We created an online survey to assess attitudes towards the network theory of mental disorders. The target respondents were mental health professionals, including clinical psychologists (persons with a university degree in psychology without additional training, but working in the field of mental health), psychotherapists (either graduated or in training), and physicians specialized in psychiatry and psychotherapy, psychosomatic medicine, neurology, or child and adolescent psychiatry (either graduated or in training). In Germany, mental health treatments are mostly provided by these professions. All three professions require a university degree in psychology or medicine, and psychotherapists and specialized physicians additionally receive extensive training in psychotherapy and/or their respective field of specialization.

The survey was made available online in English and in German using LimeSurvey (LimeSurvey GmbH, 2024). It was distributed by reaching out to 111 international mental health related associations (e.g. all national organizations of the International Association for Analytical Psychology and of the European Association for Behavioural and Cognitive Therapies), to 137 German mental health related associations (e.g., all regional chambers of psychotherapists and regional and national psychotherapeutic and psychiatric organizations), and to 77 large European psychiatry clinics. Further, the survey was shared on social media (X, LinkedIn, Instagram) and by reaching out to colleagues. Previous knowledge of the network theory was not required and participants received no incentives.

Before proceeding with the survey, it was pointed out that participation is voluntary and the survey would take a maximum of 20 minutes. All participants had to consent to the privacy statement (only personally non-identifiable data were collected) and to take part in this study. The data collection started in October 2023, and the last included response was recorded in June 2024. We aimed to achieve a sample size of 250 respondents, as simulations indicated that sufficient precision for descriptive analyses can be achieved with this sample size. The study design was preregistered on the open science framework: https://osf.io/g7zy3. We did not specify specific hypotheses. The study was approved by the local psychological ethics committee of the centre of psychological medicine at the University Medical Centre Hamburg Eppendorf (LPEK-0607).

The questionnaire was based on previously published presentations of the network theory by Borsboom and Cramer (Borsboom, 2017; Borsboom & Cramer, 2013). It was first developed in English and then translated to German with the help of ChatGPT (OpenAI, 2023). The first version was tested and adjusted for better comprehensibility with the help of two psychotherapists. Several clinical researchers assessed the final implemented version for mistakes and intelligibility.

The survey included 48 questions in total. Eight questions assessed demographic variables. Subsequently, a description of the network theory of mental disorders was shown. It was described how the network theory explains the development and persistence of psychological symptoms and what consequences this has for the treatment of mental disorders. The network approach was contrasted with the disease model of psychopathology, which states that symptoms share a common underlying disorder. Both models were described using the example of depression and depicted with corresponding diagrams (see Supplementary Materials [Schumacher & Kriston, 2025S] for the full description). Participants were asked if they knew about the network theory and if they had knowingly used a treatment planning tool or an intervention based on the network approach before. Then, participants were asked how strongly they agree or disagree with specific propositions of this theory. These propositions related to three domains: the phenomenology, the aetiology, and the treatment of mental disorders. All items referred to mental disorders or symptoms in general and not to a specific mental health problem or disorder. One statement related to the overall agreement with the network theory and four statements relating to the disease model were also incorporated.

Additionally, we were interested in the acceptability and appropriateness of possible applications of the network approach to clinical practice. Acceptability and appropriateness were defined as important outcomes for successful implementation of new practices (Proctor et al., 2011). We asked for the level of agreement with the acceptability and appropriateness of specific applications, mainly relating to the use of symptom networks for treatment planning and outcome monitoring. Wording was inspired by a validated measure of acceptability (“I would like to use...”) and appropriateness (“It seems fitting to use…”, Weiner et al., 2017). Finally, we assessed the extent that mental health professionals already implicitly implemented the network approach in their practice, i.e. if they previously considered symptom interactions in their practice. The level of agreement for each proposition could be answered on a seven-point Likert scale from “strongly disagree” (-3) to “strongly agree” (3).

Participants could also indicate if they would like to use tools or techniques based on the network approach in any other way in their practice and give any other comment in two open text questions. Two questions were included to test for automated responding. The complete questionnaire can be found in the Supplementary Materials.

To evaluate if mean scores for the agreement in each domain can be calculated, we assessed the measurement properties of the questionnaire. More specifically, we used confirmatory factor analyses to assess the fit of one-factor models for measuring agreement with the network theory regarding the phenomenology, aetiology, and treatment of mental disorders and the acceptability and appropriateness of possible applications. An acceptable to good fit was shown for measuring agreement with the network theory’s propositions on the phenomenology and aetiology of mental disorders and for measuring acceptability and appropriateness. The one-factor model for agreement with the network theory’s propositions regarding the treatment of mental disorders showed poor fit. Here, a two-factor model showed good fit. One factor indicated agreement with the network theory regarding treatment effects, e.g. that effective treatment leads to a reduction in symptom interactions, and the other factor measured agreement with treatment targets as proposed by the theory, e.g. that treatment should target individual symptoms and their interaction. Thus, a mean score was computed for agreement with the network theory’s propositions regarding the phenomenology and aetiology of mental disorders, treatment targets, treatment effects, and for the acceptability and appropriateness of possible applications. Previous considerations of symptom interactions in practice can be interpreted as formative indicators of a composite latent variable, thus, a mean score was also computed for this domain. The mean score was also computed when the response was missing for one item of the respective domain. The complete psychometric evaluation of the questionnaire is reported in the Supplementary Materials.

Data from all respondents that answered at least one question regarding the network theory of mental disorders were included. Data were excluded if the respondent indicated no mental health profession or answered both attentiveness checks wrong. The main outcome was descriptive statistics for the agreement with propositions of the network theory and the acceptability and appropriateness of possible applications. For each statement, we evaluated the level of agreement using mean scores, standard deviations, and ranges. Further, we used Bayesian linear regression analyses to assess which variables were associated with agreement in each domain of the network theory, the acceptability, appropriateness of its application, and previous considerations of symptom interactions in practice. First, age, being in training, mental health profession, work experience, and theoretical background were entered as predictors. Second, we added previous knowledge of the network theory and previous use of a treatment planning tool or intervention based on the network approach as predictors. We used Markow-Chain-Monte-Carlo sampling with weakly informative priors and 40,000 iterations to estimate the regression model with brms (Bürkner, 2017) in R 4.3.3 (R Core Team, 2013). Data is openly available (https://osf.io/zan2q/files).

While the online survey was accessed 356 times, 265 persons answered at least one question regarding the network theory. Data was excluded for five persons who answered both attentiveness checks wrong, for one person who answered “neither agree nor disagree” for all questions, and for two persons who indicated that they only work as scientists. Demographic information of all 257 included participants is displayed in Table 1.

Most participants were female, between 31 and 50 years old, and from Germany. About one third of the sample was currently in training, and of these, the majority had completed at least 60% of their training. Nearly three-quarters of the sample were psychotherapists, 17.5% specialized physicians (specialization in child and adolescent psychiatry, psychiatry, psychosomatic medicine, and/or neurology) and 11.3% clinical psychologists. Nearly 70% of the mental health professionals indicated a background in cognitive behavioural therapy, 32.3% in psychodynamic or psychoanalytic treatment, and about 20% indicated a different theoretical background (48 participants indicated multiple theoretical backgrounds). The mental health professionals had a mean of 16 years of work experience, ranging from one to 50 years.

The majority of the sample did not hear about the network theory before (70.8%), never used this approach for treatment planning (93.0%) and never knowingly administered an interventions based on the network approach (84.8%). The majority of the participants filled out the survey in German (93.0%) and the mean completion time was 15.8 minutes. All questions related to the phenomenology, aetiology, and treatment of mental disorders were answered completely by 245 participants, and 236 participants also answered all questions regarding the acceptability and appropriateness of possible applications, i.e., finished the complete survey.

We found consistent agreement with propositions of the network theory regarding the phenomenology of mental disorders (Figure 1). The vast majority of participants rated all network theory’s propositions on the phenomenology at least moderately positively. The mean score for the agreement in this domain was 1.4, SD = 0.8 (scale ranges between -3 indicating strong disagreement, to 3 indicating strong agreement). Propositions of the disease model, this means the statements that most symptoms do not strongly impact each other and that most symptoms have the same underlying cause were, on average, evaluated negatively, M = -1.7, SD = 1.0 and M = -0.1, SD = 1.4. Participants strongly differed in their ratings of the network theory’s propositions regarding the aetiology of mental disorders, with a mean level of 0.7, SD = 1.1 (Figure 2). Still, we found that participants rated the proposition that strong symptom interactions pose a risk for the development of mental disorders moderately positively, M = 1.1, SD = 1.1.

Attitudes towards propositions of network theory regarding the treatment of mental disorders differed between two aspects of treatment (Figure 3). The evaluation of statements proposing that treatment targets should be selected according to the network theory was very mixed, M = 0.5, SD = 1.2. On the other hand, we found strong support for propositions regarding treatment effects, i.e., that treatment changes the activation of specific symptoms and symptom interactions, M = 1.5, SD = 0.8. Finally, we found strong agreement for the disease model’s proposition that treatment should target the underlying cause of the mental disorder and that effective treatment leads to global improvements across all symptoms, M = 1.5, SD = 1.1 and M = 1.41, SD = 1.09.

Agreement with the network theory’s propositions regarding the phenomenology and aetiology of mental disorders, treatment targets, and treatment effects correlated moderately with each other (see Table S1 in the Supplementary Materials).

Detailed results of the Bayesian linear regression analyses are presented in Table S2 in the Supplementary Materials. Age, gender, being in training, work experience, mental health profession, and theoretical background could explain agreement to the network theory’s propositions regarding the phenomenology, aetiology, treatment targets and treatment effects only to a limited extend as indicated by the Bayesian R² ranging between .01 and .28. Still, a background in cognitive behavioural therapy was related to a more positive evaluation of the network theory regarding the phenomenology of mental disorders and treatment targets, b = 0.40 [95% credibility interval: 0.01 – 0.79] and b = 0.69 [0.20 – 1.18]. Having heard about the network theory and having used an intervention based on the network approach was related to stronger agreement with the network approach regarding the phenomenology of mental disorders, b = 0.36 [0.13, 0.60] and b = 0.43 [0.10, 0.76], and regarding important treatment targets, b = 0.33 [0.03, 0.64] and b = 0.55 [0.11, 0.98].

We found that applications of the network approach to clinical practice were rated, on average, somewhat appropriate and slightly less acceptable, M = 1.2, SD = 1.1 and M = 0.9, SD = 1.3, respectively (see Figure S1 and S2 in the Supplementary Materials). However, especially the evaluation of acceptability varied to a large extend between the participants. Further, most participants considered symptom interactions in their clinical practice in general, M = 1.1, SD = 1.3, but few have used the symptom interactions to plan or monitor treatment, M = 0.1, SD = 1.7, and M = -0.1, SD = 1.7, respectively. While previous consideration of symptom interactions correlated moderately with the other domains, the correlation of appropriateness and acceptability with agreement in the other domains was rather high (see Table S1 in the Supplementary Materials).

Detailed results of the Bayesian linear regression analyses on associated factors for the acceptability and appropriateness of possible applications and the previous consideration of symptom interactions are displayed in Table S3 in the Supplementary Materials. Age, gender, being in training, work experience, mental health profession, and the theoretical background of the participants could explain 28.2 and 26.4% of the variance in the acceptability and appropriateness ratings of possible applications. Being a physician (in comparison to psychotherapists) and having a background in cognitive behavioural therapy was associated to a stronger agreement with the acceptability and appropriateness of possible applications of the network approach. Being between 31 years and 50 years old (compared to being 18 to 30 years old) and having a background in psychodynamic or psychoanalytic therapy was related to lower ratings of the acceptability and appropriateness. Additionally, we found that physicians (in comparison to psychotherapists) and participants with a background in cognitive behavioural therapy more strongly agreed to have considered symptom interactions in their work before.

Thirty-six participants (14.0%) provided additional comments on how the network approach could be used in clinical practice and 66 participants (25.7%) provided general additional comments. Most comments indicated a negative evaluation of the network theory. First, it was emphasized that the theory does not provide new insights, because considering symptom associations is already an established clinical practice. Second, it was commented that the focus on symptoms and their interactions is too superficial and disregards the underlying (biological) cause. Some participants also commented positively that the network approach provides a useful framework for an intuitive clinical practice.

The network approach could potentially provide a framework for systematically considering symptom interactions in the diagnosis and treatment of mental disorders (von Klipstein et al., 2020). Person-specific symptom networks could provide a standardized basis for case conceptualization (Burger, Epskamp, et al., 2022; Vogel et al., 2025; von Klipstein et al., 2020) and might be used for psychoeducation and selection of treatment targets (Levinson et al., 2022; Meier et al., 2022). Further, the network approach is reflected in calls for process-based therapy, which focuses on targeting symptom maintaining processes (Hofmann & Hayes, 2019). However, similar to previous studies, we found rather mixed agreement with the acceptability and appropriateness of possible applications of the network approach. Methodological challenges for person-specific networks based on empirical symptom data remain, and guidelines for network construction are only slowly emerging (Mansueto et al., 2023; Schumacher et al., 2024; Vogel et al., 2025). Given the currently available methods, the interpretation of person-specific networks should be seen as mainly hypothesis generating (von Klipstein et al., 2020). It also remains difficult to derive direct treatment applications from a person-specific symptom network (Levinson et al., 2021). Given the juvenile state of current applications, mixed findings in regard to mental health professionals ratings of the acceptability and appropriateness of possible applications are not surprising. Before applications of the network theory should be established, more knowledge on the theory itself and on the specific methodologies to derive symptom networks is indispensable.

Our results might not generalize to mental health professionals working in other countries, as the majority of the sample was living in Germany. Second, due to privacy reasons, we did not collect data on racial/ethnic identification, measures of income, and socioeconomic status; thus, the impact of these variables could not be assessed. Similarly, we did not assess which problems were typically treated by the included mental health professionals. Although all survey items related to mental disorders or symptoms in general, participants might have considered a specific disorder when answering the questions, possibly influencing their answers. Finally, mental health professionals rated the acceptability and appropriateness of hypothetical applications of the approach. Ratings might have differed for actual implementations, i.e., if mental health professionals would have actually used tools based on the network approach in their practice.

Future research needs to further assess the validity of the network theory of mental disorders. On the one hand, studies might analyse empirical symptom data, for example assessing how certain network structures relate to mental health (problems). Hereby, methodological challenges for network analyses and a match between theoretical assumptions and the statistical model should be considered. On the other hand, existing clinical expertise could be further considered, for example by conducting interviews with experienced mental health professionals. Existing clinical theories might be mapped onto the network conceptualization. It needs to be considered to what extent a thorough understanding of the network theory is needed to evaluate its validity and usefulness for clinical practice and if the network approach is applicable to all kinds of different mental health problems. Combining various sources of information, i.e. empirical symptom data and clinical expertise, will hopefully lead to a better understanding of mental disorders.

We showed that mental health professionals agree with central claims of the network theory, however, most of them were hesitant in regard to the applications of this approach. Mixed results for the appropriateness and acceptability of possible applications in clinical practice might reflect the early state of these applications. The theory itself, methods for network construction, and the consequences of the network theory for treatment planning and evaluation must be better understood before this approach can be responsibly applied to clinical practice. Still, the conceptualization of mental disorders as symptom networks seems to resonate with mental health professionals and might be an important component of a potential new paradigm for the diagnosis and treatment of psychopathology.