The association between attachment style and social functioning in patients with non-affective psychotic disorders, unaffected siblings and healthy controls

Objective: In patients with psychotic disorders, insecure attachment styles and impaired social functioning are highly prevalent. Our aim was to explore the multi-cross-sectional and longitudinal associations between attachment style and social functioning. Methods: This study was performed in a subsample of 119 patients with non-affective psychotic disorders, 128 unaffected siblings and 66 healthy controls within the Genetic Risk and Outcome of Psychosis (GROUP) Study. Attachment style (using the Psychosis Attachment Measure) and three social functioning domains, namely withdrawal, interpersonal behaviour and pro-social activities as measured with the Social Functioning Scale (SFS) were assessed on two moments in time. Generalized linear mixed models and linear regression models were used. Bonferroni correction for multiple testing was applied. Results: In the patient group, a significant negative association was found between avoidant attachment and pro-social functioning. In the sibling and control group, we found significant negative associations between avoidant attachment and the social functioning domains withdrawal and interpersonal behaviour. We also found a significant negative association between anxious attachment and the social functioning domain withdrawal in siblings. Higher levels of insecure anxious attachment at baseline predicted a reduction of the social functioning domain interpersonal behaviour in siblings over a period of three years on a trend level. Conclusions: Findings indicate that levels of insecure attachment are elevated in patients with psychotic disorders and have a negative association with social functioning in both patients, siblings and controls. These findings warrant specific attention for attachment style in the treatment of patients with psychotic disorders and individuals with an increased vulnerability for developing psychosis.


Introduction
Impairment in social functioning is an important feature of psychotic disorders. A substantial proportion of patients with a psychotic disorder experiences impairments in the ability to work or study, to establish social relationships or to participate in recreational and community activities (Pearse et al., 2020). Moreover, impaired social functioning is often already present in individuals at risk for psychosis (Velthorst et al., 2017;Hans et al., 2000;Ballon et al., 2007), and in first-degree relatives of individuals with psychotic disorders (Mondragón-Maya et al., 2017). The impairments in social functioning are related to adverse outcomes (Bellido-Zanin et al., 2015;Degnan et al., 2018) and contribute to a reduced quality of life (Mow et al., 2020). Therefore, social functioning has become an important outcome measure in the treatment of psychotic disorders and is recognized as an important determinant of treatment success (Burns and Patrick, 2007). Greater insight into the underlying factors leading to impairments in social functioning, can enhance our understanding of psychotic disorders and can contribute to more targeted interventions for patients with a psychotic disorder.
A concept relevant to social functioning is attachment style. According to the attachment theory, the pattern of interaction between children and their primary caregivers leads to the development of internal working models, which include mental representations about self and others. These representations function as guidance for thoughts, feelings and behaviour in social interactions into adulthood (Bowlby, 1996). Attachment style is typically classified as secure or insecure. Insecure attachment has been conceptualized as two underlying dimensions: anxious attachment and avoidant attachment (Brennan, 1998). Anxious attachment is characterized by a continuous desire for affirmation and a fear of rejection, making physical or emotional distance difficult to tolerate. Avoidant attachment is characterized by avoidance of intimacy and a continuous desire for control and autonomy, making physical or emotional closeness difficult to tolerate.
A meta-analysis of the relationship between attachment style and psychosis showed that the prevalence of insecure attachment styles was higher in individuals with psychosis than in non-clinical samples (76 and 38 %, respectively) (Carr et al., 2018). Given the higher prevalence of impaired social functioning (Pearse et al., 2020;Velthorst et al., 2017;Ballon et al., 2007;Hans et al., 2000), as well as the higher prevalence of insecure attachment styles in patients with a psychotic disorder compared to non-clinical samples (Carr et al., 2018;van Dam et al., 2014;Korver-Nieberg et al., 2013), the relationship between attachment style and social functioning in patients with psychotic disorders is a potentially important area of research. However, to date it is seldom investigated (Pearse et al., 2020). The majority of studies have explored the association between attachment style and interpersonal relationships, such as the therapeutic relationship (Berry, 2020). Overall findings indicate that specifically avoidant attachment is associated with more difficulties in the therapeutic relationship (McGonagle et al., 2021). A few studies focused on the association between attachment and (social) functioning (Berry et al., 2015;Couture et al., 2007;Hansen et al., 2013;Olbert et al., 2016;Palmier-Claus et al., 2016;Tyrrell et al., 1999), they show little consistency or overlap. Taken together, there is some evidence to suggest that secure attachment is associated with better functioning, and insecure attachment is associated with impairments in functioning (Berry et al., 2015;Couture et al., 2007;Hansen et al., 2013;Olbert et al., 2016;Palmier-Claus et al., 2016;Pearse et al., 2020;Tyrrell et al., 1999). Inconsistencies between the results may be due to heterogeneity regarding the included population, sample size or measurement of attachment style or (social) functioning.
To the best of our knowledge, this is the first study investigating the relationship between attachment style and three social functioning domains, namely withdrawal, interpersonal behaviour and pro-social activities in a sample of patients with a psychotic disorder, unaffected siblings and non-clinical controls across two different assessments times (multi-cross-sectional). As social functioning deficits in patients with psychosis are highly prevalent and clinically relevant, understanding the potential predictive value of attachment style would provide important knowledge. Moreover, investigating these associations across different vulnerability groups, by including healthy controls and un-affected siblings has not been done before. Therefore, we believe it is very relevant to study the association between attachment style and social functioning.
Our primary aim was to evaluate whether attachment style was associated with social functioning across two different assessment times. Furthermore, we aimed to explore the association between baseline adult attachment style and change in social functioning over time. We hypothesized that a more insecure attachment style (higher baseline scores on avoidant or anxious attachment) is associated with worse social functioning over time. Moreover, we expected that attachment style is associated with change in social functioning in its trajectory over time, with participants with higher levels of avoidant and/or anxious attachment showing less long-term improvement in social functioning over time. Ultimately, a clearer understanding of the role of attachment style in social functioning may be informative for more targeted interventions for patients with a psychotic disorder and/or those at-risk for psychosis.

Study design and participants
Patients in the present study were included at the Amsterdam site of the Genetic Risk and Outcome of Psychosis (GROUP) study. GROUP is a Dutch multi-site naturalistic follow-up study (after three and six years) designed to study risk and protective factors influencing the onset and course of psychotic disorders in patients, their unaffected family members, and non-related controls. The full procedure of the study has been described elsewhere (Korver et al., 2012). The present study consists of a subsample of 415 participants (161 patients, 181 siblings and 73 controls) of whom complete datasets for attachment and social functioning were available at least one assessment time (see Supplement 4). All participants were invited to take part in a follow up measurement. The study was approved by the Medical Ethics Committee of the Academic Medical Centre of Utrecht. Written informed consent was obtained from participants before they were enrolled in the study.

Social functioning
Social functioning was measured with the Social Functioning Scale (SFS) (Birchwood et al., 1990). This 24 item self-report questionnaire consists of seven different domains, with higher scores reflecting higher levels of social functioning. As we aimed to investigate social functioning (rather than general functioning), we only wanted to include SFS domains that reflect social functioning. In line with Schneider and colleagues (2017), we omitted the SFS subscales independence-competence, independence-performance and employment/ occupation and also recreation as this domain assesses engagement in hobbies and activities and is not specifically linked to interaction to others. In the current study, we used three out of the seven social functioning domains, namely withdrawal, interpersonal behaviour and pro-social activities.

Attachment style
Adult attachment style was measured with the Psychosis Attachment Measure (PAM) (Berry, 2006;Korver-Nieberg, 2014). The PAM is a 15item questionnaire with seven items reflecting avoidant attachment and eight reflecting anxious attachment. Average item scores were calculated for avoidant and anxious attachment, with higher scores reflecting higher levels of insecure attachment (Berry, 2006;Berry et al., 2008).

Covariates
To correct for possible confounding effects, the covariates age, gender, years of education as a proxy for socioeconomic status and severity of psychopathology were selected based on previous reported associations with attachment style or social functioning (Pearse et al., 2020;van Dam et al., 2014;Velthorst et al., 2017;Carr et al., 2018). Sociodemographic data was evaluated using a self-reported questionnaire, specifically developed for the GROUP-study. The presence and severity of psychotic symptoms was measured with the Community Assessment of Psychotic Experiences in all participants (CAPE) (Konings et al., 2006). This 42-item self-report questionnaire measures the lifetime prevalence of positive, negative and depressive symptoms. A mean total score was calculated for the different subscales.

Statistical analyses
All analyses were performed using Statistical Package for the Social Sciences (IBM SPSS Statistics) version 26.0. Normality was checked for all numerical variables visually with probability plots and scatterplots of standardized residuals. Differences between groups regarding baseline demographic and clinical characteristics were compared by one-way (welch) ANOVA, Chi 2 test or Kruskal-Wallis tests, for normal, categorical and non-normal distributed data, respectively. Kruskal-Wallis tests were followed by Games-Howell and Mann-Whitney U post-hoc tests where appropriate. Moreover, correlations between attachment style, social functioning and psychopathology were assessed using Spearman's correlation (Supplementary Table 1).
To investigate potential between-group differences in the association between attachment style and social functioning we tested attachment style x status group interaction effects using generalized liner mixed models (GLMM) in the overall sample for the three social functioning outcome domains. Because these models showed significant differences between status groups (see Supplement 2), we subsequently fitted models investigating the association between attachment style and social functioning domains for each group separately.
To examine the association between attachment style and social functioning at baseline and three years later (multi-cross-sectional), the three social functioning outcome domains were subsequently added as dependent variables and the attachment domains, time and covariates (age and gender) were added as fixed effects with separate models for both attachment domains. We also added intercepts for subjects and random slopes for time as random effects as repeated assessments were nested within subjects. In a second set of models, the covariates education and severity of psychopathology were added. Models were computed with the unstructured covariance setting. Since three different outcome variables were tested (three domains of SFS), we used a Bonferroni correction to minimize the risk of type I errors. Thus, the twotailed significance threshold was set at 0.017 (0.05/3). All models were fitted using restricted maximum likelihood (REML) estimation, allowing for the use of all available data under the assumption that data is missing at random. We tried to make this tenable by comparing individuals lost to follow up to those providing complete data sets on social functioning outcome measures.
To examine the prospective long-term associations between baseline attachment style and change in social functioning, we calculated change scores for all three social functioning domains over the 3-year follow-up period. Given the previously demonstrated stability of attachment style over time, as conducted by our group, we used baseline scores for both attachment domains (Korver-Nieberg, 2014). Subsequently, we ran multiple linear regression models with these change scores as outcome and age, gender, education and severity of psychopathology added as covariates. Again, we used separate models for both attachment domains.

Study sample demographics and characteristics
At baseline, data on PAM and SFS were available for 119 patients, 128 siblings and 66 controls. Differences in demographic data between status groups at baseline are presented in Table 1. Baseline characteristics of attachment style and social functioning are listed in Table 2. Correlations between attachment style, social functioning and psychopathology at baseline are shown in Supplementary Table 1. Our results indicate that negative and depressive symptomatology was more strongly correlated to insecure adult attachment style than positive symptomatology across all groups.
Three years later, 64.5 % of the participants were lost to follow-up. Therefore, the sample size for the longitudinal analyses with complete datasets on PAM and SFS at both assessments time was reduced to 45 patients, 41 siblings and 25 controls. We compared participants included in these longitudinal analyses with those lost to follow up. Analyses yielded that patients and siblings lost to follow up did not differ in baseline attachment style, social functioning, age, gender, education, illness duration, severity of psychopathology compared to patients and siblings with complete social functioning data. Patients were also comparable regarding use of antipsychotic medication and DSM diagnosis. Controls lost to follow up were more often male (n = 35 vs n = 12, p = 0.025) younger (M = 27.90, SD = 7.87, p = 0.002) and higher educated (M = 18.63, SD = 2.30, p = 0.015) at baseline than those with complete data (age: M = 36.80, SD = 12.22; education M = 17.04, SD = 2.75). There were no differences in baseline attachment style, social functioning or severity of psychopathology.

Multi-cross-sectional associations between attachment style and social functioning
Multi-cross-sectional associations between the attachment domains and the three social functioning domains were first assessed by fitting generalized linear mixed models, while controlling for age and gender. In patients, we found significant negative associations between both attachment domains and all three social functioning domains, except the association between anxious attachment and pro-social functioning which did not reach significance at the 0.017 level (Supplementary Table 3). In siblings, significant negative associations between both attachment domains and the social functioning domains withdrawal and interpersonal behaviour were found. In controls, analyses yielded significant negative associations between both attachment domains and the social functioning domain withdrawal. After adding additional covariates (education and severity of psychopathology), the negative association between avoidant attachment and pro-social activities remained significant in the patient group (estimate − 5.757, SE 1.738, p 0.001) ( Table 3). In siblings and controls, all of the previous found significant associations remained significant. In siblings, both avoidant and anxious attachment were associated with withdrawal (estimate − 7.731, SE 1.527, p < 0.001 and estimate − 9.933, SE 2.104, p < 0.001, respectively) and interpersonal behaviour (estimate − 7.257, SE 2.014, p < 0.001 and estimate − 9.267, SE 2.785, p < 0.001, respectively). In controls, both avoidant and anxious attachment were associated with withdrawal (estimate − 9.078, SE 3.043, p 0.004 and estimate − 10.624, SE 4.162, p 0.012, respectively). No other significant associations were found.

Associations between attachment style and change in social functioning
The predictive value of attachment style on the course of the social functioning domains was investigated through multiple regression analyses in a subsample of 45 patients with psychotic disorders, 41 unaffected siblings and 25 healthy controls, while controlling for all covariates. In siblings, higher scores on anxious attachment predicted a reduction in interpersonal behaviour over time on trend level (estimate − 18.034, SE 7.267, p 0.018) ( Table 4). In patients and controls, attachment style at baseline did not significantly predict change in social functioning domains over time.

Summary of findings
The present study aimed to explore the multi-cross-sectional and prospective association between attachment style and three social functioning domains, namely withdrawal, interpersonal behaviour and pro-social activities in a sample of patients with a non-affective psychotic disorder, unaffected siblings and healthy controls over a period of three years. Our main finding is that in patients, siblings and controls, attachment style is negatively associated with social functioning, although the social functioning domains involved differ between groups. In patients higher scores on avoidant attachment were associated with lower scores on all three social functioning domains and notably, the association between avoidant attachment and pro-social functioning remained after correcting for level of education and severity of psychopathology. In siblings and controls, both attachment domains were associated with withdrawal, whereas in siblings they were also associated with interpersonal behaviour.
In line with our second hypothesis, attachment style predicted change in social functioning over time in siblings on a trend level.

Our findings compared to previous research findings
We confirm the higher prevalence of insecure attachment in patients with psychotic disorders compared to siblings and controls, in accordance with previous studies (van Dam et al., 2014;Korver-Nieberg et al., 2013;Carr et al., 2018).
In line with our main finding, three other cross-sectional studies found associations between attachment style and social functioning, although these studies used other measurements (Olbert et al., 2016;Palmier-Claus et al., 2016;Couture et al., 2007). Olbert et al. (2016) found associations between attachment style and community functioning ratings in patients with schizophrenia. Palmier and colleagues (2016) studied a combined sample of patients with schizophrenia, first episode psychosis and at ultrahigh risk for psychosis and found a significant negative effect of anxious attachment on social functioning. In patients with first episode psychosis, Couture et al. (2007) reported associations between avoidant attachment and social and independent living skills. Our study validates these previous results and contributes new insight regarding associations across different vulnerability groups by including unaffected siblings and healthy controls and correcting for relevant confounders.
Partly in contrast to our results, two cross-sectional studies (Tyrrell et al., 1999;Hansen et al., 2013) did not find an association between attachment and (social) functioning. However, both studies used different predictor or outcome measurements and included a more heterogeneous sample consisting of patients with psychotic disorders or bipolar disorders.
To the best of our knowledge no previous studies evaluated the longitudinal association between attachment style and social functioning over time. One previous study evaluated the longitudinal association between attachment style (PAM) and global functioning (GAF) over time (Berry et al., 2015) in a sample of patients with psychotic disorders and comorbid substance disorders and showed that avoidant attachment was an independent predictor for GAF scores at one and two years follow up when corrected for baseline GAF score and alliance of the therapist and patient. In that study no correction for severity of psychopathology or substance use was applied. In line, in the current study, we found attachment style of siblings to be a predictor of change in the social functioning domain interpersonal behaviour over time on a trend level when correcting for covariates.

Interpretations
As mentioned, we found multi-cross-sectional associations between attachment style and social functioning across all groups and a longitudinal association in siblings across three year follow-up. This suggests that attachment style has a significant bearing on social functioning, independent of illness related factors. This is of interest in patients in whom an insecure attachment style is more prevalent and social contact might already be limited by psychotic symptoms (Pearse et al., 2020) and in siblings, who share an increased risk for psychosis. Previously, impairment in premorbid social functioning has been described an important prognostic factor of onset and course of psychotic disorders (Bailer et al., 1996;Cechnicki et al., 2020). Extending these findings, the current study underlines the importance of addressing impairments in social functioning and its association with attachment style before the onset and during the course of psychotic disorders. Hence, early targeted interventions aimed at improving attachment style to prevent social impairment might be particularly important in insecure attached individuals at risk for psychosis (Alvarez-Jimenez et al., 2018). According to a systematic review, there is evidence to suggest that psychotherapy can improve adult attachment security in non-psychotic groups (Taylor et al., 2015). These improvements seemed to be achieved across different therapy settings, different patient groups, and different therapy forms (Taylor et al., 2015). These findings need to be confirmed in future studies in patients with psychotic disorders or those at-risk for psychosis. As studies developing evidence-based interventions are currently missing, for now, clinicians may consider to focus on strengthening the therapeutic alliance and thereby possibly increasing secure attachment representations as a first step .
A possible explanation for the absence of a significant association between attachment style and change in social functioning, might be the relatively small sample size of our subsample, attachment style might not have enough power to bring about change in social functioning. An alternative explanation in patients might be an overarching effect of psychotic symptoms on change in social functioning. It is an interesting suggestion for future studies to study the mechanisms between attachment style, psychopathology and social functioning.
Our findings indicate that an avoidant attachment style is more strongly associated with social functioning than an anxious attachment style. Intuitively, this can be explained by its very definition, as people with an avoidant attachment style tend to avoid intimacy and physical or emotional closeness (Bowlby, 1996). Although, anxious attachment styles also lead to ruptures in social contact given the overdependence on others and hypersensitivity to rejection of people with an anxious attachment style, our results indicate anxious attachment is a less robust predictor of social functioning.
We can only speculate about the reasons for the different combinations of associations between attachment domains and social functioning domains. Other unmeasured factors related to attachment style may also play a role in determining social functioning, such as sensitivity towards rejection, coping style or emotion recognition (Pena-Garijo and Monfort-Escrig, 2021; Korver-Nieberg et al., 2014;Berry et al., 2007). These processes are likely to display a negative influence on social functioning and the development or maintenance of psychotic symptoms (Pena-Garijo and Monfort-Escrig, 2021; Korver-Nieberg et al., 2014;Berry et al., 2007). Interventions focused on these factors by e.g. improving communication and stress coping skills could be valuable across the different groups in preventing deterioration as well as improving deficits. It could also be argued that the differential findings across groups are due to the preponderance of males in the patient group (van Dam et al., 2014), however gender did not seem to have a significant effect on social functioning in all models. Unfortunately, the sample is too small to enable separate analyses by gender. Future studies are required to validate our findings and elucidate the interplay of other mediating or moderating factors in order to enhance our understanding how attachment styles are associated with specific social functioning domains.

Strengths and limitations
The current study extends the findings of previous studies in several ways. We were the first to compare patients with psychotic disorders, unaffected siblings and healthy controls in a multi-cross-sectional and prospective longitudinal design. We controlled for several relevant a priori selected confounders namely age, gender, education and severity of psychopathology. Moreover, we investigated specific associations with social functioning by selecting three subdomains of social functioning instead of a social functioning total score. As the significant associations differ across the subdomains, the current study underlines the importance of distinguishing different social functioning domains in future studies.
In interpreting our findings, some limitations need to be taken into consideration. First, self-report questionnaires have been used to assess attachment style and social functioning. Therefore, the reported perception of participants about attachment and social functioning might not overlap entirely with the actual measured constructs, which introduces the possibility of self-report biases (Olbert et al., 2016). Second, in general the naturalistic design of the study does not allow for any conclusions regarding causation. Further longitudinal and intervention studies are needed as it is possible that the association we found is due to reversed causation (impaired social functioning may generate a more insecure attachment style). Third, in our prospective longitudinal subgroup analyses the sample size of the separate groups (in particular the control group) was relatively small, but sufficient, for conducting mixed model analyses. Fourth, it is widely assumed that attachment style is relatively stable over time (Waters et al., 2000;Chris Fraley, 2002), therefore we presume that the adult attachment style as measured with the PAM is a moderate reflection of early attachment style. Future longitudinal studies are warranted to test whether adult attachment as measured with the PAM is indeed a reflection of early attachment as well. Fifth, our findings indicate that our clinical sample consists of a rather high functioning group of patients. In future studies, it is of interest to examine the association between attachment and social functioning also in a more heterogeneous group with respect to functioning. Finally, a note of caution regarding the measurements is due to a possible overlap between items of the PAM, CAPE and SFS. Nevertheless, most items do not resemble each other as closely and correlations of PAM and SFS were modest (see Supplementary Table 1) and as such we consider it unlikely that current results can be explained away by item overlap.
In conclusion, we showed that attachment style is negatively associated with social functioning. Given the associations across all groups, attachment style may be an important factor contributing to social functioning across the vulnerability spectrum of psychosis. That we found associations between attachment style and social functioning across groups suggests a general association, irrespective of illness or vulnerability status. However, as insecure attachment and impaired social functioning is more prevalent in individuals with psychotic disorders and those at-risk for psychosis it is of additional interest in these groups to focus on attachment style as possible contributing factor to social functioning. Further research investigating causal relationships between attachment style and social functioning is needed.

Declaration of competing interest
None.