Longitudinal evolution of the transdiagnostic prodrome to severe mental disorders: a dynamic temporal network analysis informed by natural language processing and electronic health records.

TY - JOUR

T1 - Longitudinal evolution of the transdiagnostic prodrome to severe mental disorders: a dynamic temporal network analysis informed by natural language processing and electronic health records.

AU - Arribas, Maite

AU - Barnby, Joseph

AU - Patel, Rashmi

AU - McCutcheon, Robert A.

AU - Kornblum, Daisy

AU - Shetty, Hitesh

AU - Krakowski, Kamil

AU - Stahl, Daniel

AU - Koutsouleris, Nikolaos

AU - McGuire, Philip

AU - Fusar-Poli, Paolo

AU - Oliver, Dominic

PY - 2025/1/15

Y1 - 2025/1/15

N2 - Importance: Modelling the prodrome to severe mental disorders (SMD), including unipolar mood disorders (UMD), bipolar mood disorders (BMD) and psychotic disorders (PSY), should consider both the evolution and interactions of symptoms and substance use (prodromal features) over time. Temporal network analysis can detect causal dependence between and within prodromal features by representing prodromal features as nodes, with their connections (edges) indicating the likelihood of one feature preceding the other. In SMD, node centrality could reveal insights into important prodromal features and potential intervention targets. Community analysis can identify commonly occurring feature groups to define SMD at-risk states.Objective: To develop a global transdiagnostic SMD network of the temporal relationships between prodromal features, and to examine within-group differences with sub-networks specific to UMD, BMD and PSYDesign: Retrospective (2-year), real-world, electronic health records (EHR) cohort study. Validated natural language processing algorithms extracted the occurrence of 61 prodromal features every three months from two years to six months prior to SMD onset. To construct temporal networks of prodromal features, we employed generalized vector autoregression panel analysis, adjusting for covariates.Setting: South London and Maudsley NHS Foundation Trust EHRs.Participants: 6,462 individuals with an SMD diagnosis (UMD:2,066; BMD:740; PSY:3,656).Main Outcomes: Edge weights (partial directed correlation coefficients, z) in autocorrelative, unidirectional and bidirectional relationships. Centrality was calculated as the sum of (non autoregressive) connections leaving (out-centrality, cout) or entering (in-centrality, cin) a node. The three sub-networks (UMD, BMD, PSY) were compared using permutation analysis. Community analysis was performed using Spinglass. Results: The SMD network was characterised by strong autocorrelations (0.04  z  0.10), predominantly positive connections, and aggression (cout=.103) and tearfulness (cin=.134) as the most central features. The UMD sub-network showed few significant differences compared to PSY (3.5%) and BMD (0.8%), and BMD-PSY showed even fewer (0.4%). One positive psychotic (delusional thinking-hallucinations-paranoia) and two behavioural communities (aggression-cannabis use-cocaine use-hostility, aggression-agitation-hostility) were the mostcommon.

AB - Importance: Modelling the prodrome to severe mental disorders (SMD), including unipolar mood disorders (UMD), bipolar mood disorders (BMD) and psychotic disorders (PSY), should consider both the evolution and interactions of symptoms and substance use (prodromal features) over time. Temporal network analysis can detect causal dependence between and within prodromal features by representing prodromal features as nodes, with their connections (edges) indicating the likelihood of one feature preceding the other. In SMD, node centrality could reveal insights into important prodromal features and potential intervention targets. Community analysis can identify commonly occurring feature groups to define SMD at-risk states.Objective: To develop a global transdiagnostic SMD network of the temporal relationships between prodromal features, and to examine within-group differences with sub-networks specific to UMD, BMD and PSYDesign: Retrospective (2-year), real-world, electronic health records (EHR) cohort study. Validated natural language processing algorithms extracted the occurrence of 61 prodromal features every three months from two years to six months prior to SMD onset. To construct temporal networks of prodromal features, we employed generalized vector autoregression panel analysis, adjusting for covariates.Setting: South London and Maudsley NHS Foundation Trust EHRs.Participants: 6,462 individuals with an SMD diagnosis (UMD:2,066; BMD:740; PSY:3,656).Main Outcomes: Edge weights (partial directed correlation coefficients, z) in autocorrelative, unidirectional and bidirectional relationships. Centrality was calculated as the sum of (non autoregressive) connections leaving (out-centrality, cout) or entering (in-centrality, cin) a node. The three sub-networks (UMD, BMD, PSY) were compared using permutation analysis. Community analysis was performed using Spinglass. Results: The SMD network was characterised by strong autocorrelations (0.04  z  0.10), predominantly positive connections, and aggression (cout=.103) and tearfulness (cin=.134) as the most central features. The UMD sub-network showed few significant differences compared to PSY (3.5%) and BMD (0.8%), and BMD-PSY showed even fewer (0.4%). One positive psychotic (delusional thinking-hallucinations-paranoia) and two behavioural communities (aggression-cannabis use-cocaine use-hostility, aggression-agitation-hostility) were the mostcommon.

M3 - Article

SN - 1359-4184

JO - Molecular Psychiatry

JF - Molecular Psychiatry

ER -