Psychiatry Investig Search

CLOSE


Psychiatry Investig > Volume 21(4); 2024 > Article
Park, Lee, Jeong, Chung, Kim, and Jeong: The Mediating Role of Anxiety/Depression Between Auditory Verbal Hallucinations and the Level of Insight in Schizophrenia

Abstract

Objective

Auditory verbal hallucination (AVH) is a prominent symptom of schizophrenia causing profound distress. The influence of AVHs on insight appears to be intricate and contingent on other accompanying symptoms. This study investigated the relationship and possible mediators between AVHs and the degree of insight.

Methods

One hundred patients with schizophrenia participated in the study. Scales were used to evaluate the hallucinatory experience, the level of insight and other psychopathology. Complex relationships between variables were envisaged as a path model, whose initial structure was constructed via Gaussian Graphical Model. The validity of the final model was verified by Structural Equation Modeling. Separate analyses were performed for self-reported and clinician-rated data to enhance the model’s robustness.

Results

The greater the severity of the physical aspects of AVHs, the lower the level of insight observed. Conversely, higher emotional distress was associated with increased insight. These relationships were only evident in the self-reported results and were not reflected in the clinician-rated results. The path model suggested that the Positive and Negative Syndrome Scale (PANSS) anxiety/depression factor was an important mediator that linked the found association. Notably, the PANSS negative symptom had the opposite effect on the PANSS anxiety/depression factor and insight, making it difficult to define its overall effect.

Conclusion

The findings of this study provided one possible route for the positive influence of AVH experience in gaining insight. The mediating role of anxiety/depression modified by negative symptoms emerged as a valuable concept for clarifying this intricate relationship.

INTRODUCTION

Auditory verbal hallucinations (AVHs) are a key feature of schizophrenia, occurring in approximately 60%-75% of diagnosed cases [1]. However, unlike delusions, which are often understood as errors in attribution or interpretation about the situations, AVHs are less amenable to straightforward explanations [2]. Their elusive and uncontrollable quality makes it challenging for patients to understand the true nature of these experiences and to devise effective coping strategies.
Insight in the context of psychosis refers to the patient’s ability to recognize both the existence and pathological nature of their symptoms [3]. As some insight is essential for participation in the treatment process, it is a critical determinant of treatment success and overall prognosis [4]. Contrary to the popular misconception that the definition of psychosis inherently involves a lack of insight, individuals with schizophrenia can exhibit varying degrees of insight [5]. Furthermore, although the rigid belief structure of delusions often precludes insight, the precognition and sensory nature of AVHs often provides a rare opportunity for patients to realize the abnormality of their experience [6].
However, the relationship between AVHs and insight remains ambiguous. Engh et al. [6] observed that patients who experienced AVH without concomitant delusions demonstrated higher levels of insight than those without AVH. Conversely, Lera et al. [7] reported reduced levels of insight among patients with persistent AVH. Interestingly, patients who believed that voices originated in their own heads exhibited higher levels of insight than those who did not, suggesting that certain ways of experiencing AVH might influence the degree of insight [8]. Likewise, certain characteristics of AVHs, such as the physical nature, content, and emotional valence of the voices, influence insight [9]. Negative content, intimidating voices, and commanding AVHs are difficult to ignore and compel listeners to respond, undermining any attempts at resistance [10].
Exacerbating emotional distress could plausibly precipitate worsening of psychotic symptoms and loss of insight [11]. Cognitive behavioral therapy for hallucinations aims to mitigate this risk through the distress reduction technique [12]. However, the “insight paradox” posits that greater insight is correlated with increased anxiety/depression or even increased suicidal risk [13]. If the direction of cause and effect could be both ways, this paradox would also mean that greater emotional distress may contribute to the gain of insight. This raises the question of whether stress reduction invariably contributes to a healthy interpretation of their psychotic experiences. In this context, it should be noted that the positive reappraisal coping strategy for AVH leads to lower distress but is paradoxically associated with a lower level of awareness of the disease [14,15]. The score difference between the two domains of the Beck Cognitive Insight Scale (R-C index), self-reflectiveness and self-certainty, reflects the overall degree of cognitive insight [16]. Several studies have confirmed that the R-C index was significantly elevated in depressed patients with schizophrenia [17,18]. Since the ability to self-reflect is gradually impaired along with the cognitive symptoms, the denial or overly positive interpretation of their illness can indicate cognitive impairment [19]. Therefore, feeling anxious and depressed could mean that one is accurately assessing their current situation in conjunction with a relatively intact ability for self-reflection.
The multifaceted nature of both AVHs and insight necessitates a nuanced approach to their relationship [20]. In order to address this complexity, we attempted to build a path model by which different dimensions of AVHs influence the level of insight. First, the rough skeleton of this model was constructed via Gaussian Graphical Model (GGM). Subsequently, we decided on a final path model by fixing the direction of each edge based on research objectives and domain knowledge. Then, the model’s validity was verified by Structural Equation Modeling (SEM). In addition to the multidimensional assessment of AVHs and insight, the fits of the model to both self-reported and clinician-rated data were compared to enhance the robustness of our model. We hope that the results will contribute to a deeper understanding of the relationship between AVHs and insight, which is mediated and/or modified by other psychopathologies.

METHODS

Participants

Participants were recruited among patients with schizophrenia from two general hospitals. Patients aged 18 and over who had been diagnosed with schizophrenia according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition criteria were included. While patients with underlying brain disorders, intellectual disabilities, dementia, and those who reported never experiencing AVHs were excluded. Experience with AVHs and the level of insight were assessed using self-reported and clinician-rated scales. Other accompanying symptoms were measured using the Positive and Negative Syndrome Scale (PANSS). All patients voluntarily participated in the study after being informed of the protocol and signed informed consent. This study was approved by the Institutional Review Board of Ilsan Hospital, Dongguk University (approval no 2020-02-050-002) and the Institutional Review Board of Boramae Hospital, Seoul (approval no 16-2016-121).

Assessment scales

Hamilton program for schizophrenia voices questionnaire

To assess the nature of AVHs and their subjective impact on the patient, the 13-item Hamilton program for schizophrenia voices questionnaire (HPSVQ) was developed as a self-report tool [21]. The first 9 items measure the subjective experience of AVHs on a Likert scale from 0 to 4, and rest 4 items are consisted of questions which ask qualitive feature of hallucination. Severity of hallucination is calculated by summing the scores of the first 9 items. A total HPSVQ score of 0-7 was considered absent or minimal hallucination; therefore, participants with an HPSVQ score of less than seven were excluded from the actual analysis. The Korean version was standardized by Kim et al. [22] and a two-factor model consisting of physical and emotional factors was used in this study.

Psychotic symptom rating scales-auditory hallucination

The psychotic symptom rating scales (PSYRATS) is a semi-structured interview developed by Haddock et al. [23]. This scale was designed to assess patients’ experiences of both hallucinations and delusions. The section for auditory hallucinations (PSYRATS-AH) were used in the study. The Korean version was standardized by Jung et al. [24]. Three factors (physical, emotional, and cognitive) were extracted from this scale.

PANSS

The PANSS is made up of 30 items that represent different domains of schizophrenia symptoms. While the original scale proposed by Kay et al. [25] consists of three factors (positive, negative, and general symptoms), the popularity of the five-factor model has steadily increased afterwards [26]. PANSS has been standardized in Korean by Yi et al. [27]. In this study, we adopted the five-factor model proposed by Hwang et al. [28]: positive, negative, activation, autistic preoccupation, and anxiety/depression factors.

VAGUS insight into the psychosis scale

The VAGUS scale was designed by Gerretsen et al. [29]. The VAGUS Self-report (VAGUS-SR) and the VAGUS Clinician-rated (VAGUS-CR) are composed of ten and 5 items, respectively, on a Likert scale between 0 and 10. The Korean versions of VAGUS-SR and -CR were standardized by Jeong et al. [30].

Analysis

Extracting factors from the used scales

First, factors of HPSVQ (emotional, physical), PSYRATS-AH (emotional, physical), PANSS (positive, negative, anxiety/depression), and VAGUS-SR/-CR (awareness, attribution, need for treatment) were extracted by referring to existing literature [22,28,30]. The extracted factor scores were segregated into two datasets: 1) self-reported (HPSVQ and VAGUS-SR) and 2) clinically rated (PSYRATS-AH and VAGUS-CR). The data extracted from the PANSS were included in both datasets.
The complex multivariate relationship among the factors was modeled as a network via GGM. Using partial correlation coefficients as edge weights, a pair of network skeletons was obtained for each data set [31]. By integrating this pair of skeletons with domain knowledge, the final path model was proposed.

GGM

Given a set of n nodes (=variables), there can be n×(n-1)/2 possible edges between them. GGM selects important edges that maintain independent connectivity after controlling for the effects of all other nodes. Regularization methods such as EBICglasso (a graphical lasso based on extended Bayesian information criteria) are commonly used to select significant edges, but these methods cannot provide the uncertainty of edge inclusion [32]. In order to overcome this limitation, bootstrap or Bayesian approaches, such as the ones used in this study, are often employed [33].
Bayesian estimation of GGM provided credible intervals (CI) for each edge weight. The Bayesian factors (BF) that indicate the ratio of two competing statistical models (the inclusion or exclusion of this edge) could also be calculated. With these numerical indices, the final edges were selected by applying the following criteria: 1) CI should not include 0, and 2) BF should exceed 10 [34].

Validation of the path model by SEM

The initial network skeletons consisted of undirected edges that lacked causal directional information. Determining causal directions required more than just data; it involved the application of logic, common sense, domain knowledge, and a research agenda [35]. The fixation of causal directions produced the final path model containing 1) causal arrows (=regression), 2) covariation among factors, and 3) factor loadings of indicator variables.
The validity of the path models was verified by evaluating their fit with the data. Factors associated with AVH experience and PANSS factors were considered latent variables in themselves, while the three VAGUS factors were treated as indicators of a shared factor, named “insight.”

Multigroup SEM

Multigroup SEM assesses the similarities or differences between structural models applied to two or more groups [36]. Although their objectives can differ according to the research agenda, we investigated if the regression coefficients of the path model did not differ when applied to the self-report and the clinician-rated data. First, we established measurement invariance through a recommended sequence of incremental constraints (configural → weak → strong invariance). Then we examined whether the mean latent factor and regression coefficients were also identical. This sequential verification process was done by incremental chi-square tests [37].
In a strict sense, the dataset was not suitable for performing multigroup SEM because it violated the independence assumption among data. Moreover, it was unclear whether the constructs measured by HPSVQ and VAGUS-SR accurately aligned with those measured by PSYRATS and VAGUS-CR. However, the result of multigroup SEM would enable us to confirm model invariance despite group differences or identify the sources of discrepancies.
All statistical analyses were performed using the open-source software package R version 4.3.1 (R Core Team 2023, Vienna, Austria) [38]. Bayesian GGM was performed using the easyBGM package [39]. and the network comparison test (NCT) was performed using the NCT package [40]. SEM and multigroup SEM were conducted with the help of the lavaan package [41].

RESULT

Demographic characteristics of the participants

Of the original 100 patients who participated in the study, 17 patients with an HPSVQ score of seven or less, who had been considered to have no or minimal AVHs, were excluded, leaving 83 patients for final analysis. Of these, 36 were men and 47 were women. The mean age was 37.2±10.6 years, and the mean duration of the disease was 12.1±8.0 years. Twenty-six patients (31.3%) were using clozapine as their main treatment modality (Table 1).

Influence of the AVH experience on insight

In regression models based on self-report data, the physical and emotional factors of HPSVQ together explained a small but significant amount of variance in the VAGUS-SR total score (F[2, 80]=3.53, p=0.034, R2-adjusted=0.058). As expected, the physical factor showed a negative regression coefficient (β=-0.43, p=0.034), while the emotional factor showed a positive regression coefficient (β=0.69, p=0.011). In contrast, when clinician-rated data were used, the regression model hardly explained the variance of the VAGUS-CR score (F[2, 80]=1.15, p=0.320, R2-adjusted=0.004) (Figure 1).

Construction of network skeletons and the proposal of a path model

Network skeleton obtained from self-report scales

As expected, the factors belonging to the same scale were closely connected in the network skeleton (Figure 2A). As an exception, the PANSS positive factor based on Hwang’s model (PANSS-pos) was more closely related to the physical factors of HPSVQ rather than forming a cluster with other PANSS factors. Among the three factors that comprised VAGUS-SR, disease awareness (VAGUS-awa) occupied the most central position, forming a tight cluster with the other two factors.
Since the research objective was to explore the influence of AVHs on insight, it was noteworthy that only indirect routes between these two constructs were suggested by GGM. A single exception was between the physical subscale of HPSVG (HQ-phy) and awareness of the need for treatment of VAGUS-SR (VAGUS-tx), but the associated edge weight was almost negligible. The absence of a direct edge implied that much of the correlation between AVHs and insight was mediated by other psychopathologies. The most promising mediator appeared to be PANSS anxiety/depression based on Hwang’s model (PANSS-anx/dep). It occupied both the middle position of this indirect route and the central position of the entire network.
Another intriguing finding was that HQ-phy and PANSS-pos did not form any connections with VAGUS-awa. Both factors reflected the severity of the psychotic symptom itself, so it could be inferred that the severity of the symptom itself had only a minor impact on the awareness of the illness at most. Unlike PANSS-pos, the PANSS negative factor based on Hwang’s model (PANSS-neg) was observed to influence multiple aspects of insight directly and indirectly (through PANSS-anx/dep). However, while the indirect route formed a positive connection with VAGUS-awa, the direct route formed a negative connection, mutually canceling each other. Therefore, it was difficult to ascertain how PANSS-neg would affect insight in overall.

Network skeleton obtained from clinician-rated scales

Figure 2B shows the network skeleton obtained from the clinician-rated scales. Similar to the self-report network, direct connections between the AVHs and related insights could not be found. Similarly, PANSS-anx/dep was placed on the indirect route between the two constructs, which suggests that it plays a mediating role. However, the edge strength between it and VAGUS-CR, VAGUS-awa was smaller than that of the self-report network, and the direct relationships between the PANSS-neg and insight subcomponents disappeared. Instead, a new connection appeared on the network between PANSS-pos and awareness of the need for VAGUS-tx.
Although there were discrepancies between the self-reported and clinician-rated networks in terms of edge presences and strengths, the overall structures were not so dissimilar. The results of the NCT also substantiated this impression. The overall network structure (network invariance test: M=0.213, p=0.733) and edge strengths (global strength invariance test: S=0.236, p=0.911) were not significantly different between the two networks.

Centrality plot of the network skeletons

Figure 3 shows the centrality plot of the network skeletons obtained by both self-report scales and clinician-rated scales. PANSS-anx/dep had the highest closeness and betweenness centrality index among the included variables.

Building and validating a path model

Figure 4 displays the final path model along with the coefficients fitted to the datasets. Both the self-report (χ2 [16]=17.51, p=0.354) and clinician-rated data (χ2 [16]=17.57, p=0.350) did not deviate significantly from the model, and other fit measures (comparative fit index [CFI], root mean square error of Approximation [RMSEA], RMSEA p-value) also indicated a satisfactory fit (Table 2).

Model robustness by multigroup SEM

From structural to strict invariance, the measurement invariance could be established between the two datasets. Finally, even after the mean of the latent factor and regression coefficients were constrained to be the same between them, the model fit did not deteriorate significantly (Table 3). This result suggested that the same path model, up to the same regression coefficient, could consistently be applied whether evaluated by self-report or clinician-rating.

DISCUSSION

In this study, we explored the relationship between AVH and insight in patients with schizophrenia, while also investigating other psychopathologies that could serve as mediators or moderators. The self-reported experience of AVH, as assessed by HPSVQ, was significantly related to the level of insight self-reported by VAGUS-SR. However, no such relationship was observed when clinicians evaluated the same constructs using PSYRATS-AH and VAGUS-CR. This discrepancy underscores the notion that the influence on insight was more apparent in patients’ inner experiences than in observable behaviors and speech.
In addition, we explored the role of accompanying symptoms, specifically positive, negative, and anxiety/depression symptoms, as measured by PANSS. In particular, the network skeletons revealed no direct links between AVHs and insight. This suggested the presence of mediating variables, most likely anxiety/depression between them. Starting from the skeletons, we formulated a path model that was closely aligned with domain knowledge, which showed a satisfactory fit to the dataset. Multigroup SEM confirmed the invariance of the model regardless of the data collection method. With this path model, it could be hypothesized that the emotional aspect of AVH experiences intensifies patients’ overall anxiety/depression, which in turn positively influences their level of insight.
Contrary to mainstream perspectives such as insight paradox or post-psychotic depression (PPD), our results indicated that anxiety/depression could play a facilitative role in acquiring insight [42-44]. Although the insight paradox and PPD frameworks subtly imply that a patient must have a certain degree of insight to become depressed, clinicians often encounter cases in which some degree of internal distress helps patients gain insight [14,42]. Insight is not an all-or-nothing phenomenon and is not always necessary for experiencing anxiety/depression [45]. From an evolutionary standpoint, anxiety/depression functions as a warning system to keep us alert to potential dangers and to encourage us to find the cause [46]. Insight is a form of pattern finding in the natural world that provides an understanding of a specific causal relationship within a particular context [47]. The only question to be answered is whether the patterns found are correct and useful or wrong and harmful. Our data suggest that the anxiety/depression resulting from AVHs may catalyze self-reflection of the condition in accordance with the above theory.
We believed that anxiety/depression was a cause of insight rather than an outcome not only because of research agendas but also because network topology supported this direction. The causal direction of AVHs → anxiety/depression was evident. If the other causal arrow had been fixed as insight → anxiety/depression, a triadic structure would have formed, as shown in Figure 5A. However, since no direct edges connecting AVHs and insight were suggested in the network skeletons, the causal structure would become as in Figure 5B. This structure has traditionally been called a collider, which blocks the correlation between the two ends unless the intervening variable (=anxiety/depression) is controlled [48]. Since we found a significant correlation between AVH and insight not controlling for anxiety/depression, the structure of the collider was less likely. Then, the only remaining triadic structure would be the mediating structure, as shown in Figure 5C.
Contextual factors, such as accompanying symptoms or environmental conditions, should be considered to better understand the mediating role of anxiety/depression on insight, especially as induced by AVHs [28,49-51]. Although the defining feature of AVHs is indistinguishability from actual perceptions, many patients recognize something unusual in their hallucinatory experiences [7,52]. In some patients with intact cognitive function, the aberrant nature of AVHs may prompt them to seek help [53]. Cultural and religious perspectives are also important, as they could either facilitate or obstruct the acquisition of correct insight [54]. Most persons believe that hearing voices is often considered to be the utmost evidence of insanity, such that many patients receive their first psychiatric evaluation simply because they hear “voices.” [55,56]
Anxiety/depression showed a positive relationship with the emotional factor and a negative relationship with the physical factor in the self-report data, which coincided with the signs of those observed between insight and the same two factors. This consistency also supported the hypothesis that certain dimensions of the AVH experience contributed to gain insight through the mediation of anxiety/depression. However, the overall influence of AVHs on insight may be a double-edged sword. Severe AVHs cause patients to lose reality testing and insight. The negative influence of the physical factor of AVHs on the insight reflected this point. Meanwhile, how the patient accepts and interprets these strange experiences may be also important. Applying social rank theory [57], Birchwood et al. [42] argued that a patient’s beliefs about the power and authority of the voices determine the extent of their depression. The authors also observed that depressive feelings, combined with a sense of loss, humiliation, or defeat, were positively correlated with a higher level of insight.
A similar, dualistic influence was also observed in the relationship between negative symptoms and insight. Direct effects and indirect effects through anxiety/depression on insight had opposing signs that effectively nullified each other. It is well documented that negative symptoms have a detrimental effect on insight [51,58]; indicative of cognitive impairment, they can hinder a patient’s ability to objectively understand their own situation. However, negative symptoms exacerbate anxiety/depression or often disguise themselves in the form of anxiety/depression. If not properly distinguished, the following contradiction could not easily be resolved: negative symptoms had a positive correlation with depression, and depression positively correlated with insight, but contradictorily, negative symptoms were negatively correlated with insight. To resolve this, anxiety/depression may have to be viewed as a mixture of different psychopathological domains; for example, anxiety/depression derived from normal aspect of human nature and those from deepening psychosis. The lack of consensus on defining depression comorbid with schizophrenia exacerbates this confusion [59]. Anxiety/depression paired with intact cognitive function could foster normal distress and self-reflection, while anxiety/depression paired with impaired cognition could aggravate fear and dread of psychotic experiences and undermine fragile insight. In order to address these challenges, it is crucial to phenomenologically dissect the anxiety/depression displayed by psychotic patients and develop tools for evaluating each aspect [59].
The difference between self-reported and clinician-rated data warrants further discussion. The key findings of this study were primarily derived from self-reported data. Although we verified that the same path model could be applied to both datasets, neither AVH factor significantly influenced the insight in the clinician-rated data. The rating tools used in this study have been subjected to comparative studies, highlighting the similarities and differences between self-report and clinician-rating forms (HPSVQ vs. PSYRATS, VAGUS-SR vs, -CR) [22,30]. Previous studies have suggested that clinicians often focus more on individual symptom improvement rather than overall disease progression, associating disappearance with relief of distress. Conversely, patients’ self-reports typically reflect subjective pain and social maladjustment. Therefore, it is preferable to prioritize the inner experiences when investigating the mediating role of anxiety/depression [60].
This study has some clear limitations. Notably, the small sample size might have undermined the reliability of the SEM results. The minimum sample size for SEM would be 5 or 10 observations per estimated parameter or 100 or 200 observations in total [61]. Since the number of estimated parameters was approximately 20 (numbers of regression coefficients, residual variances, factor loadings, and covariances were all combined), the required sample size would be at least 200. Furthermore, our decision to focus solely on a selected subset of PANSS factors to minimize the number of parameters (excluding disorganization and hostility factors) may lack a solid rationale; however, further analysis that included all factors showed similar patterns (results not shown).
In the process of transforming the network skeletons into a path model, we made several discretionary decisions. The common sense and domain knowledge used in deriving the path model were not absolute criteria; hence, different path models could have been obtained [62,63]. Although it might be possible to evaluate all candidate models and choose the most suitable one, this approach would contradict our study’s objective of confirmatory analysis. Meanwhile, attempts to show the robustness of the path model through multigroup SEM also had notable challenges. Essentially, for a successful multigroup SEM, observations from different groups must be independent [37]. However, this study deviated from this basic requirement by measuring the same person in two different ways.
In conclusion, we studied the effects of AVHs on insight and investigated how other symptoms of schizophrenia influenced this relationship. We used GGM to determine a rough skeleton of the path model, and then validated our path model through confirmatory SEM. Our findings showed that the physical and emotional aspects of the AVH experiences significantly influenced the level of insight. We also found evidence suggesting that patients’ anxiety/depression was a key mediator of this influence. Despite some controversies and limitations, our results indicated that anxiety/depression might not only be the outcome of insight but also the cause of it, especially if it arose from AVHs experiences. However, this should not be interpreted as suggesting that to help patients gain insight, it is necessary to exacerbate the patient’s anxiety/depression. Only adaptive responses to AVH, responses expected from any ordinary individual facing such adverse experiences, should be acknowledged as valid forms of anxiety/depression. Any other form closer to negative symptoms should be adequately addressed through active treatment.

Notes

Availability of Data and Material

The datasets generated or analyzed during the study are not publicly available due the possible infringement of data privacy, but are available from the corresponding author on reasonable request.

Conflicts of Interest

The authors have no potential conflicts of interest to disclose.

Author Contributions

Conceptualization: In Won Chung, Yong Sik Kim, Seong Hoon Jeong. Data curation: Nam Young Lee, Hee Yeon Jeong. Formal analysis: Sang Won Park, Seong Hoon Jeong. Investigation: Nam Young Lee, Hee Yeon Jeong. Methodology: Nam Young Lee, In Won Chung, Seong Hoon Jeong. Project administration: Sang Won Park, Nam Young Lee. Resources: Sang Won Park, In Won Chung. Software: Seong Hoon Jeong. Supervision: Yong Sik Kim, Seong Hoon Jeong. Validation: Sang Won Park, Seong Hoon Jeong, Yong Sik Kim. Visualization: Sang Won Park, Seong Hoon Jeong. Writing—original draft: Sang Won Park, Seong Hoon Jeong. Writing—review & editing: Sang Won Park, Seong Hoon Jeong.

Funding Statement

None

ACKNOWLEDGEMENTS

None

Figure 1.
The pairwise bivariate relationship between factors of AVH experience (physical and emotional) and the level of insight (measured by VAGUS total score). A and B: When measured by self-report scales (HPSVQ for AVH experiences and VAGUS-SR for insight). C and D: When measured by clinician-rated scales (PSYRATS for AVH experiences and VAGUS-CR for insight). AVH, auditory verbal hallucination; VAGUS-SR, VAGUS Self-report; VAGUS-CR, VAGUS Clinician-rated; HPSVQ, Hamilton program for schizophrenia voices questionnaire; PSYRATS, psychotic symptom rating scales.
pi-2023-0396f1.jpg
Figure 2.
Network skeletons depicting the most relevant relations among the involved variables. They were obtained by Bayesian GGM. Two skeletons were obtained respectively from (A) self-report data and (B) clinician-rated data. Numbers attached on the edges are partial correlation coefficients (=edge strengths), and the edge widths reflect these coefficients. The size of a node is proportional to the node-weighted centrality (sum of all the connected edge strengths). Please refer to the main text for the heuristic criteria of edge inclusion. GGM, Gaussian Graphical Model; HQ-emo, emotional subscale of HPSVQ; HQ-phy, physical subscale of HPSVQ; P-neg, PANSS negative factor based on Hwang’s model; P-pos, PANSS positive factor based on Hwang’s model; P-anx, PANSS anxiety-depressive factor based on Hwang’s model; V-att, symptom attribution subscale of VAGUS-SR; V-awa, disease awarness subscale of VAGUS-SR; V-tx, awareness of the need for treatment subscale of VAGUS-SR; PSY-emo, emotional subscale of PSYRATS; PSY-phy, physical subscale of PSYRATS.
pi-2023-0396f2.jpg
Figure 3.
Centrality plot of the network skeletons obtained by GGM. Four kinds of centrality measures (strength, closeness, betweenness, expected influence) were plotted. Overall, PANSS-anx/dep and VAGUS-awa occupied the most central position in the network skeletons. GGM, Gaussian Graphical Model; VAGUS-tx, awareness of the need for treatment of VAGUS-SR; VAGUS-awa, disease awarness; VAGUS-att, symptom attribution subscale of VAGUS insight scale; PANSS-pos, PANSS positive factor based on Hwang’s model; PANSS-neg, PANSS negative factor based on Hwang’s model; PANSS-anx/dep, PANSS anxiety-depressive factor based on Hwang’s model; Hal-physical, physical subscale of hallucination scale (Self-report: HPSVQ, Clinician rated: PSYRATS); Hal-emotional, emotional subscale of hallucination scale (Self-report: HPSVQ, Clinician rated: PSYRATS).
pi-2023-0396f3.jpg
Figure 4.
The final path model fitted to the two sets of data (self-report and clinician-rated). The numbers attached on the edges were the regression coefficients and the factor loadings. Unidirectional and bidirectional arrows implied causal relationship and covariation respectively. A: Path model estimated from self-rated data. B: Path model estimated from clinician-rated data. HQ-emo, emotional subscale of HPSVQ; HQ-phy, physical subscale of HPSVQ; PANSS-neg, PANSS negative factor based on Hwang’s model; PANSS-pos, PANSS positive factor based on Hwang’s model; PANSS-anx/dep, PANSS anxiety-depressive factor based on Hwang’s model; VAGUS-att(s), symptom attribution subscale of VAGUS-SR; VAGUS-awa(s), disease awarness subscale of VAGUS-SR; VAGUS-tx(s), awareness of the need for treatment subscale of VAGUS-SR; PSY-emo, emotional subscale of PSYRATS; PSY-phy, physical subscale of PSYRATS; VAGUS-att(c), symptom attribution subscale of VAGUS-CR; VAGUS-awa(c), disease awarness subscale of VAGUS-CR; VAGUS-tx(c), awareness of the need for treatment subscale of VAGUS-CR.
pi-2023-0396f4.jpg
Figure 5.
Hypothetical causal relationship among hallucination, insight, and anxiety/depression. If we assume that both hallucination and insight cause anxiety/depression, then a triadic structure as in (A) can be considered. However, since no direct relationship between hallucination and insight had been found in our dataset, the only possible pathway forms a collider structure (B). Collider structure is unlikely because there was significant correlation between hallucination and insight. Therefore, a mediating structure with the causal direction from anxiety/depression to insight is more reasonable (C).
pi-2023-0396f5.jpg
Table 1.
Demographic characteristics of the participants and the summary statistics of the measured scales
Male (N=36) Female (N=47) Total (N=83) p*
Age (yr) 35.9±9.9 38.1±11.0 37.2±10.6 0.353
Clozapine use 9 (25.0) 17 (36.2) 26 (31.3) 0.396
Hallucinatory experience
 HPSVQ
  Physical 13.4±5.4 12.4±4.7 12.8±5.0 0.387
  Emotional 9.2±4.1 7.6±3.3 8.3±3.8 0.044
  Total 22.6±8.6 20.0±7.4 21.1±8.0 0.137
 PSYRATS
  Physical 6.8±4.7 6.7±3.6 6.7±4.1 0.939
  Emotional 6.0±4.8 4.3±4.5 5.0±4.7 0.107
  Cognition 7.2±4.9 7.0±4.1 7.1±4.5 0.879
  Total 16.1±12.6 14.1±10.3 15.0±11.4 0.445
Level of insight
 VAGUS-SR
  Awareness 5.7±2.5 4.8±2.8 5.2±2.7 0.128
  Attribution 4.9±1.5 4.6±1.5 4.7±1.5 0.435
  Need for treatment 6.8±2.4 6.2±2.4 6.4±2.4 0.250
  Total 23.3±7.0 20.4±6.8 21.7±7.0 0.067
 VAGUS-CR
  Awareness 6.2±3.0 5.9±3.0 6.0±3.0 0.609
  Attribution 5.4±2.7 5.6±2.8 5.5±2.8 0.773
  Need for treatment 8.2±2.5 6.5±3.1 7.2±2.9 0.011
  Total 25.8±8.0 24.1±9.1 24.8±8.7 0.387
Accompanying symptoms
 PANSS
  Positive 13.3±3.7 12.3±4.2 12.8±4.0 0.278
  Negative 18.9±5.5 16.1±5.6 17.3±5.7 0.027
  Anxiety/depression 10.1±3.8 8.2±3.1 9.0±3.5 0.015

Values are presented as mean±SD or number (%).

* p values were calculated by student t-tests for continuous variables and χ²-tests for categorical variables.

HPSVQ, Hamilton program for schizophrenia voices questionnaire; PSYRATS, psychotic symptom rating scales; VAGUS-SR, VAGUS Self-report; VAGUS-CR, VAGUS Clinician-rated; PANNS, Positive and Negative Symptom Scale; SD, standard deviation

Table 2.
The result of the SEM. The self-report data and the clinician-rated data were separately fitted to the final path model
χ² df p CFI RMSEA RMSEA p AIC BIC
Self-report 17.505 16 0.354 0.993 0.034 0.565 3,287.803 3,355.531
Clinician-rated 17.569 16 0.350 0.993 0.034 0.561 3,445.870 3,513.598

Both data showed adequate fit. Therefore, it can be concluded that the model was invariant to data source (self-report or clinician-rated).

SEM, Structural Equation Modeling; χ², chi square distribution; df, degree of freedom; CFI, comparative fit index; RMSEA, root mean square error of approximation; RMSEA p, p-value of root mean square error of approximation; AIC, Akaike’s information criterion; BIC, Bayesian information criterion

Table 3.
The result of the multigroup SEM. Measurement invariance (structural to strict invariance) and structural invariance were verified through multigroup SEM
df AIC BIC χ2 Δχ2 RMSEA Δdf p
No grouping 34 3403.5 3571.5 45.106
Structural invariance 36 3404.1 3565.9 49.689 4.5827 0.1247 2 0.1011
Weak invariance 43 3390.1 3530.1 49.689 0 0 7 1
Strong invariance 51 3378.4 3493.5 54.020 4.3314 0 8 0.8261
Strict invariance 53 3374.7 3483.6 54.326 0.3053 0 2 0.8584
Equal regression coefficients 59 3367.1 3457.3 58.704 4.3779 0 6 0.6257

The sequential restriction from “no grouping” to “equal regression coefficients” produced no significant deterioration in the model fit from less restricted model. Therefore, it can be concluded that the model was invariant to data source (self-report or clinician-rated). SEM, Structural Equation Modeling; χ2, chi square distribution; df, degree of freedom; CFI, comparative fit index; RMSEA, root mean square error of approximation; AIC, Akaike’s information criterion; BIC, Bayesian information criterion

REFERENCES

1. Nayani TH, David AS. The auditory hallucination: a phenomenological survey. Psychol Med 1996;26:177-189.
crossref pmid
2. Honda T, Suzuki H, Iwai K, Fujiwara Y, Kawahara H, Kuroda S. Autochthonous experience, heightened awareness, and perception distortion in patients with schizophrenia: a symptomatological study. Psychiatry Clin Neurosci 2004;58:473-479.
crossref pmid
3. Amador XF, Strauss DH, Yale SA, Flaum MM, Endicott J, Gorman JM. Assessment of insight in psychosis. Am J Psychiatry 1993;150:873-879.
crossref pmid
4. Lysaker PH, Pattison ML, Leonhardt BL, Phelps S, Vohs JL. Insight in schizophrenia spectrum disorders: relationship with behavior, mood and perceived quality of life, underlying causes and emerging treatments. World Psychiatry 2018;17:12-23.
crossref pmid pmc pdf
5. Belvederi Murri M, Amore M. The multiple dimensions of insight in schizophrenia-spectrum disorders. Schizophr Bull 2019;45:277-283.
crossref pmid pmc
6. Engh JA, Friis S, Birkenaes AB, Jónsdóttir H, Klungsøyr O, Ringen PA, et al. Delusions are associated with poor cognitive insight in schizophrenia. Schizophr Bull 2010;36:830-835.
crossref pmid pmc
7. Lera G, Herrero N, González J, Aguilar E, Sanjuán J, Leal C. Insight among psychotic patients with auditory hallucinations. J Clin Psychol 2011;67:701-708.
crossref pmid
8. Docherty NM, Dinzeo TJ, McCleery A, Bell EK, Shakeel MK, Moe A. Internal versus external auditory hallucinations in schizophrenia: symptom and course correlates. Cogn Neuropsychiatry 2015;20:187-197.
crossref pmid pmc
9. Toh WL, Moseley P, Fernyhough C. Hearing voices as a feature of typical and psychopathological experience. Nat Rev Psychol 2022;1:72-86.
crossref pdf
10. West ML. Clinical insights: addressing command hallucinations. Schizophr Res 2018;197:63-64.
crossref pmid
11. Norman RM, Malla AK. A prospective study of daily stressors and symptomatology in schizophrenic patients. Soc Psychiatry Psychiatr Epidemiol 1994;29:244-249.
crossref pmid pdf
12. Pontillo M, De Crescenzo F, Vicari S, Pucciarini ML, Averna R, Santonastaso O, et al. Cognitive behavioural therapy for auditory hallucinations in schizophrenia: a review. World J Psychiatry 2016;6:372-380.
crossref pmid pmc
13. Belvederi Murri M, Amore M, Calcagno P, Respino M, Marozzi V, Masotti M, et al. The “insight paradox” in schizophrenia: magnitude, moderators and mediators of the association between insight and depression. Schizophr Bull 2016;42:1225-1233.
crossref pmid pmc
14. Cooke M, Peters E, Fannon D, Anilkumar AP, Aasen I, Kuipers E, et al. Insight, distress and coping styles in schizophrenia. Schizophr Res 2007;94:12-22.
crossref pmid pmc
15. Tsang A, Bucci S, Branitsky A, Kaptan S, Rafiq S, Wong S, et al. The relationship between appraisals of voices (auditory verbal hallucinations) and distress in voice-hearers with schizophrenia-spectrum diagnoses: a meta-analytic review. Schizophr Res 2021;230:38-47.
crossref pmid
16. Beck AT, Baruch E, Balter JM, Steer RA, Warman DM. A new instrument for measuring insight: the Beck Cognitive Insight Scale. Schizophr Res 2004;68:319-329.
crossref pmid
17. Warman DM, Lysaker PH, Martin JM. Cognitive insight and psychotic disorder: the impact of active delusions. Schizophr Res 2007;90:325-333.
crossref pmid
18. Palmer EC, Gilleen J, David AS. The relationship between cognitive insight and depression in psychosis and schizophrenia: a review and meta-analysis. Schizophr Res 2015;166:261-268.
crossref pmid
19. van der Meer L, de Vos AE, Stiekema AP, Pijnenborg GH, van Tol MJ, Nolen WA, et al. Insight in schizophrenia: involvement of self-reflection networks? Schizophr Bull 2013;39:1288-1295.
crossref pmid pmc
20. Ouzir M, Azorin JM, Adida M, Boussaoud D, Battas O. Insight in schizophrenia: from conceptualization to neuroscience. Psychiatry Clin Neurosci 2012;66:167-179.
crossref pmid
21. Van Lieshout RJ, Goldberg JO. Quantifying self-reports of auditory verbal hallucinations in persons with psychosis. Can J Behav Sci 2007;39:73-77.
crossref
22. Kim SH, Jung HY, Hwang SS, Chang JS, Kim Y, Ahn YM, et al. The usefulness of a self-report questionnaire measuring auditory verbal hallucinations. Prog Neuropsychopharmacol Biol Psychiatry 2010;34:968-973.
crossref pmid
23. Haddock G, McCarron J, Tarrier N, Faragher EB. Scales to measure dimensions of hallucinations and delusions: the psychotic symptom rating scales (PSYRATS). Psychol Med 1999;29:879-889.
crossref pmid
24. Jung SM, Kim MK, Lee JB, Choi JH, Jung BJ, Byun WT. [Reliability and validity of the Korean version of the psychotic symptom rating scale]. J Korean Neuropsychiatr Assoc 2007;46:201-213. Korean.

25. Kay SR, Fiszbein A, Opler LA. The positive and negative syndrome scale (PANSS) for schizophrenia. Schizophr Bull 1987;13:261-276.
crossref pmid
26. Shafer A, Dazzi F. Meta-analysis of the positive and negative syndrome scale (PANSS) factor structure. J Psychiatr Res 2019;115:113-120.
crossref pmid
27. Yi JS, Ahn YM, Shin HK, An SK, Joo YH, Kim SH. [Reliability and validity of the Korean version of the positive and negative syndrome scale]. J Korean Neuropsychiatr Assoc 2001;40:1090-1105. Korean.

28. Hwang SS, Chang JS, Lee KY, Kim SH, Ahn YM, Kim YS. Causal model of insight and psychopathology based on the PANSS factors: 1-year cross-sectional and longitudinal revalidation. Int Clin Psychopharmacol 2009;24:189-198.
crossref pmid
29. Gerretsen P, Remington G, Borlido C, Quilty L, Hassan S, Polsinelli G, et al. The VAGUS insight into psychosis scale--self-report and clinician-rated versions. Psychiatry Res 2014;220:1084-1089.
crossref pmid pmc
30. Jeong SH, Chung IW, Jung HY, Hwang SS, Kim SH, Youn T, et al. Comparison of clinician-rated and self-report insight in Korean patients with schizophrenia using VAGUS insight scale. Psychiatry Res 2017;258:93-100.
crossref pmid
31. Epskamp S, Waldorp LJ, Mõttus R, Borsboom D. The Gaussian graphical model in cross-sectional and time-series data. Multivariate Behav Res 2018;53:453-480.
crossref pmid
32. Epskamp S, Fried EI. A tutorial on regularized partial correlation networks. Psychol Methods 2018;23:617-634.
crossref pmid
33. Williams DR. Bayesian estimation for Gaussian graphical models: structure learning, predictability, and network comparisons. Multivariate Behav Res 2021;56:336-352.
crossref pmid
34. Huth KB, de Ron J, Goudriaan AE, Luigjes J, Mohammadi R, van Holst RJ, et al. Bayesian analysis of cross-sectional networks: a tutorial in R and JASP. Adv Meth Pract Psychol Sci 2023;6:25152459231193334

35. Glymour C, Zhang K, Spirtes P. Review of causal discovery methods based on graphical models. Front Genet 2019;10:524
crossref pmid pmc
36. Jöreskog KG. Simultaneous factor analysis in several populations. Psychometrika 1971;36:409-426.
crossref pdf
37. Yuan KH, Chan W. Measurement invariance via multigroup SEM: issues and solutions with chi-square-difference tests. Psychol Methods 2016;21:405-426.
crossref pmid
38. R Core Team. R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing; 2021.

39. Huth K, Keetelaar S, Sekulovski N, van den Bergh D, Marsman M. Simplifying Bayesian analysis of graphical models for the social sciences with easybgm: a user-friendly R-package. PsyArxiv [Preprint] 2023;[cited November 14, 2023]. Available at: https://doi.org/10.31234/osf.io/8f72p.
crossref
40. van Borkulo CD, van Bork R, Boschloo L, Kossakowski JJ, Tio P, Schoevers RA, et al. Comparing network structures on three aspects: a permutation test. Psychol Methods 2023;28:1273-1285.
crossref pmid
41. Rosseel Y. lavaan: an R package for structural equation modeling. J Stat Softw 2012;48:1-36.

42. Birchwood M, Iqbal Z, Upthegrove R. Psychological pathways to depression in schizophrenia: studies in acute psychosis, post psychotic depression and auditory hallucinations. Eur Arch Psychiatry Clin Neurosci 2005;255:202-212.
crossref pmid pdf
43. Upthegrove R, Marwaha S, Birchwood M. Depression and schizophrenia: cause, consequence, or trans-diagnostic issue? Schizophr Bull 2017;43:240-244.
crossref pmid pmc
44. Guerrero-Jiménez M, Carrillo de Albornoz Calahorro CM, GirelaSerrano B, Bodoano Sánchez I, Gutiérrez-Rojas L. Post-psychotic depression: an updated review of the term and clinical implications. Psychopathology 2022;55:82-92.
crossref pmid pdf
45. David AS. Insight and psychosis. Br J Psychiatry 1990;156:798-808.
crossref pmid
46. Bateson M, Brilot B, Nettle D. Anxiety: an evolutionary approach. Can J Psychiatry 2011;56:707-715.
crossref pmid pdf
47. Mayer RE. The search for insight: grappling with Gestalt psychology’s unanswered questions. In: Sternberg RJ, Davidson JE, editor. The nature of insight. Cambridge: MIT Press, 1995, p. 3-32.

48. Holmberg MJ, Andersen LW. Collider bias. JAMA 2022;327:1282-1283.
crossref pmid
49. Amador XF, Gorman JM. Psychopathologic domains and insight in schizophrenia. Psychiatr Clin North Am 1998;21:27-42.
crossref pmid
50. Joseph B, Narayanaswamy JC, Venkatasubramanian G. Insight in schizophrenia: relationship to positive, negative and neurocognitive dimensions. Indian J Psychol Med 2015;37:5-11.
crossref pmid pmc pdf
51. Mintz AR, Dobson KS, Romney DM. Insight in schizophrenia: a meta-analysis. Schizophr Res 2003;61:75-88.
crossref pmid
52. Cavelti M, Kvrgic S, Beck EM, Rüsch N, Vauth R. Self-stigma and its relationship with insight, demoralization, and clinical outcome among people with schizophrenia spectrum disorders. Compr Psychiatry 2012;53:468-479.
crossref pmid
53. Johns LC, Kompus K, Connell M, Humpston C, Lincoln TM, Longden E, et al. Auditory verbal hallucinations in persons with and without a need for care. Schizophr Bull 2014;40(Suppl 4):S255-S264.
crossref pmid pmc
54. Larøi F, Luhrmann TM, Bell V, Christian WA Jr, Deshpande S, Fernyhough C, et al. Culture and hallucinations: overview and future directions. Schizophr Bull 2014;40(Suppl 4):S213-S220.
crossref pmid pmc
55. Kapur P, Hayes D, Waddingham R, Hillman S, Deighton J, Midgley N. The experience of engaging with mental health services among young people who hear voices and their families: a mixed methods exploratory study. BMC Health Serv Res 2014;14:527
crossref pmid pmc pdf
56. Lonergan A. The meaning of voices in understanding and treating psychosis: moving towards intervention informed by collaborative formulation. Eur J Psychol 2017;13:352-365.
crossref pmid pmc pdf
57. Gilbert P, Miles JN. Sensitivity to social put-down: it’s relationship to perceptions of social rank, shame, social anxiety, depression, anger and self-other blame. Pers Individ Dif 2000;29:757-774.
crossref
58. Cuesta MJ, Peralta V, Zarzuela A. Psychopathological dimensions and lack of insight in schizophrenia. Psychol Rep 1998;83(3 Pt 1):895-898.
crossref pmid pdf
59. Hausmann A, Fleischhacker WW. Differential diagnosis of depressed mood in patients with schizophrenia: a diagnostic algorithm based on a review. Acta Psychiatr Scand 2002;106:83-96.
crossref pmid pdf
60. Kyzar EJ, Denfield GH. Taking subjectivity seriously: towards a unification of phenomenology, psychiatry, and neuroscience. Mol Psychiatry 2023;28:10-16.
crossref pmid pdf
61. Sim M, Kim SY, Suh Y. Sample size requirements for simple and complex mediation models. Educ Psychol Meas 2022;82:76-106.
crossref pmid pmc pdf
62. VanderWeele TJ, Hernán MA, Robins JM. Causal directed acyclic graphs and the direction of unmeasured confounding bias. Epidemiology 2008;19:720-728.
crossref pmid pmc
63. Lipsky AM, Greenland S. Causal directed acyclic graphs. JAMA 2022;327:1083-1084.
crossref pmid


ABOUT
AUTHOR INFORMATION
ARTICLE CATEGORY

Browse all articles >

BROWSE ARTICLES
Editorial Office
#522, 27, Seochojungang-ro 24-gil, Seocho-gu, Seoul 06601, Korea
Tel: +82-2-717-0892    E-mail: psychiatryinvest@gmail.com                

Copyright © 2024 by Korean Neuropsychiatric Association.

Developed in M2PI

Close layer
prev next