Computational Psychiatry

Hoarding disorder (HD) affects approximately 2-3% of adults and is associated with substantial functional disability and limited access to evidence-based care. The aim of the current analysis was to examine the naturalistic effectiveness of therapist-delivered video cognitive-behavioral therapy (CBT) for HD in a large real-world sample, and to characterize individual-level treatment response, time-to-response, and moderators of outcome. This retrospective, observational analysis examined clinical data from 305 adults diagnosed with HD who received therapist-delivered video CBT through an online specialty therapy platform between September 2021 and February 2026. Hoarding symptom severity was assessed using the Hoarding Rating Scale-Self Report (HRS-SR). Linear mixed models examined symptom change from baseline to three timepoints: session 10, session 20, and each patient's final session. HRS-SR scores decreased from M = 22.4 (SD = 7.6) at baseline to M = 16.4 (SD = 8.2) at final session (Hedges' g = 0.81, 95% CI: 0.68-0.94). By the final session, median percent improvement was 25.0% [IQR: 3.0-46.7%]. A total of 39.3% of patients achieved [≥]35% HRS-SR reduction, 27.4% of patients who began above the clinical threshold achieved remission, 36.4% demonstrated reliable improvement, and 22.9% of eligible patients achieved clinically significant change. Among patients who achieved and maintained [≥]35% reduction through their final session (n = 120), median time to first response was session 9, with 54.2% responding within 10 sessions. Analyses of secondary outcomes showed significant improvements in clutter severity, depressive and anxiety symptoms, stress, quality of life, and functional disability (Hedges' g = 0.21-0.47). Greater baseline severity, more sessions, and longer treatment duration significantly moderated outcomes; prior OCD treatment history did not. Findings suggest that therapist-delivered video CBT for HD, delivered remotely in a real-world setting, produces outcomes consistent with controlled trials and may be a clinically effective and scalable approach for a condition historically underserved by mental health systems.

I ntroduction: Prospective memory (PM) deficits in children with attention-deficit/hyperactivity disorder (ADHD) significantly impact academic and daily functioning. Through two experiments, this study investigated how cognitive load and encoding strategies modulate PM performance. Methods: Experiment 1 included 43 children (21 ADHD, 22 typically developing) who completed an n-back task under high and low cognitive load. Experiment 2 included 44 children with ADHD who were randomly assigned to either a standard encoding group or an implementation intention encoding group, also completing the n-back task under both load conditions. Results: Experiment 1 showed that children with ADHD had significantly lower PM accuracy than typically developing peers. Signal detection analysis revealed that this deficit stemmed from a more conservative response bias rather than impaired perceptual sensitivity. Unexpectedly, PM accuracy and perceptual sensitivity were higher under high cognitive load than low load for both groups. Experiment 2 demonstrated that implementation intention encoding significantly enhanced PM accuracy and perceptual sensitivity in children with ADHD, with stable effects across load conditions and no interference with ongoing task performance. Discussion: These findings indicate that PM deficits in children with ADHD reflect a conservative response strategy rather than an inability to detect target cues. Implementation intention encoding provides an effective, load-independent cognitive strategy for enhancing PM performance. These results offer novel insights into the cognitive mechanisms underlying PM deficits in ADHD and provide evidence-based guidance for targeted interventions.

Background: Internet-based cognitive behavioral therapy (iCBT) is efficacious for panic disorder (PD), yet the mechanisms of change remain underspecified. Anxiety sensitivity (AS) is theoretically central to PD maintenance, but its role as a mediator has not been formally tested in Spanish-speaking populations using minimal-contact formats. This study evaluates the efficacy of the "Free from Anxiety" iCBT program and examines AS as a mediator of clinical outcomes. Methods: In a randomized controlled trial, 95 adults meeting DSM-IV-TR criteria for PD were assigned to an 8-week iCBT program with optional email support (n = 49) or a waiting-list control (n = 46). Primary outcome was PD severity (PDSS); secondary outcomes included anxiety sensitivity (ASI-3), general anxiety (BAI), and depression (BDI-II). Mediation was assessed via Baron and Kenny's framework with bootstrapping (5,000 resamples) to estimate the indirect effect of ASI-3 change on PDSS reduction. Results: The treatment group showed significant improvements across all measures compared to controls (PDSS: d = 0.76, 95% CI [0.10, 1.42]; mean d = 1.30). Mediation analysis confirmed that ASI-3 change partially mediated the treatment effect on PDSS (indirect effect = 1.85, 95% CI [0.36, 3.70]), accounting for 27.4% of the total effect. The direct effect remained significant (b = 4.89, p < .001). Intent-to-treat (ITT) analyses supported robustness (d = 0.47 to 1.47). Gains were maintained at 6-month follow-up (d = 1.19 to 1.26). Conclusions: iCBT reduces anxiety sensitivity as a partial mechanism of change, aligning with cognitive models of panic. These findings support Free from Anxiety as an evidence-based, viable first-step intervention for Spanish-speaking clinical populations within stepped-care pathways.

The ability to adjust to changing environments (cognitive flexibility) and optimal decision-making are pivotal brain functions that govern successful human behavior. Anxiety and depressive disorders are strongly pervasive psychiatric conditions across the lifespan that profoundly disrupt mechanisms of attention, working memory, and decision-making. Although existing task evidence documents impaired decision-making and flexibility outcomes for both anxiety and depression, there is a growing need to systematically evaluate the role of anxiety and depression and to quantitatively compare the effects of these disorders on these domains. In the present study, we conducted a meta-analysis of anxiety and depression on decision-making and cognitive flexibility. We utilized a random-effects approach, given that a large amount of between-subject heterogeneity was anticipated. Given the scope of this meta-analysis, we used the machine learning tool asReview to more efficiently conduct a meta-analytic search. Across all outcomes, results showed anxiety and depression were associated with reduced cognitive flexibility and decision-making. These effect sizes were then tested for significance using a fixed-effects (plural) model. Subgroup analyses revealed no significant differences between anxiety and depression for either decision-making or flexibility outcomes, consistent with a transdiagnostic perspective. Results are contextualized in light of the biopsychosocial model and potential transdiagnostic factors.

Adolescence is a critical developmental window for the emergence of substance use and psychosis-spectrum symptoms, yet early risk for these outcomes remains poorly understood. Using longitudinal data from the Adolescent Brain Cognitive Development (ABCD) Study (n=10,134), we tested whether demographic, clinical, and structural and functional neuroimaging measures assessed in childhood (mean baseline age=9.96 years) predict later adolescent substance use, psychotic-like experiences, and/or their co-occurrence. Multivariate machine learning models reliably predicted later emergence of psychotic-like experiences (AUROC=0.780) and their co-occurrence with substance use (AUROC= 0.828), as well as substance use on its own (AUROC=0.626). Distinct patterns of functional brain connectivity, task-related brain activation, demographic, and clinical factors differentiated each outcome. Findings suggest that partially dissociable developmental risk profiles are detectable as early as childhood, and results underscore the importance of explicitly modeling comorbidity when interrogating risk factors for mental health outcomes.

AimsTo examine the longitudinal dynamic interactions of craving and drug use in the course of treatment of stimulant use disorders. DesignCross-lagged residual dynamic structural equation modeling (R-DSEM) was used to examine the reciprocal (bidirectional) longitudinal associations between craving and drug use. SettingPooled data from 11 randomized controlled trials of pharmacotherapies for methamphetamine and cocaine use disorders in the United States sponsored by the National Institute on Drug Abuse. Participants1,936 adults with cocaine or methamphetamine use disorder. MeasurementsCraving was measured using Brief Substance Craving Scale (BSCS), drug use was measured using Timeline Followback and urine drug screen (UDS). FindingsCraving and stimulant drug use were dynamically associated over time (within-person association). Daily craving significantly predicted drug use in subsequent days (estimate=0.092, 95% credible interval [CrI]=0.081, 0.103 for self-reported drug use and estimate=0.081, 95% CrI=0.069, 0.095 for UDS-ascertained drug use). In turn, drug use predicted subsequent craving (estimate=0.361, 95% CrI=0.325, 0.398 and estimate=0.060, 95% CrI=0.028, 0.094, respectively). There was substantial between-person heterogeneity in these cross-lagged effects, as reflected in the coefficients of variation ranging from 0.78 to 2.88. ConclusionsThere is a bidirectional interaction between stimulant drug craving and drug use. The heterogeneity in the interaction of craving with stimulant drug use may partly explain between-person variability in responses to anti-craving medications in treatment of stimulant use disorders.

Formal thought disorder (FTD) is a core psychosis feature. Disentangling its dimensions requires tasks simple enough for formal modeling yet sensitive enough to capture individual variation across the psychosis spectrum. The semantic verbal fluency task offers precisely this: a structured behavioral trace of semantic memory sampling, amenable to computational analysis using distributed word embeddings. We hypothesized that this sampling process is governed by two dissociable mechanisms mapping onto FTD dimensions: initial retrieval drive (d0), quantifying the motivational resource sustaining production, and semantic search precision (), quantifying how strongly similarity to the preceding word constrains each retrieval step from near-random to highly structured. We hypothesized that reduced d0 would track negative psychosis symptoms and alogia, while degraded would track language disorganization and left inferior longitudinal fasciculus (ILF) fractional anisotropy. We tested these predictions in a primary (N = 120) and an independent replication sample (N = 249) of German-speaking individuals across the psychosis spectrum. Both parameters decreased with greater psychosis severity and, in the primary sample, they dissociated regarding their clinical correlates. d0 correlated negatively with negative symptoms, general psychopathology, and poverty of speech, consistent with a computational signature of alogia. correlated negatively with positive symptoms and cognitive flexibility, and, in individuals with psychosis, positively with left ILF fractional anisotropy. The association between d0 and negative symptoms was replicated in the independent sample. These findings pave the way for mechanistic, automatically derived FTD markers capturing subclinical variation across the psychosis spectrum and mapping onto underlying cognitive and neural processes.

Background: Dissociative experiences encompass a variety of discontinuities in awareness and perception that are elevated in the dissociative disorders and associated with extensive comorbid symptomatology. Accumulating evidence points to developmental trauma and trait responsiveness to verbal suggestions (REVS) as factors that confer risk for severe dissociative symptoms, but they have typically been studied in isolation. This study integrated these measures using prediction modelling to better understand their predictive value for the risk of dissociative psychopathology. Method: 1,104 non-clinical participants completed measures of trauma, dissociation and trait REVS. The predictive model was developed using elastic net logistic regression, internally validated with 10-fold cross-validation, and assessed using receiver operating characteristic (ROC) curve and area under the ROC (AUROC). Variables entered into the model were components of REVS, trauma, age, and their interactions. Results: A dissociative psychopathology at-risk group (7%) was characterised by younger age, greater trauma and elevated REVS, particularly involuntariness during cognitive-perceptual suggestions. The prediction model retained nine of ten predictors, with an AUROC of .77 [95% CI: .73, .82], reflecting good discrimination with moderate sensitivity (78%) but modest specificity (67%). Conclusions: These findings reinforce trauma and trait REVS as risk factors for dissociative psychopathology and demonstrate that they can be integrated in a model that can identify at-risk individuals. Further validation and extension of the model is necessary to improve the identification of individuals at risk for severe dissociative symptomatology and the diagnosis of dissociative disorders with implications for outcome trajectories.

Background: Conventional psychiatric screening instruments summarize symptoms within individual scales and prioritize cases with high single-instrument additive score severity. This design treats items as independent within instruments and ignores cross-instrument covariance structure, making it insensitive to respondents whose responses are distributed across multiple domains in unusual combinations that remain below threshold on every individual scale. Methods: We analyzed two cohorts spanning older and younger adults. Item prompts from depression, stress, anxiety, and sleep instruments were embedded into a shared semantic space using a pretrained sentence encoder. Principal component analysis of the item-prompt embeddings alone---with no use of respondent data at this stage---was used to construct a low-dimensional subspace retaining 80\% of variance in the item embedding matrix. Normalized participant responses were then projected into this subspace, with Jaccard-based stability analysis used as a check on dimensional robustness. Multivariate deviation from the cohort norm was quantified with Mahalanobis distance using Ledoit-Wolf covariance regularization. Candidate outliers were defined by the empirical 95th percentile of the cohort-specific distance distribution. To isolate response configurations not already captured by conventional single-instrument extreme-value logic, we excluded all outlier respondents who had endorsed any individual item at the maximum value of its Likert scale on any instrument. For the remaining outliers, anomalous components were backtracked to their original item loadings for interpretation. Results: In the older-adult Health and Retirement Study (HRS) cohort, principal component analysis of 27 item-prompt embeddings showed that a 10-dimensional subspace provided a stable representation of cross-instrument semantic structure. In the younger-adult Xinxiang cohort the corresponding stable solution was 16-dimensional. In each cohort, seven respondents remained as multivariate outliers despite falling below every single-instrument extreme-value threshold. These cases were not characterized by uniformly severe symptom scores but by unusual cross-domain response configurations that became visible only in the shared semantic covariance subspace. The response structure of the retained configurations differed across cohorts: older-adult cases more often involved weak endorsement of mood-labeled items alongside nonzero body- and sleep-related responses, whereas younger-adult cases more often involved incomplete response configurations spanning mood, sleep, stress, and self-harm-related items. Conclusions: A semantically aligned, auditable covariance subspace provides a practical tool for flagging unusual multivariate response configurations that single-instrument additive screening may not flag. The method is interpretable at the level of original item contributions. It should be understood as a hypothesis-generating screen for unusual response configurations requiring further clinical assessment, not as a diagnostic instrument. Outcome validity remains to be established by prospective study.

RationaleThe progression of psychostimulant abuse is associated with a shift from recreational to habitual use (R2H-shift). Because this R2H-shift can be modeled using behavioral economics, we developed a novel Behavioral Economic model for the Analysis of Self-administration Time-curve (BEAST) to obtain R2H-shift variable(s). The relationship(s) between R2H-shift variables and drug intake (under normal and/or punishment conditions) is/are unknown. Our goal was to determine if the R2H-shift variable and intake variables obtained during the initial self-administration training phase were related to 1) drug intake at that time, and subsequent drug intake under 2) normal, 3) punishment, 4) post-punishment, and 5) price-constrained conditions. MethodLong Evans rats self-administered methamphetamine (METH, males n = 16, females n = 14), sucrose (males n = 22, females n = 22) and/or saline (males n = 3, females n = 10) under FR1 for 6 h per day for 20 days to obtain 1) followed by the assessment of subsequent drug intake under different conditions (2-5 above). We obtained all variables referenced above. We determined the relationships between all variables (multivariate analysis). ResultsThere were no sex differences detected in the METH and sucrose studies. For METH and sucrose, prior drug intake levels could predict drug intake under normal/punishment but not under price-constrained conditions. The R2H-shift variable could predict drug intake under a consumption-price curve but could not predict intake under normal/punishment conditions. ConclusionsWhile related to economic demand, the recreational-to-habitual shift rate was unrelated to drug intake levels (under normal and punishment conditions).

In a single-blind randomized controlled trial, we investigated the effectiveness of real-time fMRI neurofeedback delivered in 7 runs over three sessions across two weeks in N = 65 patients with alcohol use disorder. The intervention targeted modulation of ventral striatal cue reactivity to alcohol-related cues as well as enhancement of prefrontal control mechanisms in the right inferior frontal gyrus. The study design incorporate three experimental groups that either were instructed to downregulate a ventral striatum signal, upregulate the right inferior frontal gyrus, or upregulate negative functional connectivity between these two structures. In two active control groups participants were instructed to either up- or downregulate the primary auditory cortex. We did not find an effect of ventral striatal downregulation or negative connectivity feedback, and a reduced striatal activation in the right inferior frontal gyrus upregulation group was accompanied by concurrent lower activation in the target structure, suggesting that our intended modulation approaches were not effective. Identified problems that might have contributed to this unexpected outcome might have been the use of continuous feedback presentation that potentially confuses regulation target and reward processing in the ventral striatum, counterintuitive regulation directions, a lack of explicit strategy guidance and transparency about the targeted process, and generally the difficulty to recruit a sufficient number of eligible voluntary participants for a well-powered study with a complex design. These insights emphasize the complex challenges of real-time fMRI neurofeedback interventions for the treatment of substance use disorders and could provide guidance for the development of more effective future approaches.

Background. Something in discourse with a person experiencing psychosis often "feels off" before formal assessment is completed, yet this disturbance has not been quantified at the level of ongoing dyadic conversation. Prior work has largely treated patient speech in isolation, limiting our capacity to measure how communicative disruption emerges within clinical exchange. Methods. We applied a three-level decomposition of conversational alignment in 109 patients with psychotic disorders (26 female) and 60 healthy controls (22 female) at baseline and 12 months (n = 115). Register divergence (dAUCnorm) captured lexical distance between interviewer and patient; embedding-based synchrony (rembed) measured semantic trajectory coupling; within-speaker coherence was computed separately for each speaker. We used linear mixed-effects models adjusted for timepoint and participant clustering. Results. Patients showed significantly greater lexical-semantic divergence from the interviewer (d = 0.48, p < .001) and reduced embedding-based synchrony (d = -0.59, p < .001), both effects replicating at each time point. Critically, the interviewer's within-speaker coherence was reduced during conversations with patients (d = -0.33, p = .016), indicating that the disruption extends beyond the patient to the interaction itself. Register divergence tracked impoverished thinking and synchrony tracked disorganized thinking (both FDR-corrected q = .038). Group differences were persistent at 12 months, indicating a partially stable profile. Conclusions. Conversational alignment in psychosis reveals a dyadic failure of semantic coordination that destabilizes the interviewing clinician's coherence even when patient narrative continuity is preserved. These transcript-derived alignment metrics offer a scalable approach to quantifying interpersonal communicative function from routine clinical encounters.

Introduction: Youth mental health presentations are largely heterogenous, making it difficult to match individuals to the most appropriate interventions. Personalised, measurement-based care has the potential to improve clinical decision-making and support shared decision-making, but remains challenging to implement in routine practice. Advances in digital monitoring and causal modelling offer new opportunities to identify individual-level processes driving mental health difficulties and to generate personalised decision-support. This pilot study aims to evaluate the feasibility and acceptability of the Minding Your Mind computational decision-support approach, a newly developed approach integrating routine outcome monitoring, individual-level causal modelling, and personalised feedback to support shared decision-making between young people and their clinicians. Methods and analysis: The study involves two phases. Phase 1 will recruit young people aged 15-25 years and mental health clinicians to participate in workshops to co-design the decision-support approach and its implementation into routine practice. Phase 2 is a prospective, single-arm feasibility study involving young people receiving mental health care and their treating clinicians. Primary outcomes include feasibility, acceptability, appropriateness, and usability of the decision-support approach, assessed via self-report and objective process indicators. Secondary outcomes include changes in use and experiences with shared decision-making, and clinical and functional outcomes. Quantitative analyses will be primarily descriptive, with exploratory pre-post comparisons and sensitivity analyses. Qualitative interviews will explore user experiences and implementation barriers and facilitators. Ethics and dissemination: This study has been approved by the Sydney Local Health District (RPAH Zone) Human Research Ethics Committee (X25-0341). All participants will provide informed consent prior to participation. Findings will be disseminated through peer-reviewed publications, conference presentations, and accessible summaries co-developed with young people with lived experience.

BackgroundMost studies seeking to identify youth at increased risk for depression have developed prediction models using a limited set of risk factors in general population samples. It is unclear whether these models generalize to high-risk youth. Here, we developed machine learning algorithms to predict first-onset depression in youth from the general population and high-risk youth with attention-deficit/hyperactivity disorder (ADHD). MethodsParticipants were 4803 unrelated children from the ABCD study with no prior mood disorder and complete data at baseline (age 9-10 years) and 2-year follow-up. Support Vector Machine, Random Forest, and Elastic Net models were used to predict first-onsets from clinically-relevant risk factors spanning mental and physical health, cognitive, dispositional, interpersonal, and socio-environmental domains. Predictive performance was evaluated in the full sample and separately in participants with ADHD (N=584, 12.16%). ResultsModels trained on the full sample achieved good discriminative predictive power (area under the curve [AUC]=0.70 and accuracy=0.70-0.82). Predictors that replicated across models included earlier pubertal development, higher behavioral inhibition and aggression, and more time spent passively watching media content. In the ADHD subsample, model performance declined (AUC=0.46-0.61) and predictors only partly overlapped with those identified in the full sample. ConclusionsModels effectively predicted depression in the general population but showed poor generalization to high-risk youth with ADHD, suggesting different risk factors in this group. These findings highlight that models trained in general population samples may not generalize to high-risk groups, pointing to the need for more tailored efforts to predict depression in youth at increased risk.

Confirmation bias impacts judgments and decisions across a range of domains including finance, policy and science. Here we examine whether explicitly labelling information as true or false disrupts a core underlying computational mechanism that can generate this pervasive bias - asymmetric learning. Human participants (Study 1: N=47; Study 2: N=57) completed a 2 alternative forced choice (2AFC) task previously used to test for the presence of confirmation bias. Participants made choices between pairs of options that could win or lose money and received either factual or counterfactual feedback after each choice. We introduced a key novel feature into the task - providing explicit cues that signalled to participants whether feedback they had seen was true (verified) or false (debunked). Learning in response to feedback was attenuated under false compared to true labels but was present under both. Fitting participants choices to computational models enabled us to examine how sensitivity to the feedback varied as a function of both the label (true/false) and confirmation (confirmatory/disconfirmatory). This revealed a distinct pattern of learning rates typical of confirmation bias (enhanced learning from positive prediction errors for chosen options and from negative prediction errors for unchosen options) in response to both true and false labels. The findings highlight how confirmation bias plays an important role in the effectiveness of interventions designed to verify true and/or debunk false claims. Verification is less likely to succeed when information disconfirms prior beliefs. Conversely, debunking false claims is unlikely to succeed when the information confirms ones prior beliefs.

Problematic Sexual Behaviour (PSB) is defined as difficult to control recurrent sexual behaviours that continue despite adverse consequences, leading to social and functional impairment. There is debate whether PSB is a disorder of compulsion or addiction; PSB often co-occurs with neuropsychiatric disorders, but further elucidation regarding underlying biology is required. A deficiency in reward neurotransmitter systems (reward deficiency syndrome: RDS) may underlie a shared vulnerability to addiction. We conducted the first case-control genome wide association study (GWAS) of PSB in patients (n=448), and comparison participants with (n=196) and without PSB (n=1488). We used polygenic risk scores (PRS) to test genetic overlap with related psychiatric, behavioural and personality phenotypes. Three models were used: 1) All-PSB (patient + comparison) vs. controls, 2) Patient-PSB vs controls, and 3) RDS (yes/no). Results suggested genetic overlap of PSB with psychiatric conditions, with PRS for major depression, substance use, and others predicting PSB status. PRS for related behavioural phenotypes (e.g., externalizing, age at first sex, number of lifetime sexual partners) and personality traits also predicted PSB. The patient model showed stronger associations than the All-PSB model, and RDS had both shared and distinct genetics with PSB. As expected with the sample size, only suggestive hits were observed with single variant and gene-based tests. PSB may share genetic mechanisms with various conditions. Further research in larger cohorts is needed to better understand the underlying genetics and environmental factors involved, and to improve diagnostic classification, intervention and treatment prospects.

BackgroundThe mechanisms underlying pharmacological treatments for stimulant use disorders are poorly understood. This study examined whether changes in craving, depressive symptoms, and/or impulsivity mediate treatment effect in pharmacotherapy with combined naltrexone and bupropion for methamphetamine use disorder. MethodsThe study was based on secondary analysis of data from the Accelerated Development of Additive Pharmacotherapy Treatment for methamphetamine disorder (ADAPT-2) trial which randomized adults with methamphetamine use disorder to combined treatment with injectable naltrexone (380 mg every three weeks) plus oral bupropion (450 mg daily) versus placebo. A total of 403 adults with methamphetamine use disorder participated in the first Stage; 225 of first Stage participants in the placebo arm who did not respond to treatment were re-randomized in the second Stage. Mediation effects were examined using longitudinal multi-level structural equation modeling. ResultsNaltrexone-bupropion treatment was associated with decreases in drug use, craving, depressive symptoms, and impulsivity. The indirect effect of treatment through change in craving was significant (self-reported use=-0.21, 95% Credible Interval [CrI]=-0.35, -0.09; drug screen-ascertained use=-0.36, 95% CrI=-0.63, -0.16). Change in craving mediated 56% of the treatment effect on self-reported use and 45% of the effect on drug screen-ascertained use. Estimates for mediated effects for depressive symptoms and impulsivity were smaller in magnitude and non-significant. ConclusionReduction in craving mediates the effect of naltrexone-bupropion pharmacotherapy in methamphetamine use disorder. Craving may serve as a surrogate measure of treatment efficacy in short-term trials and help identify promising candidate medications to be tested in larger and longer-term trials. Trial RegistrationClinicalTrials.gov number: NCT03078075.

BackgroundScreen use is pervasive in childhood and adolescence, yet its role in antisocial behaviour (ASB) remains uncertain. While cross-sectional studies consistently link higher screen use to elevated ASB, longitudinal evidence is mixed, and few studies have controlled adequately for prior behaviour and genetic liability. Thus, it remains unclear whether these associations reflect prospective influences of screen exposure, or underlying vulnerabilities shared with ASB. We investigated whether screen use is a modifiable risk factor or a marker of underlying vulnerability. MethodsWe analysed data from up to 41,562 children in the Norwegian Mother, Father, and Child Cohort Study (MoBa). ASB traits and ICD-10-based conduct disorder (CD) diagnoses were assessed at ages 5, 8 and 14 years, together with screen use (total exposure and modality). Cross-sectional logistic regression models examined associations between screen use and ASB traits/CD at each age, adjusting for sex and parental education. Polygenic risk scores for ASB (PRSASB) were used to assess genetic susceptibility and gene-environment interplay. Lagged logistic models tested whether screen use predicted later ASB, adjusting for prior ASB. Linear mixed-effects models examined developmental patterns across age. ResultsHigher screen use was positively associated with ASB traits and CD across all ages, with dose-response patterns across screen-use modalities. Social media showed the strongest modality-specific association at adolescence. In lagged models, screen use did not predict later ASB after adjustment for prior ASB. Longitudinal models showed significant but attenuating associations across development. PRSASB was independently and additively associated with ASB outcomes but did not interact with screen use. ConclusionsWe found that higher screen use was consistently associated with antisocial outcomes across childhood and adolescence. However, the absence of prospective associations after accounting for prior behaviour, together with independent genetic contributions, suggests that screen use may be better understood as a marker of underlying vulnerability rather than an independent driver of antisocial development.

Human behavioral and mental health outcomes arise from interactions among genetic, environmental, and neurobiological systems. Existing frameworks often model these components jointly, but many treat variables independently or use static representations. This limits their ability to capture system-level dynamics and changes over time. To address this, we developed DynoSys, a unified framework that integrates these signals using three layers: predictive models, relationship exploration models, and mechanism-oriented explanation models. Building on this framework, we introduce DynoSys 2.0, a graph-based temporal modeling approach inspired by the free-energy principle by Karl Friston. In this framework, each individual is represented as a dynamic graph that evolves over time. We hypothesize that healthy development and adverse mental health outcomes correspond to different system states and trajectories. Using longitudinal data from the Adolescent Brain Cognitive Development (ABCD) Study, we construct time-indexed graphs that integrate polygenic risk scores (PRS), multi-domain environmental features, and neuroimaging-derived representations. We study six phenotypes: externalizing behavior, internalizing behavior, and sub-stance use initiation (alcohol, nicotine, cannabis, and any substance). In these graphs, nodes represent domain-level features, and edges capture relationships derived from data-driven feature selection and temporal dependencies. We model graph evolution using recurrent neural networks and graph-temporal learning methods. We also define system-level measures, including graph energy and state transitions, to quantify dynamic patterns. Our results show that DynoSys 2.0 can model behavioral development using longitudinal multi-domain data. The framework achieved meaningful prediction for both continuous behavioral symptoms and substance-use initiation outcomes, but performance differed by outcome type. Externalizing behavior was predicted more accurately than internalizing behavior, and alcohol and any substance initiation showed stronger prediction than cannabis and nicotine initiation. Graph-derived energy measures showed clearer separation for high-versus low-symptom externalizing and internalizing groups, suggesting that continuous behavioral symptoms may be linked to different latent system states over time. Overall, DynoSys 2.0 provides a flexible framework for studying behavioral risk as a dynamic developmental process, while rare-event prediction and detailed graph-level interpretation require further work.

BackgroundMental health systems face escalating demand that exceeds clinician capacity, making accurate severity-based triage a critical bottleneck. Severity assessment guides treatment intensity, resource allocation, and risk management, yet most clinically relevant information remains embedded in unstructured electronic health record (EHR) narratives, limiting its utility for scalable decision support. ObjectivesThis study evaluates whether a single large language model (LLM) can autonomously extract clinical factors from psychiatric EHR narratives, derive predictive weights from those factors, and use the resulting structured representation to predict clinician-implied severity at scale. MethodsFrom a Mayo Clinic repository of more than 2.7 million encounters, 15,000 de-identified psychiatric notes were sampled into a 5,000-patient discovery cohort and a 10,000-patient replication cohort. The same LLM (Llama 3 8B Instruct) extracted 17 background clinical factors and 3 treatment-action factors from each note. Severity reference labels were derived from the treatment-action factors using pre-specified clinical criteria. The LLM independently derived two factor-weight dictionaries from the discovery cohort: one capturing risk-oriented predictors of severe presentations and one capturing protective predictors. Five weighting conditions were then evaluated against the severity labels: the two LLM-derived dictionaries, two controls (LLM-derived variables with randomized weights; clinically irrelevant variables with arbitrary weights), and an unweighted zero-shot baseline. Performance was assessed across 928 valid iterations in the replication cohort. ResultsLLM-derived structured conditions significantly outperformed all controls and the baseline, with statistically equivalent performance between the two structured conditions. Improvements in precision and recall were balanced, indicating gains in discriminative capacity rather than threshold shifts. The variables and weights the LLM derived as predictors of severe presentations aligned closely with established clinical determinants of psychiatric severity. ConclusionA single LLM can derive clinically meaningful factor weights from unstructured EHR narratives and use them to predict psychiatric severity at scale, supporting a viable path toward interpretable, scalable triage in resource-constrained mental health systems.