Keywords
Negative symptoms, network analysis, neurocognition, schizophrenia, social cognition
Negative symptoms, network analysis, neurocognition, schizophrenia, social cognition
Minor revisions:
This version mainly elaborates the discussion section.
See the author's detailed response to the review by Haruo Fujino
See the author's detailed response to the review by Pichai Ittasakul
Schizophrenia is associated with a broad range of symptoms including positive symptoms, negative symptoms and neurocognition impairments. Recently, social cognition deficits also have been increasingly reported.1,2 Despite the increase in new medications and treatments, functional outcome is not as good as expected. Functional recovery is observed in less than 25% of patients with schizophrenia.3
While all of the symptoms are believed to affect functional outcome to some extent, negative symptoms, neurocognitive, and social cognition seem to have the highest impact on functioning.4 For example, previous studies generally demonstrate that negative symptoms have associations with poorer levels of functioning.5 Neurocognitive impairment is a predictor for low levels of a real-world functioning.6 Moreover, in the last decade, social cognition, a relatively new concept, appears to be a strong predictor of social abilities and real-world function. Theory of mind and emotion recognition are the most extensively studied domains in social cognition study, and impairment in these tasks are risk factors for poor social and functional outcome in patients with schizophrenia.4,6
However, the interplay between these symptoms and functioning is highly complex. Since not only can negative symptoms, neurocognition, and social cognition predict functional outcome, all of these symptoms are also interrelated to each other. For example, previous studies demonstrated that negative symptoms were closely associated with social cognition.7,8 Social cognition and neurocognition are also usually correlated to some degree.9 Moreover, some studies suggest that social cognition might be a mediator between neurocognition and symptomatology.10 Therefore, as noted before, a traditional statistical approach might not be suitable for examining complex relationships of interconnected variables. For instance, a simple correlation approach could not control for the influence of other variables. Linear regression and structural equation models require a priori assumptions regarding the selection of predictors, mediators, precursors, and outcomes. Furthermore, traditional analysis could not demonstrate which variables connect to other variable more often and which ones play a more or less central role than others.
Network analysis is a relatively new and powerful methodological approach to investigate complex relationship patterns. It is a data-driven technique that does not require a priori assumption of relationships among variables. With a network analysis approach, all phenomena are conceptualized as systems of causally connected signs and symptoms. The system can be analysed and presented in its full complexity.11 In the network, the key variables will be located at the center of the network, while less important and connected variables will be in the periphery. The aim of this study was to explore the interplay of positive symptoms, five groups of negative symptoms, neurocognition, two forms of social cognition (theory of mind, emotion recognition), and functioning in patients with clinically stable schizophrenia using network analysis.
64 outpatients with schizophrenia were recruited from a mental health clinic at Thammasat University Hospital, Thailand between February 2018 and August 2019. They were between 20 and 60 years old and had at least four years of education. All patients met criteria for schizophrenia based upon the Diagnostic and Statistical Manual of Mental Disorders-5 (DSM-5) criteria 12 and were in a clinically stable phase, defined as no changes in treatment or symptoms for the previous three months. Exclusion criteria were having a major neurological disorder or other psychiatric disorders (i.e., intellectual disability, active major depressive disorder and substance dependence excluding smoking). The participants were invited to participate in this study by our research assistants during outpatient visits. The study was approved by the Human Research Ethics Committee of Thammasat University (No. MTU-EC-PS-0-191/60). Written informed consent forms were obtained from all participants. Participants had a right to withdraw from the study at any time.
Positive symptoms
Positive symptoms were measured using the Scale for the Assessment of Positive Symptoms (SAPS).13,14 Positive symptoms were presented as a single unitary construct using the sum of global ratings. The global ratings range from 0 to 20 with the higher score indicating more severe positive symptoms.
Negative symptoms
The Scale for the Assessment of Negative Symptoms (SANS) was administered to assess negative symptoms.15,16 In this study, negative symptoms were classified into five subdomains i.e., blunted affect, alogia, anhedonia, avolition, and asociality as suggested by the NIMH-MATRICS consensus statement on negative symptoms.17 Each subdomain has a maximum score of five, which is computed from the average scores of the relevant items of each subdomain. It should be noted that inattention subdomain was not included in the analysis because of its overlap with neurocognitive assessment.
We used five dimensions of negative symptoms instead of a unidimensional approach because several recent factor analysis studies suggest that a unidimensional model of negative symptoms does not adequately capture the complexity of negative symptoms.17,18 Furthermore, previous study showed that negative symptoms subdomains seem to have a different relationship with functioning. For example, avolition and apathy seem to be a stronger predictor of poor outcomes than blunted affect.4,19 Therefore, utilizing a multidimensional model would avoid losing information relevant to the connection with other variables and the ability of each subdomain to predict functioning.
Neurocognition
The Addenbrooke’s Cognitive Examination III (ACE-III) was used to assess neurocognitive function. It assesses five neurocognitive domains: attention, verbal fluency, memory, language, and visuospatial ability with a maximum score of 100 points.20,21A previous study demonstrated that ACE-III was sensitive to detect cognitive impairment in patients with schizophrenia.22 A Thai version of the ACE-III had an excellent inter-rater reliability with a Pearson’s correlation of 1.0 (p < 0.001).20
Social cognition
The social cognition assessments included tests of emotion recognition (Faces test),23 and theory of mind (Reading the Mind in the Eyes test: RMET).24,25 The Faces test consists of 20 photographs of people faces showing a variety of emotions. The participants were required to match emotions to faces. The maximum score is 20 with a higher score suggesting better emotion recognition ability.23 The RMET includes 36 pictures of persons’ eye regions where participants must select which of four words best describes the mental state of a target person. The RMET has a score range of 0-36 with a higher score indicating better theory of mind capability.24,25
Real-life functioning
The functioning of the participants was measured using the Personal Social Performance scale (PSP).26 It assessed the patients’ functioning based on four dimensions: useful activities, social relationships, self-care and disturbing/aggressive behaviors. The score ranges from 1 to 100. A score of 91-100 indicates excellent function, while 1-10 suggests lack of autonomy in basic function.26 The Thai version of the PSP had an acceptable inter-rater reliability with the intraclass correlation coefficients of 0.75 (p < 0.001).27
Characteristics and clinical data were retrieved from medical records. Psychiatrists interviewed the patients and rated the SAPS, SANS and PSP. Then, the ACE-III, Faces, and RMET tests were administered by independent clinical psychologists on the same day. All measures were paper versions.
Descriptive statistic and partial correlation were analyzed using STATA version 14.0. Network analysis was conducted using R version 4.0.5 (R Foundation for Statistical Computing). 10 variables were selected for the network analysis. The least absolute shrinkage and selection operator network (LASSO) was used as type of network. The centrality measures of the network were also analyzed. The network was visualized using qgraph package.
A network was created comprising variables, which are represented by ‘nodes’ (circle), and the links between the nodes called ‘edges’ (solid line). Thicker edges represent stronger relationships. Blue edges represent positive correlation, and red edges represent negative correlations. The algorithm places strongly associated nodes at the center of the network and weakly associated variables at the periphery.
The centrality indices are the method to examine the relative centrality of constructs within the network. They reveal which is the most important variable in the network. The common centrality measures are ‘strength’, ‘betweenness’ and ‘closeness’. Node strength reflects how strongly a node is directly connected to other nodes. It was determined by the sum weighted associations to other nodes. Betweenness of a node is defined as the number of times that a node is part of the shortest path between two other nodes. The closeness of the node implies how easy it is to reach all other nodes. A high closeness index indicated a short average distance from an interest node to all other nodes. For each measure, higher values indicated more centrality in the network.
The accuracy and stability of the centrality indices of the estimated network were examined using edge weight accuracy and case-dropping bootstrap method on the 500 samples. We utilized recommended cut-off for Coefficient Stability (CS-coefficient, correlation = 0.7) at least 0.25 for considering stable network.28
64 patients with schizophrenia were recruited into the study. Characteristics and descriptive statistics of the network analysis variables are reported in Table 1. The patients had a mean age of 37 (standard deviation (SD) 12.6) years and level of education of 13.3 (SD 3.4) years. The average duration of illness was 8 (SD 9) years. Table 2 presents the partial correlation matrix between 10 network analysis variables. There was no missing data in all variables.
The network structure and centrality measures are demonstrated in Figures 1 and 2. The negative symptoms and cognitive function (neurocognition and social cognition) variables seem to be as two separate communities on upper and lower side of the network. The positive symptoms node appears as an isolate and farthest node from the central.
ACE: Addenbrooke’s Cognitive Examination III; Blunt: blunted affect; Eyes: Reading the Mind in the Eyes test; Faces: Faces test; PSP: Personal and Social Performance scale; SAPG: global Scale for the Assessment of Positive Symptoms score.
ACE: Addenbrooke’s Cognitive Examination III; Asocial: asociality; Blunt: blunted affect; Eyes: Reading the Mind in the Eyes test; Faces: Faces test; PSP: Personal and Social Performance scale; SAPG: global Scale for the Assessment of Positive Symptoms score.
The functioning was found to be the most central node and is displayed in the center of the network structure. It connected and demonstrated inverse correlations to three subdomains of negative symptoms (asociality, avolition and blunt affect) and positive correlations to emotion recognition and neurocognition performance.
In the social cognition domain, only emotion recognition was directly connected to functioning, whereas theory of mind was connected to emotion recognition and neurocognition but not directly connected to functioning. Among cognitive function variables, neurocognition had the highest centrality index and was the strongest connection to functioning.
Among negative symptoms variables, the highest centrality index variables were asociality and avolition subdomains. Except for blunt affect, the other four negative symptoms nodes were interconnected to all other negative symptoms’ nodes.
Between negative symptoms and cognitive function, only alogia and blunt affect showed association with emotion recognition and neurocognition, whereas asociality, avolition and anhedonia showed no direct association to cognitive function.
Figures 3 and 4 show the stability of central indices and accuracy of the estimated network. In brief, the CS-coefficient indicates that strength had the highest and acceptable stability (CS-coefficient 0.28); whereas closeness (CS-coefficient 0.13) and betweenness (CS-coefficient 0.2) are not stable under sub-setting cases.
This study used a network analysis technique to explore the complex relationships among positive symptoms, negative symptoms, neurocognition, social cognition and real-life functioning. To our knowledge, there are few studies examining this association with network approach.29-31
Our results illustrate that real-life functioning is the most central and key role in the network. It has the highest centrality index in strength, betweenness and closeness. It is also connected to negative symptoms, neurocognition, and social cognition. This finding suggests that real-life functioning should be one of the main targets of schizophrenia treatment and research. Rehabilitation for patients with schizophrenia should be provided and should focus on training of everyday real-life skills.32 Besides functioning, avolition and asociality appear to be the second most central and important symptoms. These two negative symptoms also firmly correlate to functioning and connect to other subdomains of negative symptoms. These results confirm the previous findings that, generally, negative symptoms and cognitive symptoms are the strongest predictors of functional outcome in schizophrenia.4,33
The network also shows that neurocognition and emotion recognition are interconnected and close to real-life functioning. Therefore, neurocognition and emotion recognition training could be implemented and would benefit patients with schizophrenia. It is interesting that emotion recognition might directly improve functioning or could have an indirect effect by decreasing blunt affect symptoms. This association is in line with previous studies that blunted affect seems to be closely related to emotion recognition.7,34
Conforming to previous studies, our network analysis confirms that negative and cognitive symptoms are the most important and influential on functioning in clinically stable patients with schizophrenia.31,35 The result underlines the importance of evaluating and treating negative symptoms and cognitive symptoms, as positive symptoms alone have only minimal effect on real-life function. Furthermore, medications that specifically target negative symptoms, social cognition, or neurocognitive symptoms are urgently needed for patients with schizophrenia.
Regarding cognitive function and in line with previous studies on schizophrenia, this study shows that social cognition is highly correlated with neurocognition.7,8 The social cognition node that directly links to functioning is emotion recognition. This connection highlights the important role of emotion recognition in patients’ real-life functioning. The result supports previous studies that found social cognition to be a strong prediction of function outcome in patients with schizophrenia.4,36 On the other hand, though, theory of mind is strongly correlated to neurocognition and emotion recognition; it is not directly connected to functioning and is the most distant cognitive function node.
Compared with the existing network analysis studies on determinants of functioning in schizophrenia, overall, our results are consistent with other studies.29,31 First, these studies support that functioning was found to be the most central node, whereas positive symptoms appeared as the farthest node. Second, our results confirm that negative symptoms, neurocognition and social cognition are important contributors to functioning, and better understood as a separate construct.29,31 Also consistent with Hajduk et al., the RMET was strongly connected to emotion recognition but not directly connected to functioning.31 However, comparing details of the networks between these studies should be interpreted with caution. Each study used significantly different measurements and variables. For example, each study used different functional assessment scales; and our study used five dimensions of negative symptoms, while the others used two and unidimensional approaches.29,31
The strength of this study is that we used network analysis to examine the complex relationship among real-life functioning, clinical symptoms and cognitive functions. In addition, we also used five dimensions of negative symptoms instead of a unidimensional approach in order to avoid losing information relevant to connections with other factors and the ability of each subdomain to predict functioning.37
Our study has some limitations. First, the sample size was relatively modest. Therefore, positive symptoms and neurocognition were constructed as single-measured variables to minimize the number of parameters. Second, closeness and betweenness should be interpreted with caution since their estimates showed less stability. Therefore, a larger sample might be needed to ensure a more stable estimated network. Third, on account of the inclusion of only clinically stable patients with schizophrenia, most had none or very few positive symptoms, therefore, the results might not be able to be generalized to patients with actively psychotic episodes. Lastly, the cross-sectional nature of the data could not allow proper testing of the direction of the causes and effects. Longitudinal study with long term follow-up period would be more appropriate for evaluating the effects of predictors to future functional outcome.
There are restrictions on publicly sharing the dataset, because of the confidential nature of the data and the informed consent given by the study participants, which has been approved by the Human Research Ethics Committee, Faculty of Medicine, Thammasat University. However, thenetwork analysis dataset (ten variables without demographic data) may be requested by contacting the corresponding author (TC) and access will be granted to researchers affiliated with an accredited institution, and reviewers. Full name, title, institution, and purpose for using the dataset should be included in an email.
The author would like to thank Tiraya Lerthattasilp, MD and our psychologists for help in collecting the data.
Views | Downloads | |
---|---|---|
F1000Research | - | - |
PubMed Central
Data from PMC are received and updated monthly.
|
- | - |
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Research on cognitive and social functioning in schizophrenia
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Electroconvulsive therapy, schizophrenia, mood disorders
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Research on cognitive and social functioning in schizophrenia
Is the work clearly and accurately presented and does it cite the current literature?
Partly
Is the study design appropriate and is the work technically sound?
Partly
Are sufficient details of methods and analysis provided to allow replication by others?
Partly
If applicable, is the statistical analysis and its interpretation appropriate?
Partly
Are all the source data underlying the results available to ensure full reproducibility?
Partly
Are the conclusions drawn adequately supported by the results?
Partly
References
1. Hajdúk M, Penn DL, Harvey PD, Pinkham AE: Social cognition, neurocognition, symptomatology, functional competences and outcomes in people with schizophrenia - A network analysis perspective.J Psychiatr Res. 2021; 144: 8-13 PubMed Abstract | Publisher Full TextCompeting Interests: No competing interests were disclosed.
Reviewer Expertise: Research on cognitive and social functioning in schizophrenia
Is the work clearly and accurately presented and does it cite the current literature?
Yes
Is the study design appropriate and is the work technically sound?
Yes
Are sufficient details of methods and analysis provided to allow replication by others?
Yes
If applicable, is the statistical analysis and its interpretation appropriate?
Yes
Are all the source data underlying the results available to ensure full reproducibility?
Yes
Are the conclusions drawn adequately supported by the results?
Yes
Competing Interests: No competing interests were disclosed.
Reviewer Expertise: Electroconvulsive therapy, schizophrenia, mood disorders
Alongside their report, reviewers assign a status to the article:
Invited Reviewers | ||
---|---|---|
1 | 2 | |
Version 3 (revision) 13 Apr 22 |
read | |
Version 2 (revision) 11 Mar 22 |
read | read |
Version 1 08 Dec 21 |
read | read |
Provide sufficient details of any financial or non-financial competing interests to enable users to assess whether your comments might lead a reasonable person to question your impartiality. Consider the following examples, but note that this is not an exhaustive list:
Sign up for content alerts and receive a weekly or monthly email with all newly published articles
Already registered? Sign in
The email address should be the one you originally registered with F1000.
You registered with F1000 via Google, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Google account password, please click here.
You registered with F1000 via Facebook, so we cannot reset your password.
To sign in, please click here.
If you still need help with your Facebook account password, please click here.
If your email address is registered with us, we will email you instructions to reset your password.
If you think you should have received this email but it has not arrived, please check your spam filters and/or contact for further assistance.
Comments on this article Comments (0)