Background 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 address this by representing prodromal features as nodes, with their connections (edges) indicating the likelihood of one feature preceding the other. Node centrality could reveal insights into important prodromal features and potential intervention targets. We developed a SMD network and compared sub-networks specific to UMD, BMD and PSY.

Methods We analysed 7,049 individuals with an SMD diagnosis (UMD:2,306; BMD:817; PSY:3,926) from the South London and Maudsley NHS Foundation Trust electronic health records. Using validated natural language processing algorithms, we 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. We computed edge weights (correlation coefficients, z) in autocorrelative, unidirectional and bidirectional relationships. Centrality was calculated as the sum of connections leaving (out-centrality, cout) or entering (in-centrality, cin) a node. We compared the three sub-networks (UMD, BMD, PSY) using permutation analysis.

Findings The strongest autocorrelation in the SMD network was tearfulness (z=·10). Unidirectional positive relationships were observed for irritability-agitation (z12=·03), mood instability-tearfulness (z12=·03) and irritability-aggression (z12=·03). Aggression-hostility (z12=·04, z21=·03), delusions-hallucinations (z12=·04, z21=·03) and aggression-agitation (z12=·03, z21=·03) were the strongest bidirectional relationships. The most central features included aggression (cout=·082) and tearfulness (cin=·124). The PSY sub-network showed few significant differences compared to UMD (3·9%) and BMD (1·6%), and UMD-BMD showed even fewer (0·4%).

Interpretations This study represents the most extensive temporal network analysis conducted on the longitudinal interplay of SMD prodromal features. These findings provide further evidence to support early detection services across SMD.

Evidence before this study Preventive approaches for severe mental disorders (SMD) can improve outcomes, however, their effectiveness relies on accurate knowledge of the prodromal symptoms and substance use preceding their onset and how they evolve over time. We searched PubMed from database inception to 26th January 2024 for studies investigating the dynamic prodromes for unipolar mood disorders (UMD), bipolar mood disorders (BMD) or psychotic disorders (PSY) published in English. The search terms were prodrom* AND (depression OR bipolar OR psychosis) AND (timecourse OR dynamic OR “network analysis” OR longitudinal). First, while many studies have investigated the prodromal phases of SMD, particularly for PSY, the majority of studies have taken a cross-sectional rather than longitudinal approach which are unable to detect causal dependence between and within prodromal symptoms and substance use. Second, there are no studies focusing on the evolution of features during the prodromal period. Finally, studies have focused on diagnosis-specific analyses, considering UMD, BMD or PSY alone, limiting the possibility for comparison between them.

Added value of this study We have used a temporal network analysis approach, in combination with a large electronic health record database (n=7,049) and natural language processing, to examine the dynamic evolution of symptoms and substance use in the prodrome to an SMD diagnosis in secondary mental healthcare. This is the largest network analysis investigating prodromal features in SMD, the first assessing longitudinal changes and the first to directly compare the prodromes to UMD, BMD and PSY. Our results add to the growing evidence for a transdiagnostic prodrome to SMD, by showing small differences between UMD, BMD and PSY in how symptoms and substance use evolve over the course of the prodrome.

Implications of all the available evidence Our study explores the patterns of evolution of symptom and substance use events across and within SMD diagnostic groups. We highlight the importance of understanding the dynamic progression of these prodromal features to fully characterise the prodrome to SMD. These findings, together with a growing literature base, also support the potential for broader transdiagnostic early detection services that provide preventive psychiatric care to individuals at risk for SMD.

MA has been employed by F. Hoffmann-La Roche AG outside of the current study. RP has received grant funding from Janssen, and consulting fees from Holmusk, Akrivia Health, Columbia Data Analytics, Boehringer Ingelheim and Otsuka. PFP has received research funds or personal fees from Lundbeck, Angelini, Menarini, Sunovion, Boehringer Ingelheim, Mindstrong, Proxymm Science, outside the current study.

MA is supported by the UK Medical Research Council (MR/N013700/1) and Kings College London member of the MRC Doctoral Training Partnership in Biomedical Sciences. JMB has received funding from the Wellcome Trust (WT228268/Z/23/Z). RP has received funding from an NIHR Advanced Fellowship (NIHR301690) and a Medical Research Council (MRC) Health Data Research UK Fellowship (MR/S003118/1). PFP is supported by #NEXTGENERATIONEU (NGEU), funded by the Ministry of University and Research (MUR), National Recovery and Resilience Plan (NRRP), project MNESYS (PE0000006) A Multiscale integrated approach to the study of the nervous system in health and disease (DN. 1553 11.10.2022).

I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained.

Yes

The details of the IRB/oversight body that provided approval or exemption for the research described are given below:

Clinical Record Interactive Search Permissions (CRIS) received ethical approval as an anonymised dataset for secondary analyses from Oxfordshire REC C (Ref: 23/SC/0257).

I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals.

Yes

I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance).

Yes

I have followed all appropriate research reporting guidelines, such as any relevant EQUATOR Network research reporting checklist(s) and other pertinent material, if applicable.

Yes

The data accessed by CRIS remain within an NHS firewall and governance is provided by a patient-led oversight committee. Subject to these conditions, data access is encouraged and those interested should contact Robert Stewart (robert.stewartkcl.ac.uk), CRIS academic lead. There is no permission for data sharing. Covariance matrices to estimate networks and all analysis code are available on GitHub: https://github.com/m-arribas/network_analysis.git.

https://github.com/m-arribas/network_analysis.git