{"paper_id":"37202d67-a282-4a2b-9c66-43483a3f824d","body_text":"1 \n \nDisentangling heterogeneity of psychosis expression in the general population: \nsex-specific moderation effects of environmental risk factors on symptom \nnetworks \n \nLinda T. Betz1, Nora Penzel1,2, Marlene Rosen1, Kamaldeep Bhui3,4, Rachel Upthegrove5,6, \nJoseph Kambeitz1 \n \n1 Department of Psychiatry and Psychotherapy, Faculty of Medicine and University Hospital \nof Cologne, University of Cologne, Cologne, Germany \n2 Department of Psychiatry and Psychotherapy, Ludwig-Maximilian-University, Munich, \nGermany \n3 Department of Psychiatry, University of Oxford, Oxford, United Kingdom \n4 Nuffield Department of Primary Care Health Sciences, University of Oxford, United \nKingdom \n5 Institute for Mental Health and Centre for Human Brain Health, University of Birmingham, \nBirmingham, United Kingdom \n6 Birmingham Early Intervention Service, Birmingham Women’s and Children’s NHS \nFoundation Trust, Birmingham, United Kingdom \n \nCorresponding Author:  \nLinda T. Betz, Department of Psychiatry and Psychotherapy, Faculty of Medicine and \nUniversity Hospital of Cologne, Kerpener Str. 62, 50937 Cologne. E-mail: linda.betz@uk-\nkoeln.de, Phone: +49 (0)221 – 478 7175. \n \nWord Count: 4484 (text only, excluding abstract, references, tables/figures) \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \nNOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.\n\n2 \n \nAbstract \nBackground: Psychosis expression in the general population may reflect a behavioral \nmanifestation of the risk for psychotic disorder. It can be conceptualized as an interconnected \nsystem of psychotic and affective experiences; a so-called ‘symptom network’. Differences in \ndemographics, as well as exposure to adversities and risk factors, may produce substantial \nheterogeneity in symptom networks, highlighting potential etiological divergence in \npsychosis risk.  \nMethods: To explore this idea in a data-driven way, we employed a novel recursive \npartitioning approach in the 2007 English National Survey of Psychiatric Morbidity survey \n(N = 7,242). We sought to identify ‘network phenotypes’ by explaining heterogeneity in \nsymptom networks through potential moderators, including age, sex, ethnicity, deprivation, \nchildhood abuse, separation from parents, bullying, domestic violence, cannabis use, and \nalcohol. \nResults: Sex was the primary source of heterogeneity in symptom networks. Additional \nheterogeneity was explained by interpersonal trauma (childhood abuse, domestic violence) in \nwomen and domestic violence, cannabis use, ethnicity in men. Among women, especially \nthose exposed to early interpersonal trauma, an affective loading within psychosis may have \ndistinct relevance. Men, particularly those from minority ethnic groups, demonstrated a \nstrong network connection between hallucinatory experiences and persecutory ideation. \nConclusion: Symptom networks of psychosis expression in the general population are highly \nheterogeneous. The structure of symptom networks seems to reflect distinct sex-related \nadversities, etiologies, and mechanisms of symptom-expression. Disentangling the complex \ninterplay of sex, minority ethnic group status, and other risk factors may help optimize early \nintervention and prevention strategies in psychosis. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n3 \n \nIntroduction \nRecent research has advanced our understanding of psychosis through so-called \n‘symptom networks’, i.e., causal systems of individual interacting experiences and symptoms \n(Betz et al., 2020; Hardy, O’Driscoll, Steel, van der Gaag, & van den Berg, 2020; Isvoranu, \nBorsboom, van Os, & Guloksuz, 2016; Isvoranu et al., 2019, 2017; Moffa et al., 2017; \nMurphy, McBride, Fried, & Shevlin, 2018; Robinaugh, Hoekstra, Toner, & Borsboom, \n2020). Complex interactions between specific psychotic as well as non-psychotic experiences \n(e.g., depression, anxiety) in the general population may predate onset of psychosis in clinical \nsettings (Guloksuz et al., 2016, 2015; Kelleher et al., 2012; Linscott & van Os, 2013; Murphy \net al., 2018; van Os & Reininghaus, 2016). Additional lines of evidence indicate that there is \nconsiderable etiological continuity between subclinical and clinical levels of psychosis \n(Binbay et al., 2012; DeRosse & Karlsgodt, 2015; Kelleher & Cannon, 2011; Linscott & van \nOs, 2013). Thus, examining the symptom network structure of a transdiagnostic psychosis \nphenotype, reflecting a behavioral manifestation of risk for psychotic disorder in the general \npopulation that blends gradually into clinical syndromes, may help to better understand \netiological mechanisms in psychosis and to develop prevention strategies (Bebbington, 2015; \nBinbay et al., 2012; DeRosse & Karlsgodt, 2015; Isvoranu et al., 2016; Kelleher & Cannon, \n2011; Linscott & van Os, 2013; Robinaugh et al., 2020; van Os & Reininghaus, 2016). \nImportantly, symptomatology and involved etiological mechanisms in psychosis \nexpression are highly variable by specific at risk groups (Bentall, Wickham, Shevlin, & \nVarese, 2012; Isvoranu et al., 2016; Linscott & van Os, 2013; van Os & Reininghaus, 2016). \nFor example, in line with the theory of an affective pathway to psychosis, early traumatic \nevents are strongly associated with connections between affective and psychotic \nsymptomatology (Myin-Germeys & van Os, 2007; Upthegrove et al., 2015; van Nierop et al., \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n4 \n \n2015). In the presence of heterogeneity, averaged network models of psychosis may obscure \nimportant distinctions in relevant etiological mechanisms across specific risk groups (Jones, \nMair, Simon, & Zeileis, 2020; Moriarity, van Borkulo, & Alloy, 2020). Thus far, however, \nheterogeneity in symptom networks of psychosis has been either overlooked or addressed in a \npartial way on a single candidate risk factor (such as sex, cannabis use or socioeconomic \nbackground) at specific thresholds, or using summed environmental risk scores (Betz et al., \n2020; Guloksuz et al., 2016; Isvoranu et al., 2016; Wüsten et al., 2018), which lose specificity \nand relevance for real work prevention and intervention. \nThe characterization of ‘network phenotypes’ based on a comprehensive set of \nenvironmental and demographic factors may explain heterogeneity; that is the structure of \nsymptomatology is a function of types, combinations, and intensity of etiological loads in \npsychosis expression (Jones et al., 2020; Moriarity et al., 2020). With the goal of \ncharacterization of network phenotypes in mind, the current study uses novel work on \nrecursive partitioning, a data-driven, explorative statistical technique that can sequentially \nextract isolated and combined moderation effects of a large set of environmental and \ndemographic factors on symptom networks, without a priori specification of thresholds or \ncombinations of risk factors (Jones et al., 2020; Strobl, Malley, & Tutz, 2009). Recursive \npartitioning identifies network phenotypes that are maximally distinct from each other (Jones \net al., 2020; Zeileis, Hothorn, & Hornik, 2008). \nWe used recursive partitioning to define meaningful network phenotypes of psychosis \nexpression in the general population, using the 2007 Adult Psychiatric Morbidity in England \nSurvey (APMS; (National Centre for Social Research, University of Leicester, 2017). We \nhypothesized that exposure to environmental risk, if identified as defining a network \nphenotype, would be characteristically associated with more densely connected symptom \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n5 \n \nnetworks when compared with samples not exposed to that specific environmental risk \n(Guloksuz et al., 2016, 2015; Isvoranu et al., 2016; Lin, Fried, & Eaton, 2019; Russell, \nKeding, He, Li, & Herringa, 2020). We also aimed to test whether the strength of connections \nbetween individual symptoms differed between network phenotypes. \nMethod \nData analytic strategy \nWe conducted all analyses in the R language for statistical computing, version 4.0.4. \nThroughout, we considered a significance level of α  = .05. Data of the 2007 APMS (National \nCentre for Social Research, University of Leicester, 2017) used in the analyses are available \nfrom the UK Data Service (https://ukdataservice.ac.uk/). Code to reproduce the analyses can \nbe accessed at www.github.com/LindaBetz/APMS_NetworkTree. \nSample \nWe present analyses based on the 2007 APMS of adults living in private households \naged 16 and above who were recruited using a stratified multistage random probability \nsampling strategy (N = 7,403) (McManus, Meltzer, Brugha, Bebbington, & Jenkins, 2009; \nSingleton, Bumpstead, O’Brien, Lee, & Meltzer, 2003). Methods, procedures, and full details \non sample characteristics have been described previously (McManus et al., 2009). For the \npresent analyses, we excluded participants with missing values in the variables of interest, \ngiven that the methods employed do not allow missings. For comparing sample \ncharacteristics of included and excluded participants, we used permutation tests as \nimplemented in the R package ‘coin’ (Hothorn, Hornik, van de Wiel, & Zeileis, 2008). \nAssessment of symptomatology \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n6 \n \nSelection and definition of symptom variables followed a previously published \nnetwork analysis using data from the 2007 APMS (Moffa et al., 2017), including measures \nfrom an affective domain (worry, sleep disturbance, generalized anxiety, and depression), and \nfrom a psychotic domain (persecutory ideation and hallucinatory experiences). All symptom \nvariables in the network were coded in binary form (present or absent). For details on these \nassessments, see supplementary method 1. \nAssessment of environmental and demographic risk factors \nEnvironmental risk factors comprised psychosocial adversities in the form of physical \nabuse and sexual abuse before the age of 16, separation from parents until the age of 16 (local \nauthority care and/or institutional care), lifetime experiences of bullying, and lifetime \nexperiences of domestic violence. Additionally, we included sex, age, ethnic origin (White, \nBlack, South Asian, Mixed/Other), cannabis use in the past year, alcohol use, and \nsocioeconomic deprivation. For details on these assessments, see supplementary method 2. \nIdentification of network subgroups via recursive partitioning \nIn a first step, we estimated a partial correlation network (without regularization) \nbased on the full sample, using the R package ‘qgraph’, version 1.6.5 (Epskamp, Cramer, \nWaldorp, Schmittmann, & Borsboom, 2012). A partial correlation network depicts unique \npairwise associations between variables (‘edges’ in network terminology), i.e., the share of \nthe association between two variables that remains after controlling for all other variables in \nthe network (Epskamp, Borsboom, & Fried, 2018). We estimated the underlying zero-order \ncorrelations between the binary items using Pearson’s \nφ , as recommended when employing \nrecursive partitioning on binary data (Jones et al., 2020). The stronger the partial correlation \nbetween two variables, the more likely it is that they co-occur, controlling for the other \nvariables under consideration. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n7 \n \nSecond, we used a model-based recursive partitioning approach to identify \nmeaningful subgroups of symptom networks given the included environmental and \ndemographic factors, as implemented in the R package ‘networktree’, version 1.0.1 (Jones et \nal., 2020). In brief, recursive partitioning sequentially creates a decision tree by either \nsplitting or not splitting the sample along a set of potential moderating variables (Strobl et al., \n2009; Zeileis et al., 2008). The ‘networktree’ approach (figure 1) determines sample splits \nbased on significant invariance in the correlation matrix of the network variables under \nconsideration, yielding non-overlapping partitions of the sample with maximally \nheterogeneous symptom networks (Jones et al., 2020). For a detailed account, we refer to \nsupplementary method 3 and available methodological articles (Jones et al., 2020; Strobl et \nal., 2009; Zeileis et al., 2008). For plotting, we transformed the correlation matrices to partial \ncorrelation matrices using the R package ‘qgraph’, such that edges reflect unique associations \nbetween two variables. \nComparison of identified subgroups \nTo delineate specific network differences between the identified subgroups (i.e., \ndifferences between subgroups as defined by a splitting factor in the recursive partitioning \napproach), we compared the overall strength of symptom connections, defined as the absolute \nsum of all individual partial correlation coefficients in the network (global strength; S), and \ndifferences in estimates of individual partial correlation coefficients (individual edge weights; \nρ ) within a Bayesian framework, using the R package ‘BGGM’, version 2.0.2 (Williams, \n2021; Williams & Mulder, 2019; Williams, Rast, Pericchi, & Mulder, 2020). Specifically, we \nused posterior predictive checks for assessing differences in overall connection strength \n(Williams et al., 2020), and evaluated the posterior distribution for each difference in partial \ncorrelation coefficients, where we deemed a difference significant if the 95% credible \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n8 \n \ninterval did not contain 0 (Williams, 2021). P-values derived from recursive partitioning are \ndenoted as pRP, while p-values derived from post-hoc comparisons implemented in the \npackage ‘BGGM’ are denoted as pBGGM. \n                  --- Figure 1 --- \nRobustness Analyses \nWe used the R package ‘bootnet’, version 1.4.3 (Epskamp et al., 2018) to conduct \nrobustness analyses to check stability and accuracy of the results. We investigated stability of \nsymptom networks estimated in the full sample and identified subgroups by testing sensitivity \nto dropping cases. Specifically, we assessed the degree to which edge weights remained the \nsame after re-estimating the networks with less cases via the correlation stability (CS) \ncoefficient. The CS coefficient represents the maximum proportion of cases that can be \ndropped, such that the correlation between original edge weights and edge weights of \nnetworks based on subsets is 0.7 or higher (95% confidence). The CS-coefficient should \npreferably be above 0.5 (good stability), and not be below 0.25 (acceptable stability) \n(Epskamp et al., 2018). To investigate the accuracy of individual edge weights estimates \nacross the networks in the full sample and identified subgroups, participants were randomly \nresampled 5000 times, and the bootstrapped confidence intervals (CIs) of the edge weights \nwere estimated. \nResults \nSample \nFollowing removal of 161 participants (2.2% of the whole sample) with missing \nvalues in the variables of interest, the final sample comprised 7,242 participants, 56.8% of \nwhom were women, with an average age of 50 (IQR = 30) years. Participants excluded due to \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n9 \n \nmissing data were on average older, less White and reported lower proportions of alcohol use \nand hallucinatory experiences, and higher proportions of depressive symptoms \n(supplementary table 1). \nNetwork variables and potential moderators \n         Table 1 presents positive endorsement and characteristics of the network variables \nand potential moderating risk factors in the sample. The most prevalent symptom was worry, \nand the most prevalent risk factor bullying. \nOverall symptom network structure and subgroups \nThe partial correlation network estimated in the full sample suggested positive \nrelationships between all symptoms, with a mean edge weight of 0.11. Partial correlations \nwithin each symptom domain were, on average, stronger than between the domains. \nRecursive partitioning revealed that six of the tested demographic and environmental risk \nfactors were linked to significant heterogeneity in symptom networks and split the sample \naccordingly in a hierarchical fashion: sex, childhood sexual abuse, childhood physical abuse, \ndomestic violence, cannabis use, ethnicity. Partial correlation matrices for the plotted \nnetworks are available at the linked GitHub repository. Sex was the primary source of \nheterogeneity (figure 2a, pRP < .001): the network of women was overall significantly less \nstrongly connected (Δ S = -0.17, pBGGM = .002), and featured a significantly stronger connection \nbetween depression and hallucination (Δρ  = 0.06), and a significantly weaker connection \nbetween sleep problems and persecutory ideation (Δρ  = -0.07) than the network of men. This \nmeans that in women, depression and hallucination were more likely to co-occur than in men, \nwhile sleep problems and persecutory ideation were less likely to co-occur than in men. For \nnetworks of women and men, see figure 3. \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n10 \n \n Distinct risk factors explained further heterogeneity in symptom networks of women \nand men, yielding 8 different network phenotypes in total. Among women, experiences of \nchildhood sexual abuse were the major source of heterogeneity in symptom networks (figure \n2b, pRP = .016) linked to a stronger connection between anxiety and persecutory ideation (Δρ  \n= 0.09). The difference in global strength of the symptom networks of women who reported \nsexual abuse and those who did not was not significant (Δ S = 0.08, pBGGM = .482). Among \nwomen who reported no childhood sexual abuse, exposure to childhood physical abuse \nexplained further heterogeneity (figure 2c, pRP = .015), and was associated with a significantly \nstronger association between anxiety and hallucinations (Δρ  = 0.26). Corresponding symptom \nnetworks did not differ significantly in global strength (Δ S = 0.70, pBGGM = .204). Finally, \namong those women that reported neither sexual nor physical abuse, exposure to domestic \nviolence (figure 2d, pRP = .012) was linked to a stronger connection between worry and \ndepression (Δρ  = 0.12), as well as persecutory ideation and hallucinations (Δρ  = 0.19). The \ndifference in global strength of the corresponding symptom networks was not significant (Δ S \n= 0.32, pBGGM = .113). \nAmong men, in those who reported having experienced domestic violence (figure 2e, \npRP = .007) the connection between sleep problems and anxiety was significantly stronger than \nin men who did not report past domestic violence (Δρ  = 0.17). Global strength was not \nsignificantly different between the corresponding networks (Δ S = 0.26, pBGGM = .376). Second, \nin men not reporting past domestic violence, cannabis use in the past year (figure 2f, pRP = \n.009) was associated with a significantly increased connection between worry and \npersecutory ideation (Δρ  = 0.17), and a significantly weaker connection between \nhallucination and persecutory ideation (Δρ  = -0.18). Global strength of the corresponding \nsymptom networks did not differ significantly (Δ S = 0.33, pBGGM = .126). Finally, men reporting \nneither domestic violence nor cannabis use were further split by ethnic background (figure \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n11 \n \n2g, pRP = .011): the network of men with a Black or South Asian ethnic background was \noverall significantly more strongly connected (Δ S = 0.76, pBGGM = .003), and showed stronger \nconnections between worry and depression (Δρ  = 0.25), sleep problems and anxiety (Δρ  = \n0.20), anxiety and depression (Δρ  = 0.16), depression and persecutory ideation (Δρ  = 0.21), \nas well as persecutory ideation and hallucinatory experiences (Δρ  = 0.20), and a weaker \nconnection between sleep problems and depression (Δρ  = -0.23) than the network of men \nfrom a White or Mixed ethnic background. \nAge of the respondent, alcohol use, bullying, separation experiences, and \nsocioeconomic deprivation were not identified as relevant sources of heterogeneity in \nsymptom networks. Repeating analyses based on data from women and men separately \nyielded identical results regarding sex-specific moderators (supplementary figure 1). \nRobustness analyses \nThe network estimated in the full sample, as well as all identified subgroup networks, \nshowed good stability to dropping cases (supplementary table 2). Accuracy analyses showed \nsome relatively wide bootstrapped CIs in some of the identified subgroups with smaller \nsample sizes. In these cases, we recommend caution when interpreting the strength of weaker \nedges. Still, the bootstrap mean was generally very close to the sample mean, indicating \ninterpretable results (supplementary figures 2-16). \n  \n                                                    --- Figure 2 --- \n  \n                                                   --- Figure 3 --- \n  \nDiscussion \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n12 \n \nIn the present study, we employed a novel, data-driven recursive partitioning \napproach in a large national household survey to identify networks of psychotic and affective \nexperiences in the population. Our findings point to considerable heterogeneity, which we \nexplain with several phenotypic systems: six (out of eleven) demographic and environmental \nrisk factors yielded eight different network phenotypes, with sex being the primary source of \nheterogeneity in symptom networks. Among women and men, different risk factors were \nrelated to heterogeneity in symptom networks, suggesting potentially distinct relevance and \nmechanisms of these risk factors across the sexes, in line with a multidimensional model of \nsexual differentiation in psychosis risk (Riecher-Rössler, Butler, & Kulkarni, 2018). Overall, \nour findings on sex and other environmental differences illustrate that the multifactorial and \nheterogeneous nature of psychosis expression (Isvoranu et al., 2016; Linscott & van Os, \n2013; van Os & Reininghaus, 2016) appears to be reflected in symptom networks that \ndiffered considerably depending on the type, combination and strength of demographic and \nenvironmental risk in a large general population sample. \nDifferences in symptom networks of women and men \nThe identification of multiple network phenotypes substantiates the notion that \naveraged symptom network models are likely not representative of psychosis expression in \nthe general population (Jones et al., 2020). Rather, observed differences in strength of overall \nand specific symptom connections may point to diverse etiological mechanisms operating \nacross different demographic and environmental risk factors. Corroborating a growing \nrecognition that understanding variability by sex is central for the development of \ncomprehensive etiological models of psychopathology (Hartung & Lefler, 2019; Hodes & \nEpperson, 2019; Riecher-Rössler et al., 2018; Rosen, Haidl, Ruhrmann, Vogeley, & Schultze-\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n13 \n \nLutter, 2019), the primary source of heterogeneity in symptom networks of psychosis was \nsex. \nSpecifically, our results highlight how associations between affective and psychotic \nexperiences may be differentially expressed in women and men. Prior research indicates that, \nfollowing the theory of an affective pathway to psychosis, affective alterations, in particular \ndepression and anxiety, may be fundamental driving forces of psychotic experiences (Betz et \nal., 2020; Isvoranu et al., 2017; Myin-Germeys & van Os, 2007; Upthegrove et al., 2020; \nUpthegrove, Marwaha, & Birchwood, 2017; van Nierop et al., 2018). Present findings \nsuggest a particularly strong association between depression and hallucinatory experiences in \nthe network of women compared to men, corroborating the idea that such an affective \npathway to psychosis involving depression may be expressed to a greater degree in women, \npotentially funneled by increased emotional reactivity to life events and daily hassles (Davis, \nMatthews, & Twamley, 1999; Hodes & Epperson, 2019; Myin-Germeys & van Os, 2007; \nStainton et al., 2021). In the symptom network of men, by contrast, a previously identified \nlink between sleep problems and persecutory ideation (Freeman et al., 2010) was stronger, \nand therefore, possibly more relevant, than in women. An intriguing potential clinical \nimplication to be tested is that men may, on average, profit in particular from the use of \ninterventions for sleep problems with demonstrated benefit for reducing persecutory ideation \n(Freeman et al., 2017). \nRisk factors explaining heterogeneity in symptom networks of women and men \nAmong women, heterogeneity in symptom networks of psychosis expression was \nexplained by exposure to interpersonal trauma, including childhood abuse and domestic \nviolence. Specifically, exposure to childhood abuse was linked to stronger associations \nbetween anxiety and psychotic experiences. These findings are consistent with previous \nreports of increased proportions of mixed symptom expression following childhood trauma \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n14 \n \n(Guloksuz et al., 2015; Russell et al., 2020; Upthegrove et al., 2015; van Nierop et al., 2015), \nbut extend the literature by highlighting how sex may be an important determinant in this \nrelationship. Following trauma, women are more likely to blame themselves, to view the \nworld as dangerous, and to hold more negative views of themselves (Davis et al., 1999; Tolin \n& Foa, 2002). This may facilitate a pathway from distressing interpretations of everyday \nevents, including the experience of anxiety, to threat beliefs feeding into the formation of \npsychotic experiences, as proposed in cognitive models of psychosis (Freeman, 2007; Garety, \nKuipers, Fowler, Freeman, & Bebbington, 2001; Hardy et al., 2020). Overall, the idea that a \npathway from anxiety to psychotic experiences may be particularly relevant among women \nwith a history of childhood abuse has potentially important repercussions for clinical practice \nand deserves further investigation (Bloomfield et al., 2020). Moreover, at a population level, \nit may well be that links between affective and psychotic experiences following childhood \nabuse are manifestations of personality function. The interplay between borderline \npersonality functioning and affective instability, also involving psychosis, and subclinical and \nclinical levels of psychosis warrant further investigations (Barnow et al., 2010). \nAmong men, cannabis use and minority ethnic group status were identified as \npotential sources of heterogeneity in network connections between psychotic and affective \nsymptoms. Most striking differences were evident in the symptom network of men with a \nminority ethnic group status reporting no domestic violence or cannabis use. Documented \nvariations in experience and reporting of hallucinations (Vanheusden et al., 2008) and \ndelusions (Berg et al., 2014) in minority ethnic groups seem to extend to the level of \nsymptom networks. Here, they appear to be expressed as an increased co-occurrence of \nhallucinations and persecutory ideations in men from a minority ethnic background compared \nto men from the majority White or Mixed ethnic background. This finding agrees with the \nidea that, under the influence of risk factors, hallucinations and delusions can become \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n15 \n \nconnected, which has been linked to worse prognosis and symptom persistence (Binbay et al., \n2012; Smeets et al., 2012; Smeets, Lataster, Viechtbauer, Delespaul, & G.R.O.U.P., 2014; \nvan Os & Reininghaus, 2016). Taken together with the present results, this may reinforce \nevidence that demonstrates that people from a minority ethnic background, particularly men, \nare at increased risk for poor mental health outcomes (Morgan et al., 2017; Singh et al., \n2015). With the present data, however, it cannot be excluded that ethnicity acts as a proxy \nmeasure for factors not covered by our analysis, such as specific forms of deprivation. \nDelineating how mental health outcomes in men from a minority ethnic background are \ndetermined is an outstanding task for future research and may help to design more effective \ninterventions. Identifying potential commonalities underlying minority ethnic group status \nand domestic violence, both of which were associated with increased co-occurrence of \npsychotic experiences in men and women, respectively, may prove insightful in this context. \nExcept for domestic violence, which was a relevant moderating factor in women and \nmen, different risk factors explained heterogeneity in symptom networks of women and men, \nsuggesting a likely complex interplay between sex and risk factors in impacting psychosis \nrisk. Childhood sexual and physical abuse, for instance, were sources of heterogeneity in \nsymptom networks of women, but not men. This finding adds to previous research suggesting \nparticularly detrimental effects of sexual and physical abuse on mental health of girls and \nwomen (Adams, Mrug, & Knight, 2018; Thompson, Kingree, & Desai, 2004). One reason for \nthe distinct role of adversities may lie in the sex-specific effects they have on the nervous \nsystem, against the backdrop of sex differences in maturation, structure and functioning \nthereof (DeSantis et al., 2011; Dow-Edwards, 2020; Hodes & Epperson, 2019; Popovic et al., \n2020). Moreover, characteristics of some risk factors have been shown to differ by sex: men \nare more likely to engage in more escalating and chronic patterns of cannabis use than \nwomen, for example (Hawes, Trucco, Duperrouzel, Coxe, & Gonzalez, 2019; Wagner & \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n16 \n \nAnthony, 2007). Girls, on the other hand, are more likely than boys to experience severe \nforms of sexual abuse within close victim-perpetrator relationships (Gold, Elhai, Lucenko, \nSwingle, & Hughes, 1998; Kendall-Tackett, Williams, & Finkelhor, 1993). Such variations \nmay contribute to differing patterns of relationships between risk and symptom expression in \nwomen and men. \nOverall, our results corroborate a growing realization that research should appraise \nthat mechanisms contributing to psychosis expression may, at least in parts, differ by sex \n(Hodes & Epperson, 2019; Riecher-Rössler et al., 2018; Rosen et al., 2019; Stainton et al., \n2021). As clinical research works towards early identification and individually tailored \npreventive interventions, the complex interplay between sex and environmental factors in \nimpacting psychosis risk needs to be better understood to optimize these efforts (Hartung & \nLefler, 2019; Riecher-Rössler et al., 2018; Rosen et al., 2019; Stainton et al., 2021). This \nincludes disaggregating results by sex and gender in psychosis research more consistently \n(Hartung & Lefler, 2019), for example by documenting differences and similarities in \nsymptom networks of women and men. \nLimitations \nResults from the present study should be interpreted given several limitations. First, \nposterior predictive checks used for comparing the overall network connectivity tend to be \nconservative (Williams et al., 2020), which may have resulted in low sensitivity in post-hoc \ncomparisons. This factor, and small sample sizes in some subgroups, may explain why we \nfound no evidence that exposure to risk factors was associated with more densely connected \nsymptom networks compared to non-exposure, contrary to our hypothesis. Effects of risk \nfactors on symptom networks seem to be more specific, impacting single relations between \nsymptoms rather than connectivity between all symptoms. Second, model-based recursive \npartitioning identifies those variables that reduce heterogeneity in symptom networks the \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n17 \n \nmost. Thus, age of the respondent, alcohol use, separation experiences, bullying and \nsocioeconomic deprivation may explain heterogeneity in symptom networks, but not to the \nsame extent as the other risk factors tested. Related, differential relevance of risk factors, for \nexample within ethnic groups, may have remained undetected due to small sample sizes in \nsome subgroups. For a better understanding of the mechanisms relevant in different minority \ngroups, targeted investigations in these populations with larger sample sizes are needed. \nThird, we did not incorporate complex design features of the APMS, such as weights to take \nnon-response into account, due to the lack of established methods to do so for network \nmodels (Lin et al., 2019). Related, recursive partitioning currently only allows for complete \ncase analyses. Even though the percentage of excluded participants was small, they differed \nfrom included participants in some important aspects, including hallucinatory and depressive \nsymptoms, which may have biased our results. While therefore not fully representative of the \nEnglish population, our results are based on a large national household survey, with \nsuitability for a data-intensive method, such as network-based recursive partitioning, unlike \nfor smaller samples which would not offer the same opportunity. Fourth, the retrospective \nassessment of risk factors via self-report may be prone to memory biases and so directions of \neffect may be contested. Fifth, data used in the present analyses were gathered in a large \nepidemiological study; therefore, instruments and tools used were designed such that they \nwere simple to understand and appropriate given their use in over 7,000 people. This setting \nnecessarily leads to less refined assessments of symptomatology and risk. Sixth, the analyses \nwere based on cross-sectional data, meaning that the directions of interactions among the \nsymptoms remain unknown. Longitudinal studies are therefore an important next step for this \nline of research, and extension of recursive partitioning methods to personalized network \nstructures (e.g., derived from experience sampling methods) may allow for insights into how \nrisk factors moderate dynamic associations between symptoms in individuals. Lastly, some \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n18 \n \nresearchers have expressed concerns about stability and replicability of network models (e.g., \nestimates of edges (Forbes, Wright, Markon, & Krueger, 2017); for a summary of the debate, \nsee McNally, 2021). While our robustness analysis suggests that the networks and edge \nestimates are generally stable, especially weaker links in the networks of small subgroups \nshould be interpreted with care. Given that recursive partitioning and network methodology \nare data-driven, replication of present findings in other samples is needed to establish \ngeneralizability (Fried et al., 2018). \nConclusion \nSymptom networks of psychosis expression in the general population are highly \nheterogeneous. Sex was the primary source of heterogeneity, and different risk factors \nexplained further variability in symptom networks of women and men, potentially reflecting \ndistinct sex-specific mechanisms contributing to psychosis risk. Among women, an affective \nloading within psychosis, particularly following early interpersonal trauma, may have distinct \nimportance. Among men, the symptom network of those from a minority ethnic background \nshowed a particularly strong connection between hallucinatory experiences and persecutory \nideation, which may reflect poorer outcomes including symptom resolution in this group. A \nbetter understanding and consideration of these sex differences provides an important \nopportunity to deliver high quality prevention and patient-centered care in psychosis. \n \nRequired Statements \nAcknowledgements \nNone. \n \nFinancial Support \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n19 \n \nJK has received funding from the German Research Foundation (DFG; grant agreement n° \nKA 4413/1-1). MR has received funding from the Köln Fortune program (grant agreement n° \n304/2020). \n \nConflict of Interest \nThe authors declare no conflict of interests with relation to the work reported in this \nmanuscript. \n \nEthical Standards \nThe authors assert that all procedures contributing to this work comply with the ethical \nstandards of the relevant national and institutional committees on human experimentation and \nwith the Helsinki Declaration of 1975, as revised in 2008. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n20 \n \nReferences \nAdams, J., Mrug, S., & Knight, D. C. (2018). Characteristics of child physical and sexual \nabuse as predictors of psychopathology. Child Abuse & Neglect, 86, 167–177. \nhttps://doi.org/10.1016/j.chiabu.2018.09.019 \nBarnow, S., Arens, E. A., Sieswerda, S., Dinu-Biringer, R., Spitzer, C., & Lang, S. (2010). \nBorderline personality disorder and psychosis: a review. Current Psychiatry Reports, \n12(3), 186–195. https://doi.org/10.1007/s11920-010-0107-9 \nBebbington, P. (2015). Unravelling psychosis: psychosocial epidemiology, mechanism, and \nmeaning. Shanghai Archives of Psychiatry, 27(2), 70. \nhttps://doi.org/10.11919/j.issn.1002-0829.215027 \nBentall, R. P., Wickham, S., Shevlin, M., & Varese, F. (2012). Do specific early-life \nadversities lead to specific symptoms of psychosis? A study from the 2007 the Adult \nPsychiatric Morbidity Survey. Schizophrenia Bulletin, 38(4), 734–740. \nhttps://doi.org/10.1093/schbul/sbs049 \nBerg, A. O., Andreassen, O. A., Aminoff, S. R., Romm, K. L., Hauff, E., & Melle, I. (2014). \nThe impact of immigration and visible minority status on psychosis symptom profile. \nSocial Psychiatry and Psychiatric Epidemiology, 49(11), 1747–1757. \nhttps://doi.org/10.1007/s00127-014-0897-x \nBetz, L. T., Penzel, N., Kambeitz-Ilankovic, L., Rosen, M., Chisholm, K., Stainton, A., … \nPRONIA consortium. (2020). General psychopathology links burden of recent life \nevents and psychotic symptoms in a network approach. npj Schizophrenia, 6(1), 40. \nhttps://doi.org/10.1038/s41537-020-00129-w \nBinbay, T., Drukker, M., Elbi, H., Tanık, F. A., Özkınay, F., Onay, H., … Alptekin, K. \n(2012). Testing the psychosis continuum: differential impact of genetic and nongenetic \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n21 \n \nrisk factors and comorbid psychopathology across the entire spectrum of psychosis. \nSchizophrenia Bulletin, 38(5), 992–1002. https://doi.org/10.1093/schbul/sbr003 \nBloomfield, M. A. P., Yusuf, F. N. I. B., Srinivasan, R., Kelleher, I., Bell, V., & Pitman, A. \n(2020). Trauma-informed care for adult survivors of developmental trauma with \npsychotic and dissociative symptoms: a systematic review of intervention studies. The \nLancet Psychiatry, 7(5), 449–462. https://doi.org/10.1016/S2215-0366(20)30041-9 \nDavis, M. C., Matthews, K. A., & Twamley, E. W. (1999). Is life more difficult on Mars or \nVenus? A meta-analytic review of sex differences in major and minor life events. Annals \nof Behavioral Medicine, 21(1), 83–97. https://doi.org/10.1007/BF02895038 \nDeRosse, P., & Karlsgodt, K. H. (2015). Examining the Psychosis Continuum. Current \nBehavioral Neuroscience Reports, 2(2), 80–89. https://doi.org/10.1007/s40473-015-\n0040-7 \nDeSantis, S. M., Baker, N. L., Back, S. E., Spratt, E., Ciolino, J. D., Moran-Santa Maria, M., \n… Brady, K. T. (2011). Gender differences in the effect of early life trauma on \nhypothalamic-pituitary-adrenal axis functioning. Depression and Anxiety, 28(5), 383–\n392. https://doi.org/10.1002/da.20795 \nDow-Edwards, D. (2020). Sex differences in the interactive effects of early life stress and the \nendocannabinoid system. Neurotoxicology and Teratology, 80, 106893. \nhttps://doi.org/10.1016/j.ntt.2020.106893 \nEpskamp, S., Borsboom, D., & Fried, E. I. (2018). Estimating psychological networks and \ntheir accuracy: A tutorial paper. Behavior Research Methods, 50(1), 195–212. \nhttps://doi.org/10.3758/s13428-017-0862-1 \nEpskamp, S., Cramer, A. O. J., Waldorp, L. J., Schmittmann, V. D., & Borsboom, D. (2012). \nqgraph: Network visualizations of relationships in psychometric data. Journal of \nStatistical Software, 48(4), 1–18. https://doi.org/10.18637/jss.v048.i04 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n22 \n \nForbes, M. K., Wright, A. G. C., Markon, K. E., & Krueger, R. F. (2017). Evidence that \npsychopathology symptom networks have limited replicability. Journal of Abnormal \nPsychology, 126(7), 969–988. https://doi.org/10.1037/abn0000276 \nFreeman, D. (2007). Suspicious minds: the psychology of persecutory delusions. Clinical \nPsychology Review, 27(4), 425–457. https://doi.org/10.1016/j.cpr.2006.10.004 \nFreeman, D., Brugha, T., Meltzer, H., Jenkins, R., Stahl, D., & Bebbington, P. (2010). \nPersecutory ideation and insomnia: findings from the second British National Survey of \nPsychiatric Morbidity. Journal of Psychiatric Research, 44(15), 1021–1026. \nhttps://doi.org/10.1016/j.jpsychires.2010.03.018 \nFreeman, D., Sheaves, B., Goodwin, G. M., Yu, L.-M., Nickless, A., Harrison, P. J., … \nEspie, C. A. (2017). The effects of improving sleep on mental health (OASIS): a \nrandomised controlled trial with mediation analysis. The Lancet. Psychiatry, 4(10), 749–\n758. https://doi.org/10.1016/S2215-0366(17)30328-0 \nFried, E. I., Eidhof, M. B., Palic, S., Costantini, G., Huisman-van Dijk, H. M., Bockting, C. \nL. H., … Karstoft, K.-I. (2018). Replicability and Generalizability of Posttraumatic \nStress Disorder (PTSD) Networks: A Cross-Cultural Multisite Study of PTSD \nSymptoms in Four Trauma Patient Samples. Clinical Psychological Science, 6(3), 335–\n351. https://doi.org/10.1177/2167702617745092 \nGarety, P. A., Kuipers, E., Fowler, D., Freeman, D., & Bebbington, P. E. (2001). A cognitive \nmodel of the positive symptoms of psychosis. Psychological Medicine, 31(2), 189–195. \nhttps://doi.org/10.1017/s0033291701003312 \nGold, S. N., Elhai, J. D., Lucenko, B. A., Swingle, J. M., & Hughes, D. M. (1998). Abuse \ncharacteristics among childhood sexual abuse survivors in therapy: a gender comparison. \nChild Abuse & Neglect, 22(10), 1005–1012. https://doi.org/10.1016/S0145-\n2134(98)00085-4 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n23 \n \nGuloksuz, S., van Nierop, M., Bak, M., de Graaf, R., Ten Have, M., van Dorsselaer, S., … \nvan Os, J. (2016). Exposure to environmental factors increases connectivity between \nsymptom domains in the psychopathology network. BMC Psychiatry, 16, 223. \nhttps://doi.org/10.1186/s12888-016-0935-1 \nGuloksuz, S., van Nierop, M., Lieb, R., van Winkel, R., Wittchen, H.-U., & van Os, J. \n(2015). Evidence that the presence of psychosis in non-psychotic disorder is \nenvironment-dependent and mediated by severity of non-psychotic psychopathology. \nPsychological Medicine, 45(11), 2389–2401. \nhttps://doi.org/10.1017/S0033291715000380 \nHardy, A., O’Driscoll, C., Steel, C., van der Gaag, M., & van den Berg, D. (2020). A network \nanalysis of post-traumatic stress and psychosis symptoms. Psychological Medicine, 1–8. \nhttps://doi.org/10.1017/S0033291720001300 \nHartung, C. M., & Lefler, E. K. (2019). Sex and gender in psychopathology: DSM-5 and \nbeyond. Psychological Bulletin, 145(4), 390–409. https://doi.org/10.1037/bul0000183 \nHawes, S. W., Trucco, E. M., Duperrouzel, J. C., Coxe, S., & Gonzalez, R. (2019). \nDevelopmental pathways of adolescent cannabis use: Risk factors, outcomes and sex-\nspecific differences. Substance Use & Misuse, 54(2), 271–281. \nhttps://doi.org/10.1080/10826084.2018.1517177 \nHodes, G. E., & Epperson, C. N. (2019). Sex Differences in Vulnerability and Resilience to \nStress Across the Life Span. Biological Psychiatry, 86(6), 421–432. \nhttps://doi.org/10.1016/j.biopsych.2019.04.028 \nHothorn, T., Hornik, K., van de Wiel, M., & Zeileis, A. (2008). Implementing a Class of \nPermutation Tests: The coin Package. Journal of Statistical Software, 28(8), 1–23. \nhttps://doi.org/10.18637/jss.v028.i08 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n24 \n \nIsvoranu, A.-M., Borsboom, D., van Os, J., & Guloksuz, S. (2016). A Network Approach to \nEnvironmental Impact in Psychotic Disorder: Brief Theoretical Framework. \nSchizophrenia Bulletin, 42(4), 870–873. https://doi.org/10.1093/schbul/sbw049 \nIsvoranu, A.-M., Guloksuz, S., Epskamp, S., van Os, J., Borsboom, D., & GROUP \nInvestigators. (2019). Toward incorporating genetic risk scores into symptom networks \nof psychosis. Psychological Medicine, 1–8. \nhttps://doi.org/10.1017/S003329171900045X \nIsvoranu, A.-M., van Borkulo, C. D., Boyette, L.-L., Wigman, J. T. W., Vinkers, C. H., \nBorsboom, D., & Group Investigators. (2017). A Network Approach to Psychosis: \nPathways Between Childhood Trauma and Psychotic Symptoms. Schizophrenia Bulletin, \n43(1), 187–196. https://doi.org/10.1093/schbul/sbw055 \nJones, P. J., Mair, P., Simon, T., & Zeileis, A. (2020). Network Trees: A Method for \nRecursively Partitioning Covariance Structures. Psychometrika, 85(4), 926–945. \nhttps://doi.org/10.1007/s11336-020-09731-4 \nKelleher, I., & Cannon, M. (2011). Psychotic-like experiences in the general population: \ncharacterizing a high-risk group for psychosis. Psychological Medicine, 41(1), 1–6. \nhttps://doi.org/10.1017/S0033291710001005 \nKelleher, I., Keeley, H., Corcoran, P., Lynch, F., Fitzpatrick, C., Devlin, N., … Cannon, M. \n(2012). Clinicopathological significance of psychotic experiences in non-psychotic \nyoung people: evidence from four population-based studies. The British Journal of \nPsychiatry: The Journal of Mental Science, 201(1), 26–32. \nhttps://doi.org/10.1192/bjp.bp.111.101543 \nKendall-Tackett, K. A., Williams, L. M., & Finkelhor, D. (1993). Impact of sexual abuse on \nchildren: a review and synthesis of recent empirical studies. Psychological Bulletin, \n113(1), 164–180. https://doi.org/10.1037/0033-2909.113.1.164 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n25 \n \nLinscott, R. J., & van Os, J. (2013). An updated and conservative systematic review and \nmeta-analysis of epidemiological evidence on psychotic experiences in children and \nadults: on the pathway from proneness to persistence to dimensional expression across \nmental disorders. Psychological Medicine, 43(6), 1133–1149. \nhttps://doi.org/10.1017/S0033291712001626 \nLin, S.-Y., Fried, E. I., & Eaton, N. R. (2019). The association of life stress with substance \nuse symptoms: A network analysis and replication. Journal of Abnormal Psychology, \n129(2), 204–214. https://doi.org/10.1037/abn0000485 \nMcManus, S., Meltzer, H., Brugha, T. S., Bebbington, P. E., & Jenkins, R. (2009). Adult \nPsychiatric Morbidity in England, 2007: Results of a Household Survey. \nhttps://doi.org/10.13140/2.1.1563.5205 \nMcNally, R. J. (2021). Network Analysis of Psychopathology: Controversies and Challenges. \nAnnual Review of Clinical Psychology, 17, 31–53. https://doi.org/10.1146/annurev-\nclinpsy-081219-092850 \nMoffa, G., Catone, G., Kuipers, J., Kuipers, E., Freeman, D., Marwaha, S., … Bebbington, P. \n(2017). Using Directed Acyclic Graphs in Epidemiological Research in Psychosis: An \nAnalysis of the Role of Bullying in Psychosis. Schizophrenia Bulletin, 43(6), 1273–\n1279. https://doi.org/10.1093/schbul/sbx013 \nMorgan, C., Fearon, P., Lappin, J., Heslin, M., Donoghue, K., Lomas, B., … Dazzan, P. \n(2017). Ethnicity and long-term course and outcome of psychotic disorders in a UK \nsample: the ÆSOP-10 study. The British Journal of Psychiatry: The Journal of Mental \nScience, 211(2), 88–94. https://doi.org/10.1192/bjp.bp.116.193342 \nMoriarity, D. P., van Borkulo, C., & Alloy, L. B. (2020). Inflammatory phenotype of \ndepression symptom structure: A network perspective. Brain, Behavior, and Immunity, \n93, 35–42. https://doi.org/10.1016/j.bbi.2020.12.005 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n26 \n \nMurphy, J., McBride, O., Fried, E., & Shevlin, M. (2018). Distress, Impairment and the \nExtended Psychosis Phenotype: A Network Analysis of Psychotic Experiences in an US \nGeneral Population Sample. Schizophrenia Bulletin, 44(4), 768–777. \nhttps://doi.org/10.1093/schbul/sbx134 \nMyin-Germeys, I., & van Os, J. (2007). Stress-reactivity in psychosis: evidence for an \naffective pathway to psychosis. Clinical Psychology Review, 27(4), 409–424. \nhttps://doi.org/10.1016/j.cpr.2006.09.005 \nNational Centre for Social Research, University of Leicester. (2017). Adult Psychiatric \nMorbidity Survey, 2007 (4th ed., Version 6379) [Data set]. UK Data Service. \nhttps://doi.org/10.5255/UKDA-SN-6379-2 \nPopovic, D., Ruef, A., Dwyer, D. B., Antonucci, L. A., Eder, J., Sanfelici, R., … PRONIA \nConsortium. (2020). Traces of Trauma: A Multivariate Pattern Analysis of Childhood \nTrauma, Brain Structure, and Clinical Phenotypes. Biological Psychiatry, 88(11), 829–\n842. https://doi.org/10.1016/j.biopsych.2020.05.020 \nRiecher-Rössler, A., Butler, S., & Kulkarni, J. (2018). Sex and gender differences in \nschizophrenic psychoses-a critical review. Archives of Women’s Mental Health, 21(6), \n627–648. https://doi.org/10.1007/s00737-018-0847-9 \nRobinaugh, D. J., Hoekstra, R. H. A., Toner, E. R., & Borsboom, D. (2020). The network \napproach to psychopathology: a review of the literature 2008-2018 and an agenda for \nfuture research. Psychological Medicine, 50(3), 353–366. \nhttps://doi.org/10.1017/S0033291719003404 \nRosen, M., Haidl, T. K., Ruhrmann, S., Vogeley, K., & Schultze-Lutter, F. (2019). Sex \ndifferences in symptomatology of psychosis-risk patients and in prediction of psychosis. \nArchives of Women’s Mental Health, 23, 339–349. https://doi.org/10.1007/s00737-019-\n01000-3 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n27 \n \nRussell, J. D., Keding, T. J., He, Q., Li, J. J., & Herringa, R. J. (2020). Childhood exposure to \ninterpersonal violence is associated with greater transdiagnostic integration of \npsychiatric symptoms. Psychological Medicine, 1–9. \nhttps://doi.org/10.1017/S0033291720003712 \nSingh, S. P., Brown, L., Winsper, C., Gajwani, R., Islam, Z., Jasani, R., … Birchwood, M. \n(2015). Ethnicity and pathways to care during first episode psychosis: the role of cultural \nillness attributions. BMC Psychiatry, 15, 287. https://doi.org/10.1186/s12888-015-0665-\n9 \nSingleton, N., Bumpstead, R., O’Brien, M., Lee, A., & Meltzer, H. (2003). Psychiatric \nmorbidity among adults living in private households, 2000. International Review of \nPsychiatry, 15(1-2), 65–73. https://doi.org/10.1080/0954026021000045967 \nSmeets, F., Lataster, T., Dominguez, M.-D.-G., Hommes, J., Lieb, R., Wittchen, H.-U., & van \nOs, J. (2012). Evidence That Onset of Psychosis in the Population Reflects Early \nHallucinatory Experiences That Through Environmental Risks and Affective \nDysregulation Become Complicated by Delusions. Schizophrenia Bulletin, 38(3), 531–\n542. https://doi.org/10.1093/schbul/sbq117 \nSmeets, F., Lataster, T., Viechtbauer, W., Delespaul, P., & G.R.O.U.P. (2014). Evidence That \nEnvironmental and Genetic Risks for Psychotic Disorder May Operate by Impacting on \nConnections Between Core Symptoms of Perceptual Alteration and Delusional Ideation. \nSchizophrenia Bulletin, 41(3), 687–697. https://doi.org/10.1093/schbul/sbu122 \nStainton, A., Chisholm, K., Woodall, T., Hallett, D., Reniers, R. L. E. P., Lin, A., & Wood, S. \nJ. (2021). Gender differences in the experience of psychotic-like experiences and their \nassociated factors: A study of adolescents from the general population. Schizophrenia \nResearch, 228, 410–416. https://doi.org/10.1016/j.schres.2021.01.008 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n28 \n \nStrobl, C., Malley, J., & Tutz, G. (2009). An introduction to recursive partitioning: rationale, \napplication, and characteristics of classification and regression trees, bagging, and \nrandom forests. Psychological Methods, 14(4), 323–348. \nhttps://doi.org/10.1037/a0016973 \nThompson, M. P., Kingree, J. B., & Desai, S. (2004). Gender differences in long-term health \nconsequences of physical abuse of children: data from a nationally representative survey. \nAmerican Journal of Public Health, 94(4), 599–604. \nhttps://doi.org/10.2105/ajph.94.4.599 \nTolin, D. F., & Foa, E. B. (2002). Gender and PTSD: A cognitive model. In R. Kimerling \n(Ed.), Gender and PTSD (Vol. 460, pp. 76–97). New York, NY, US: The Guilford Press. \nRetrieved from https://psycnet.apa.org/fulltext/2003-04410-003.pdf \nUpthegrove, R., Chard, C., Jones, L., Gordon-Smith, K., Forty, L., Jones, I., & Craddock, N. \n(2015). Adverse childhood events and psychosis in bipolar affective disorder. The British \nJournal of Psychiatry: The Journal of Mental Science, 206(3), 191–197. \nhttps://doi.org/10.1192/bjp.bp.114.152611 \nUpthegrove, R., Lalousis, P., Mallikarjun, P., Chisholm, K., Griffiths, S. L., Iqbal, M., … \nPRONIA Consortium. (2020). The Psychopathology and Neuroanatomical Markers of \nDepression in Early Psychosis. Schizophrenia Bulletin, 47(1), 249–258. \nhttps://doi.org/10.1093/schbul/sbaa094 \nUpthegrove, R., Marwaha, S., & Birchwood, M. (2017). Depression and Schizophrenia: \nCause, Consequence, or Trans-diagnostic Issue? Schizophrenia Bulletin, 43(2), 240–244. \nhttps://doi.org/10.1093/schbul/sbw097 \nVanheusden, K., Mulder, C. L., van der Ende, J., Selten, J.-P., van Lenthe, F. J., Verhulst, F. \nC., & Mackenbach, J. P. (2008). Associations between ethnicity and self-reported \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n29 \n \nhallucinations in a population sample of young adults in The Netherlands. Psychological \nMedicine, 38(8), 1095–1102. https://doi.org/10.1017/S0033291707002401 \nvan Nierop, M., Lecei, A., Myin-Germeys, I., Collip, D., Viechtbauer, W., Jacobs, N., … van \nWinkel, R. (2018). Stress reactivity links childhood trauma exposure to an admixture of \ndepressive, anxiety, and psychosis symptoms. Psychiatry Research, 260, 451–457. \nhttps://doi.org/10.1016/j.psychres.2017.12.012 \nvan Nierop, M., Viechtbauer, W., Gunther, N., van Zelst, C., de Graaf, R., Ten Have, M., … \nvan Winkel, R. (2015). Childhood trauma is associated with a specific admixture of \naffective, anxiety, and psychosis symptoms cutting across traditional diagnostic \nboundaries. Psychological Medicine, 45(6), 1277–1288. \nhttps://doi.org/10.1017/S0033291714002372 \nvan Os, J., & Reininghaus, U. (2016). Psychosis as a transdiagnostic and extended phenotype \nin the general population. World Psychiatry, 15(2), 118–124. \nhttps://doi.org/10.1002/wps.20310 \nWagner, F. A., & Anthony, J. C. (2007). Male-female differences in the risk of progression \nfrom first use to dependence upon cannabis, cocaine, and alcohol. Drug and Alcohol \nDependence, 86(2-3), 191–198. https://doi.org/10.1016/j.drugalcdep.2006.06.003 \nWilliams, D. R. (2021). Bayesian Estimation for Gaussian Graphical Models: Structure \nLearning, Predictability, and Network Comparisons. Multivariate Behavioral Research, \n1–17. https://doi.org/10.1080/00273171.2021.1894412 \nWilliams, D. R., & Mulder, J. (2019). BGGM: A R package for Bayesian Gaussian Graphical \nModels. Preprint. Retrieved from https://psyarxiv.com/3b5hf/download?format=pdf \nWilliams, D. R., Rast, P., Pericchi, L. R., & Mulder, J. (2020). Comparing Gaussian graphical \nmodels with the posterior predictive distribution and Bayesian model selection. \nPsychological Methods, 25(5), 653–672. https://doi.org/10.1037/met0000254 \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n30 \n \nWüsten, C., Schlier, B., Jaya, E. S., Genetic Risk and Outcome of Psychosis (GROUP) \nInvestigators, Fonseca-Pedrero, E., Peters, E., … Lincoln, T. M. (2018). Psychotic \nExperiences and Related Distress: A Cross-national Comparison and Network Analysis \nBased on 7141 Participants From 13 Countries. Schizophrenia Bulletin, 44(6), 1185–\n1194. https://doi.org/10.1093/schbul/sby087 \nZeileis, A., Hothorn, T., & Hornik, K. (2008). Model-Based Recursive Partitioning. Journal \nof Computational and Graphical Statistics, 17(2), 492–514. \nhttps://doi.org/10.1198/106186008X319331 \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n31 \n \nFigure Captions \nFigure 1. Recursive partitioning for symptom networks as applied to data from the 2007 \nAdult Psychiatric Morbidity Survey (APMS) study. The goal is to assess which of the \nincluded demographic and risk factors capture individual deviations from the correlation \nmatrix of symptom scores, which underlies symptom networks. Starting with the whole \nsample, individual deviations from the correlation matrix of symptom scores are computed \nvia a log-likelihood-based score function. The variable that explains these deviations best, as \ndetermined by a minimum p-value strategy at Bonferroni-corrected α , is selected (here: sex), \nand the sample split accordingly. Within the identified subgroups, the procedure is repeated \nrecursively until no significant deviations, i.e., heterogeneity, is detected. We compared \nsymptom networks of the identified subgroups in terms of global strength and individual edge \nweights. For a detailed account of the method, see supplementary method 3 and Jones et al. \n(2020). \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n32 \n \nFigure 2. Results from recursive partitioning, depicted as a decision tree of partial correlation \nnetworks. Numbers behind splitting factors give the sample size retained after the \ncorresponding sample split. Symptom domains are differentiated by color. The thicker and \nless transparent the edge, the stronger the partial correlation between two symptoms. Blue \n(red) edges indicate positive (negative) relationships. To ensure visual comparability, edge \nweights were scaled identically across all networks. Only connections representing edge \nweights larger than 0.01 are depicted. \n  \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n33 \n \nFigure 3. Partial correlation networks estimated in women (n = 4,115) and men (n = 3,127). \nSex was identified as the first split in the recursive partitioning approach, suggesting that sex \nwas the primary source of heterogeneity in symptom networks. Symptom domains are \ndifferentiated by color. The thicker and less transparent the edge, the stronger the partial \ncorrelation between two symptoms. Blue (red) edges indicate positive (negative) \nrelationships. To ensure visual comparability, edge weights were scaled identically across \nboth networks.  \n \n \n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint \n\n . CC-BY 4.0 International licenseIt is made available under a \n is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)\nThe copyright holder for this preprint this version posted August 10, 2021. ; https://doi.org/10.1101/2021.05.06.21256748doi: medRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}