions endorsed was 5.85 inpatient: n = 25). Paranoia was slightly higher in those that (SD = 5.17). Out of the 301 patients, 35% had at least mildly completed either measure (mean = 20.3, SD = 18.5) than elevated paranoia, 15% had at least moderate paranoia, 6% those who did neither (mean = 15.7, SD = 14.7), although this had at least high paranoia, and 3% had severe levels of par- difference was not significant (t = 1.60, df = 41.5, p = 0.12). anoia (Table 3). As shown in Tables 2 and 3, the rates of paranoia were approximately double those previously Clinician-rated problems. Bivariate associations between reported in a general population sample of adolescents. paranoia and the presence of each clinician-rated problem Paranoia in the patient sample was significantly higher in are shown in Table 4. The presence of peer relationship prob- girls than boys (t = 4.08, df = 288.2, p < 0.001), with 41% of lems had the strongest association with paranoia (β = 0.64, girls reporting at least mildly elevated levels compared to p < 0.001) and explained 11% of the variance in paranoia 24% of boys. There was no relationship between age and scores. The second largest association was for self-harm paranoia (r = 0.08, p = 0.16). The 30 inpatients had somewhat (β = 0.55, p < 0.001) which accounted for 7% of the variance higher paranoia scores overall (mean = 27.1, SD = 21.5) than in paranoia. Similar sized medium associations were also the community patients (mean = 19.2, SD = 17.7), although observed for post-traumatic stress symptoms (β = 0.54, this was not significant (t = 1.93, df = 33.5, p = 0.062). p = 0.001) and a history of abuse/neglect (β = 0.50, p = 0.013), although only 4% and 2% of the variance in paranoia was explained by these factors, respectively. It was notable that Clinical associations of the 104 patients with at least elevated paranoia, 38 (37%) had clinician-rated trauma (post-traumatic stress or history The clinician-rated Current View was completed for 272 par- of abuse/neglect). Depression and social anxiety showed ticipants (mean age = 15.0, SD = 1.77, girls: n = 166, boys: small but significant associations with paranoia that each n = 106, outpatient: n = 248, inpatient: n = 24). Paranoia did Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1171 Table 2. B-CAP item endorsement in CAMHS sample (n = 301) and previously reported weekly rates from the general population (n = 801). Non- CAMHS clinical Item 0 1 2 3 4 5 Weekly+ Weekly+ 1. People at school are trying to 135 33 68 37 17 11 44% 25% make me feel unwanted 2. I’m sure people are gossiping 120 39 76 31 12 23 47% 21% about me on social media 3. I am being pushed out of 124 54 63 28 22 10 41% 22% conversations on purpose 4. My friends or partner are 177 49 32 21 12 10 25% 10% ignoring my messages to upset me 5. People are trying to embarrass 185 39 31 24 9 13 26% 20% me in class on purpose 6. People are making sly 118 58 60 36 14 15 42% 16% comments to upset me 7. I think people are lying to me 93 44 74 47 19 24 54% 30% on purpose 8. People say things under their 143 43 48 33 18 16 38% 24% breath to wind me up 9. Nasty tricks are being played 216 32 30 14 1 8 18% 8% on me 10. People are trying to confuse 164 40 48 19 15 15 32% 17% me on purpose 11. Groups of people are planning 197 35 31 17 11 10 23% 10% against me 12. People are collecting my 237 21 23 7 4 9 14% 7% information or photos to use against me 13. I’m sure people are seeking 201 36 30 17 8 9 21% 11% revenge on me 14. I feel like I am being followed 212 23 26 18 9 13 22% 12% or stalked 15. I am scared of what strangers 124 50 45 35 22 25 42% 32% will do to me 16. People will try to kidnap me 193 42 26 22 11 7 22% 14% 17. I could be attacked at any time 132 54 43 21 25 26 38% 23% 18. I feel unsafe around people 149 46 37 20 23 26 35% 19% everywhere I go CAMHS: Child and Adolescent Mental Health Services; B-CAP: Bird Checklist of Adolescent Paranoia. Endorsement rates as reported in Bird et al. (2019). explained 6% of the variance. Small significant associations anxiety, respectively. The presence of ADHD symptoms accounting for only 4% and 2% of the variance in paranoia showed a small negative association that explained 2% of the were observed for educational difficulties and generalised variance in paranoia scores. Australian & New Zealand Journal of Psychiatry, 55(12) 1172 ANZJP Articles Table 3. Mean scores and proportions of CAMHS patients (n = 301) scoring above validated score thresholds compared to previously collected data from the adolescent general population (n = 801; Bird et al., 2019). CAMHS General population All Girls Boys All Girls Boys Mean score (SD) 20.0 (18.2) 23.1 (19.4) 15.0 (14.9) 12.5 (14.0) 15.8 (15.0) 8.2 (10.8) ⩽22 (average range) 197 (65%) 108 (59%) 89 (76%) 667 (83%) 314 (77%) 351 (92%) 23+ (mildly elevated+) 104 (35%) 76 (41%) 28 (24%) 134 (17%) 96 (23%) 31 (8%) 40+ (moderate+) 46 (15%) 34 (18%) 12 (10%) 52 (7%) 40 (10%) 8 (2%) 54+ (high+) 18 (6%) 15 (8%) 3 (3%) 16 (2%) 11 (3%) 3 (0.8%) 71+ (severe+) 10 (3%) 9 (5%) 1 (0.9%) 4 (0.5%) 2 (0.5%) 2 (0.5%) CAMHS: Child and Adolescent Mental Health Services; SD: standard deviation. In the general population sample, 9/801 participants identified as ‘other gender’. These participants were not included in the gender group comparison due to the limited sample size. There was a small-medium association between the pres- (p < 0.05) but not self-harm or post-traumatic stress. None ence of clinician-rated psychosis (hallucinations/delusions) of the other edges with paranoia were significantly different and higher paranoia (β = 0.47, p = 0.061, R = 0.01). This was in size (p > 0.05; supplementary Table S2). A total of 56% not statistically significant, most likely due to limited power of the variance in paranoia was explained by the direct edges with only 17 patients rated as having these symptoms; nota- with these seven variables (see supplementary Table S2 for bly, nine of these (53%) had at least mildly elevated para- predictability values of all nodes). The absence of edges in noia. Small associations that were not significant (p > 0.05) Figure 1(a) shows that paranoia was conditionally inde- and each accounted for only 1% of the variance in paranoia pendent from depression, distress/social impairment, hyper- were observed for substance abuse, emerging personality activity, generalised anxiety, and social anxiety, indicating disorder, separation anxiety, family relationship problems, primarily indirect relationships through other variables in panic, conduct problems and OCD (Table 4). The associa- the network. tions between paranoia and agoraphobia, extremes of mood, The shortest paths from paranoia to all other variables eating problems, and ASD were of a negligible size in Figure 1(b) shows the direct relationship was the domi- (β < 0.20) and non-significant (p > 0.05). nant pathway between paranoia and all seven variables for which an edge was present. The shortest path network then shows that the fastest route from paranoia to distress/social Network analysis. The fully estimated network between para- impairment was via peer difficulties, indicating a mediat- noia, self-report psychological problems and selected clini- ing role of peer difficulties in this relationship. Potential cian-rated symptoms is shown in Figure 1(a) (see supplement mediation pathways are also highlighted from paranoia to for 95% CIs of all edges). Once the contribution of all other hyperactivity via conduct problems, and to depression, variables was controlled, paranoia demonstrated the largest generalised anxiety, and social anxiety via panic. unique relationship with peer difficulties (edge weight = 0.35, Notably, paranoia was the only variable that both self- 95% CI = [0.22, 0.47]). Figure 1(a) shows paranoia also had harm and post-traumatic stress had a unique association with a key role in connecting peer difficulties with the rest of the once all other variables were controlled (Figure 1(a)). The network, with the paths from peer difficulties to four of the normalised accuracy (i.e. predictability) values suggested the anxiety domains, behavioural problems, self-harm, and post- single edge with paranoia accounted for 22% of the remain- traumatic stress all occurring via paranoia. ing accuracy of self-harm beyond what was predicted by the Paranoia also demonstrated direct edges with self-harm intercept model (nCC = 0.22; CC = 0.51; CC = 0.62). (edge weight = 0.17, 95% CI = [−0.05,0.38]), conduct prob- marg total Conversely, the edge with paranoia did not lead to any lems (edge weight = 0.17, 95% CI = [0.02, 0.31]), panic increase in accuracy beyond the intercept model for post- (edge weight = 0.14, 95% CI = [−0.01, 0.28]), post-traumatic traumatic stress (nCC = 0.00; CC = 0.75; CC = 0.75). stress (edge weight = 0.14, 95% CI = [−0.07, 0.36]), obses- marg total sive compulsiveness (edge weight = 0.11, 95% CI = [−0.03, 0.26]), and separation anxiety (edge weight = 0.08, 95% Paranoia persistence CI = [−0.05, 0.21]). The edge with peer difficulties was sig- A total of 105 participants (mean age = 15.1, SD = 1.71, girls: nificantly larger than the edges with conduct problems, n = 75, boys: n = 30) agreed to repeat the questionnaires several panic, obsessive-compulsiveness, and separation anxiety Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1173 Table 4. Associations between paranoia severity and the presence of clinician-rated problems. Problem type Clinician rating Linear regressions Absent Present β 95% CI p R n Mean n Mean Social anxiety 79 14.2 (13.7) 189 22.5 (19.4) 0.45 [0.23, 0.67] <0.001 0.06 Separation anxiety 172 18.5 (17.0) 97 23.1 (20.5) 0.25 [0.00, 0.50] 0.050 0.01 Generalised anxiety 97 16.8 (17.4) 171 21.6 (18.7) 0.26 [0.01, 0.51] 0.042 0.02 OCD 220 20.8 (19.1) 49 17.1 (14.9) –0.20 [–0.51, 0.11] 0.21 0.01 Panic 187 18.7 (18.2) 84 23.0 (18.7) 0.23 [–0.03, 0.49] 0.078 0.01 Agoraphobia 217 19.6 (18.1) 52 21.8 (19.3) 0.12 [–0.18, 0.43] 0.43 0.00 Depression 75 13.9 (13.2) 197 22.4 (19.5) 0.46 [0.25, 0.68] <0.001 0.06 Self-harm 143 15.3 (14.3) 129 25.4 (20.8) 0.55 [0.31, 0.79] <0.001 0.07 Eating problems 222 19.4 (18.6) 50 23.0 (17.3) 0.19 [–0.11, 0.50] 0.21 0.01 Psychosis 254 19.6 (18.1) 17 28.2 (21.6) 0.47 [–0.02, 0.96] 0.061 0.01 Bipolar 246 19.8 (18.6) 26 22.5 (16.4) 0.15 [–0.26, 0.55] 0.48 0.00 PTSD 199 17.7 (16.9) 63 27.5 (20.6) 0.54 [0.22, 0.85] 0.001 0.04 Abuse or neglect 221 18.6 (17.6) 43 27.9 (21.2) 0.50 [0.11, 0.89] 0.013 0.02 Conduct problems 218 19.3 (18.7) 52 23.2 (16.9) 0.22 [–0.09, 0.52] 0.16 0.01 Substance abuse 242 19.4 (18.5) 30 25.4 (16.4) 0.33 [–0.05, 0.71] 0.089 0.01 Emerging PD 208 18.9 (17.8) 62 23.6 (19.5) 0.25 [–0.03, 0.54] 0.080 0.01 Peer difficulties 98 12.5 (13.3) 173 24.2 (19.5) 0.64 [0.42, 0.85] <0.001 0.11 Family difficulties 111 17.6 (18.4) 157 22.0 (18.3) 0.24 [0.00, 0.48] 0.054 0.01 ADHD 196 21.6 (19.2) 74 15.9 (15.4) –0.31 [–0.55, –0.07] 0.010 0.02 ASD 172 19.6 (17.6) 93 21.0 (20.2) 0.08 [–0.18, 0.33] 0.56 0.00 Education problems – – – – 0.22 [0.08, 0.36] 0.002 0.04 β: standardised beta; CI: confidence interval; OCD: obsessive-compulsive disorder; PTSD: post-traumatic stress disorder; PD: personality disorder; ADHD: attention-deficit hyperactivity disorder; ASD: autism spectrum disorder. Mean paranoia scores shown with standard deviations in parentheses for those with and without each problem. Significant results highlighted in bold. Weighted least squares regression used due to heteroscedasticity in residuals. months later (mean = 21.3 weeks, SD = 6.52). The difference in least mildly elevated baseline paranoia, 30 had consistently baseline paranoia between those with follow-up data elevated or increasing scores, 5 showed significant reduc- (mean = 22.6, SD = 19.6) and those without (mean = 18.6, tions that remained in the elevated range, and 11 had sig- SD = 17.3) was small and not significant (t = 1.78, df = 190.9, nificant reductions into the average range. p = 0.077). Linear mixed-effects models showed that, compared to There was no overall difference in paranoia between those with low/transient paranoia (n = 55), across the two baseline (mean = 22.6, SD = 19.6) and follow-up (mean = time points, participants with persistent/increasing para- 23.7, SD = 19.4; V = 2296, p = 0.73, ES = 0.06). On an indi- noia (n = 50) had consistently higher levels of depression vidual basis, however, 18/105 participants had significant (β = 0.81, 95% CI = [0.45, 1.18], p < 0.001), panic (β = 0.75, increases (RCI > 1.96) in paranoia and 16/105 had signifi- 95% CI = [0.38, 1.12], p < 0.001), social anxiety (β = 0.75, cant decreases (RCI < −1.96). Of the 46 participants with at 95% CI = [0.38, 1.11], p < 0.001), generalised anxiety Australian & New Zealand Journal of Psychiatry, 55(12) 1174 ANZJP Articles Figure 1. (a) Network analysis of paranoia and other symptoms. Edges indicate positive associations and rings represent node predictability based on neighbouring nodes. Pink, blue, and orange rings (i.e. continuous variables) indicate R values. For binary (i.e. purple) variables, the shaded rings represent the proportion of correct classification, split into the accuracy of the intercept model (purple section) and the additional contribution of connected nodes (dark blue section). (b) Shortest paths from paranoia to all other variables, with dashed lines representing suppressed edges. (a) (b) Discussion (β = 0.74, 95% CI = [0.38, 1.10], p < 0.001), separation anx- iety (β = 0.64, 95% CI = [0.26, 1.02], p = 0.001), peer diffi- The adolescents attending CAMHS were primarily doing culties (β = 0.63, 95% CI = [0.24, 1.01], p = 0.002), conduct so because they had emotional disorders such as anxiety problems (β = 0.50, 95% CI = [0.11, 0.90], p = 0.014), and depression. This was to be expected. However, para- hyperactivity (β = 0.44, 95% CI = [0.04, 0.84], p = 0.032), noia was common in these young patients, with several sus- and distress/social impairment (β = 0.62, 95% CI = [0.23, picious thoughts occurring in one-third to one-half of the 1.01], p = 0.0026), but not OCD (β = 0.22, 95% CI = [−0.18, clinical cohort. Over half of patients regularly thought peo- 0.63], p = 0.28). ple were lying to them on purpose, over 40% felt scared of There were small paranoia group by time interactions at what strangers would do to them, and 35% felt unsafe eve- the threshold for significance for generalised anxiety rywhere around people. Overall, 35% reported at least (β = 0.38, 95% CI = [0.02, 0.74], p = 0.043) and social anxi- mildly elevated paranoia and 15% reported at least moder- ety (β = 0.34, 95% CI = [0.00, 0.68], p = 0.052), indicating ate paranoia. Rates of paranoia were approximately double those with persistent paranoia had somewhat less improve- those observed in adolescents from the general population ment in these symptoms compared to those with low/tran- sient paranoia. Group by time interactions were negligible (Bird et al., 2019). Previous findings that adolescent girls, and non-significant for all other domains (p > 0.05; sup- compared to boys, may be especially likely to report suspi- plementary Table S4). cious thinking were replicated (Bird et al., 2019; Ronald Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1175 et al., 2014). Although traditionally conceptualised as a to a dangerous world, though this does not mean it is inevita- symptom of psychotic disorders, paranoia in this adoles- ble or that it is without negative consequences. But our find- cent sample primarily occurred alongside common mental ings also show paranoia is certainly not confined to health problems and only a minority had suspected psycho- traumatised youth: the trauma variables only accounted for a sis. Although limited in size, the available follow-up data very small amount of the variance in paranoia and almost indicated that the paranoia was often persistent. Yet para- two-thirds of patients with paranoia did not have a (clinician- noia may well be overlooked: only one participant had the rated) history of trauma. presence of paranoia recorded in their clinical notes. Arguably one of the most important findings from the Paranoid thinking in the adolescent patients was associ- study is a close relationship between paranoia and peer ated with a wide range of clinician-rated problems includ- relationship difficulties. This association was the strongest ing anxiety, depression, trauma, self-harm, peer relationship, of all those assessed from both clinicians and patients, even and educational difficulties. Paranoia in CAMHS patients after controlling for the influence of all other variables in may therefore be expected to present in the context of emo- the network. Although the relationship will undoubtedly be tional problems, adverse life experiences, and impaired bidirectional to a degree, our previous work using a social functioning. It may also be particularly common in Bayesian approach to causal discovery found adolescent young people who self-harm: elevated paranoia was pre- peer difficulties are more likely to be influenced by para- sent in almost half of patients for whom emotion dysregula- noia than vice versa (Bird et al., 2019). This pathway is tion, self-harm, or suicidality was a primary reason for plausible, as the ability to trust is necessary for relation- accessing services. Network analysis also showed that once ships, whereas fear of other people will make it difficult to all other variables were controlled, the presence of self- socialise and make friends. We also found the most com- harm was solely associated with paranoia, with this edge mon pathway from emotional and behavioural problems to contributing to 22% of the predictability of self-harm peer difficulties occurred via paranoia, suggesting paranoia (beyond the intercept model). This relationship is consist- may be a common route to impairments in adolescent peer ent with findings from the adult literature (Freeman et al., relationships. At an age when peer acceptance is most 2019b) and evidence that self-harm is associated with psy- highly valued (Somerville, 2013), the potential impact on chotic experiences in general in adolescents (Hielscher friendships is likely to be a substantial cause of distress for et al., 2019; Martin et al., 2015). The co-occurrence of para- young people. In line with this, peer difficulties were the noia with so many different psychiatric symptoms could mediating link connecting paranoia and the overall distress also be an indicator of more severe presentations, with ado- and functional impact of young people’s problems. lescents who report persistent paranoia having greater lev- els of symptoms and social impairments over time. Limitations Consistent with a cognitive conceptualisation of paranoia as an unfounded threat belief (Freeman, 2016), network The study has several limitations. First, the sample was not analyses showed paranoia had unique associations with anxi- a fully representative cohort. It was not possible to invite all ety symptoms, especially panic. The network analysis fur- patients accessing participating services to take part, since ther demonstrated a relationship between paranoia and services could not be covered by the research team all the post-traumatic stress symptoms. Once all other variables time. However, attempts were made to minimise sampling were controlled, the presence of post-traumatic stress symp- bias by inviting patients to take part regardless of their rea- toms was solely related to paranoia. This relationship is con- son for accessing services or clinical diagnosis. The cohort sistent with evidence that negative interpersonal experiences also included a higher proportion of girls than boys, contribute to the development of paranoia (Freeman et al., although this may be representative of CAMHS given the 2013; Shevlin et al., 2015). It is important to emphasise, higher rates of common mental health problems in adoles- however, that justified fears of harm in relation to ongoing cent girls (NHS Digital, 2018). Nevertheless, the pattern of bullying or abuse is not paranoia (a term that only applies to associations between paranoia and other variables could be unfounded ideas). Paranoia in those with adversity occurs influenced by gender, and, as a result, the network structure when their concerns generalise excessively beyond specific may have biased understanding towards girls. However, experiences to the point they become clearly unfounded (e.g. there is a lack of clear evidence showing the relationships when an individual with past bullying develops a persistent between paranoia and causal factors differ by gender. concern that people are conspiring to humiliate them and Another notable source of sampling bias was the primarily interprets friendliness from others as a trick). Although sev- affluent catchment areas for the services included with a eral mechanisms driving this generalisation are likely, one local demographic of mostly White British individuals. As proposal is that negative experiences lead to learned beliefs experiences such as racism and child adversity are likely to about other people (i.e. as threatening) and the self (i.e. as contribute to the development of paranoia (Bentall et al., vulnerable) upon which paranoia flourishes (Freeman, 2012; Shaikh et al., 2016), clinical levels of paranoia in 2016). Paranoia can be an understandable protective response youth may differ by locality. Australian & New Zealand Journal of Psychiatry, 55(12) 1176 ANZJP Articles A strength of this study was the ability to compare the C.S., A.-L.T., L.C., H.J.S. and A.C.J. contributed to data collec- tion and management. D.F. and F.W. supervised the work and con- prevalence of paranoia in CAMHS patients with a repre- tributed to the design, theoretical interpretation, and writing. All sentative general population sample of adolescents using authors contributed to the final version of the manuscript prior to the same measure. This was not a perfect comparison, how- submission. ever, as the general population sample were slightly younger than the patients in this study. But as age was not Declaration of Conflicting Interests associated with paranoia in either sample, the effect of this The author(s) declared no potential conflicts of interest with difference on the comparison is likely to be minimal. respect to the research, authorship, and/or publication of this Another limitation was that aside from the B-CAP, our article. other measures were missing for approximately 10% of participants. This reflected the reality of routine measure- Funding ment in CAMHS where clinical pressures could prevent clinicians from completing the Current View and patients The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The sometimes left before completing all questionnaires. study was funded by an NIHR Research Professorship awarded to Notably, as follow-up questionnaires were collected as an D.F. (NIHR-RP-2014-05-003). The work was also supported by optional second stage of the study, only a third of the sam- the NIHR Oxford Health Biomedical Research Centre (BRC- ple provided longitudinal data. Planned prospective studies 1215-20005). This paper presents independent research funded by examining paranoia in representative clinical samples will the National Institute for Health Research (NIHR). The views be needed to understand fully the relationship over time expressed are those of the authors and not necessarily those of the with other mental health problems. NHS, the NIHR, or the Department of Health. It must also be acknowledged that a degree of measure- ment error is likely in self-report measures of paranoia. It is ORCID iD not possible in self-report questionnaires to determine if Jessica C Bird https://orcid.org/0000-0001-9457-1506 concerns about intended harm from others are unfounded. However, the B-CAP has shown good construct validity as Supplemental Material a measure of unfounded ideation, with evidence that scores Supplemental material for this article is available online. are distinct from bullying and are associated with adoles- cents’ ratings that their fears of others are excessive (Bird et al., 2019, 2020). Evidence also shows that, in general, References self-report paranoia questionnaires predict genuine para- Bates D, Mächler M, Bolker B, et al. (2015). Fitting linear mixed-effects noid ideation in controlled virtual reality settings (Freeman models using lme4. Journal of Statistical Software, 67(1), 1–48. Bentall RP, Wickham S, Shevlin M, et al. (2012) Do specific early-life et al., 2010, 2014). Nevertheless, clinical interview valida- adversities lead to specific symptoms of psychosis? A study from the tion of paranoia in young people accessing CAMHS would 2007 the Adult Psychiatric Morbidity Survey. Schizophrenia Bulletin be beneficial. 38: 734–740. Overall, this study highlights paranoia as a common, Bird JC, Evans R, Waite F, et al. (2019) Adolescent paranoia: Prevalence, potentially clinically important, and overlooked problem in structure, and causal mechanisms. Schizophrenia Bulletin 45: 1134–1142. young people who are accessing mental health services. Bird JC, Loe BS, Kirkham M, et al. (2020) The assessment of paranoia Greater awareness by clinicians of paranoia in patients in young people: Item and test properties of the Bird Checklist of attending CAMHS may be required. The use of validated Adolescent Paranoia. Schizophrenia Research 220: 116–122. tools such as the B-CAP may be helpful for clinicians to Borsboom D and Cramer AOJ (2013) Network analysis: An integrative identify paranoia within young people’s broader clinical approach to the structure of psychopathology. Annual Review of Clinical Psychology 9: 91–121. presentation and monitor change. Targeted intervention for Chorpita BF, Yim L, Moffitt C, et al. (2000) Assessment of symptoms of paranoia, suitably adapted for this age group, may then be DSM-IV anxiety and depression in children: A revised child anxiety appropriate to help young people feel safer in their daily and depression scale. Behaviour Research and Therapy 38: 835–855. lives. D’Agostino A, Rossi Monti M and Starcevic V (2019) Psychotic symp- toms in borderline personality disorder: An update. Current Opinion in Psychiatry 32: 22–26. Acknowledgements Dijkstra EW (1959) A note on two problems in connexion with graphs. We would like to thank all staff members at the South Oxfordshire Numerische Mathematik 1: 269–271. Child and Adolescent Mental Health Service and the Highfield Epskamp S and Fried EI (2018) A tutorial on regularized partial correla- Adolescent Unit, Oxford for supporting this research, and all the tion networks. Psychological Methods 23: 617–634. young people who gave their valuable time to take part. Epskamp S, Borsboom D and Fried EI (2018) Estimating psychological networks and their accuracy: A tutorial paper. Behavior Research Methods 50: 195–212. Author Contributions Epskamp S, Cramer AOJ, Waldorp LJ, et al. (2012) qgraph: Network visu- J.C.B. led the research design, data collection, data management, alizations of relationships in psychometric data. Journal of Statistical statistical analysis, and manuscript preparation. E.C.F., M.K., Software 48: 11–18. Australian & New Zealand Journal of Psychiatry, 55(12) Bird et al. 1177 Ford T, Macdiarmid F, Russell AE, et al. (2017) The predictors of per- Kelleher I, Keeley H, Corcoran P, et al. (2012) Clinicopathological sig- sistent DSM-IV disorders in 3-year follow-ups of the British Child nificance of psychotic experiences in non-psychotic young people: and Adolescent Mental Health Surveys 1999 and 2004. Psychological Evidence from four population-based studies. The British Journal of Medicine 47: 1126–1137. Psychiatry 201: 26–32. Foygel R and Drton M (2010) Extended Bayesian information criteria Kelleher I, Ramsay H and DeVylder J (2017) Psychotic experiences and for Gaussian graphical models. Advances in Neural Information suicide attempt risk in common mental disorders and borderline per- Processing Systems 23: 604–612. sonality disorder. Acta Psychiatrica Scandinavica 135: 212–218. Freeman D (2016) Persecutory delusions: A cognitive perspective on McGorry PD, Hartmann JA, Spooner R, et al. (2018) Beyond the ‘at risk understanding and treatment. The Lancet Psychiatry 3: 685–692. mental state’ concept: Transitioning to transdiagnostic psychiatry. Freeman D, Antley A, Ehlers A, et al. (2014) The use of immersive vir- World Psychiatry 17: 133–142. tual reality (VR) to predict the occurrence 6 months later of paranoid Martin G, Thomas H, Andrews T, et al. (2015) Psychotic experiences thinking and posttraumatic stress symptoms assessed by self-report and psychological distress predict contemporaneous and future non- and interviewer methods: A study of individuals who have been phys- suicidal self-injury and suicide attempts in a sample of Australian ically assaulted. Psychological Assessment 26: 841–847. school-based adolescents. Psychological Medicine 45: 429–437. Freeman D, Loe BS, Kingdon D, et al. (2019a) The revised Green et al., NHS Digital (2018) Mental Health of Children and Young People in Paranoid Thoughts Scale (R-GPTS): Psychometric properties, sever- England 2017. London: NHS Digital. ity ranges, and clinical cut-offs. Psychological Medicine. Epub ahead Orben A, Tomova L and Blakemore S-J (2020) The effects of social dep- of print 20 November. DOI: 10.1017/S0033291719003155. rivation on adolescent development and mental health. The Lancet Freeman D, Bold E, Chadwick E, et al. (2019b) Suicidal ideation and behav- Child & Adolescent Health 4: 634–640. iour in patients with persecutory delusions: Prevalence, symptom associa- Peralta V and Cuesta MJ (1999) Dimensional structure of psychotic symp- tions, and psychological correlates. Comprehensive Psychiatry 93: 41–47. toms: An item-level analysis of SAPS and SANS symptoms in psy- Freeman D, Pugh K, Vorontsova N, et al. (2010) Testing the continuum chotic disorders. Schizophrenia Research 38: 13–26. of delusional beliefs: An experimental study using virtual reality. R Core Team (2013) R: A Language and Environment for Statistical Journal of Abnormal Psychology 119: 83–92. Computing. R Foundation for Statistical Computing. Available at: Freeman D, Thompson C, Vorontsova N, et al. (2013) Paranoia and www.R-project.org/2013 (accessed 9 March 2020). post-traumatic stress disorder in the months after a physical assault: Romano JP and Wolf M (2017) Resurrecting weighted least squares. A longitudinal study examining shared and differential predictors. Journal of Econometrics 197: 1–19. Psychological Medicine 43: 2673–2684. Ronald A, Sieradzka D, Cardno AG, et al. (2014) Characterization of Fruchterman TMJ and Reingold EM (1991) Graph drawing by force- psychotic experiences in adolescence using the Specific Psychotic directed placement. Software: Practice and Experience 21: 1129–1164. Experiences Questionnaire: Findings from a study of 5000 16-year- Garety PA, Gittins M, Jolley S, et al. (2013) Differences in cognitive and old twins. Schizophrenia Bulletin 40: 868–877. emotional processes between persecutory and grandiose delusions. Schmidt AF and Finan C (2018) Linear regression and the normality Schizophrenia Bulletin 39: 629–639. assumption. Journal of Clinical Epidemiology 98: 146–151. Goodman R (1999) The extended version of the Strengths and Difficulties Shaikh M, Ellett L, Dutt A, et al. (2016) Perceived ethnic discrimination Questionnaire as a guide to child psychiatric caseness and consequent and persecutory paranoia in individuals at ultra-high risk for psycho- burden. Journal of Child Psychology and Psychiatry 40: 791–799. sis. Psychiatry Research 241: 309–314. Goodman R (2001) Psychometric properties of the Strengths and Shevlin M, McAnee G, Bentall RP, et al. (2015) Specificity of association Difficulties Questionnaire. Journal of the American Academy of Child between adversities and the occurrence and co-occurrence paranoia & Adolescent Psychiatry 40: 1337–1345. and hallucinations: Evaluating the stability of childhood risk in an Haslbeck JMB and Waldorp LJ (2018) How well do network models adverse adult environment. Psychosis 7: 206–216. predict observations? On the importance of predictability in network Somerville LH (2013) The teenage brain: Sensitivity to social evaluation. models. Behavior Research Methods 50: 853–861. Current Directions in Psychological Science 22: 121–127. Haslbeck JMB and Waldorp LJ (2020) mgm: Estimating time-vary- Taylor MJ, Gregory AM, Freeman D, et al. (2015) Do sleep disturbances ing mixed graphical models in high-dimensional data. Journal of and psychotic-like experiences in adolescence share genetic and Statistical Software 93: 1–46. environmental influences? Journal of Abnormal Psychology 124: Hermans K, van der Steen Y, Kasanova Z, et al. (2020) Temporal dynam- 674–684. ics of suspiciousness and hallucinations in clinical high risk and first Tibshirani R (1996) Regression shrinkage and selection via the Lasso. episode psychosis. Psychiatry Research 290: 113039. Journal of the Royal Statistical Society: Series B (Methodological) Hielscher E, Connell M, Lawrence D, et al. (2019) Association between 58: 267–288. psychotic experiences and non-accidental self-injury: Results from a Wigman JTW, Vollebergh WAM, Raaijmakers QAW, et al. (2011) nationally representative survey of adolescents. Social Psychiatry and The structure of the extended psychosis phenotype in early ado- Psychiatric Epidemiology 54: 321–330. lescence: A cross-sample replication. Schizophrenia Bulletin 37: Jacobson NS and Truax P (1991) Clinical significance: A statistical 850–860. approach to defining meaningful change in psychotherapy research. Zavos HMS, Freeman D, Haworth CMA, et al. (2014) Consistent etiology Journal of Consulting and Clinical Psychology 59: 12–19. of severe, frequent psychotic experiences and milder, less frequent Jones M, Hopkins K, Kyrke-Smithm R, et al. (2013) Current View Tool manifestations: A twin study of specific psychotic experiences in ado- Completion Guide. London: CAMHS Press. lescence. JAMA Psychiatry 71: 1049–1057. Australian & New Zealand Journal of Psychiatry, 55(12)

Journal

Australian & New Zealand Journal of Psychiatry – SAGE

Published: Jan 10, 2021

Keywords: Youth mental health; psychotic experiences; delusions; emotional disorders; network analysis

There are no references for this article.