Co-existence of clinical high-risk for psychosis and bipolar disorder: a multi-dimensional psychopathological analysis

Abstract

Background

While the clinical high-risk for psychosis (CHR-P) has been extensively studied and the clinical high-risk for bipolar disorder (CHR-BD) has gained increasing attention, overlapping symptoms—such as mood instability and subclinical psychotic features—complicate diagnostic boundaries. We aimed to investigate the psychopathological characteristics of high-risk individuals to address the feasibility of early-stage diagnostic differentiation. We hypothesized that CHR-P and CHR-BD could co-occur and that symptom structure would differ depending on overlap status.

Methods

We recruited 231 participants without formal psychiatric diagnoses from a prospective cohort established at the psychiatry clinic of Seoul National University Hospital. We identified CHR-P and CHR-BD using the Structured Interview for Psychosis-Risk Syndromes (SIPS) and the Bipolar Prodrome Symptom Interview and Scale (BPSS), respectively. Participants were classified into CHR-P-only, CHR-BD-only, and CHR-P/BD-overlap groups. Participants were then evaluated across multiple psychopathological domains, including delusional ideation, obsessive-compulsive symptoms, perceptual abnormalities, depressive and manic symptoms, and global clinical severity. Symptom severity, inter-domain correlations, and symptom network structures were analyzed and compared across groups.

Results

The overlap was substantial: among 158 individuals meeting CHR-P criteria, 106 (67.1%) also met CHR-BD criteria, and among 179 individuals meeting CHR-BD criteria, 106 (59.2%) also met CHR-P criteria. The subgroup distribution was 52 CHR-P-only, 73 CHR-BD-only, and 106 CHR-P/BD-overlap. Symptom structures differed by overlap status. CHR-P-only showed significantly greater social withdrawal compared to other groups. CHR-P/BD-overlap had more severe mood, delusional, obsessive-compulsive, and hallucinatory symptoms. Correlation analysis showed weaker interconnections among symptom domains in CHR-P-only. Network analysis revealed four fragmented communities in CHR-P-only, two tightly integrated communities in CHR-BD-only, and partial integration across three communities in CHR-P/BD-overlap. Central symptoms varied: paranoia in CHR-P-only; mood symptoms and obsessive thoughts in CHR-BD-only; mood symptoms and perceptual abnormalities with religious themes in CHR-P/BD-overlap.

Conclusions

High-risk states for psychosis and bipolar disorder frequently co-occur, suggesting that early psychopathology extends beyond the prodrome of any single disorder. Distinct patterns of symptom networks across high-risk subgroups underscore the value of dimensional and spectrum-based approaches to early psychopathology.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12888-026-08220-9.

Background

A prodrome refers to a subthreshold state that precedes the onset of a diagnosable psychiatric disorder [1–5]. With growing emphasis on early detection and prevention of major mental illnesses, increasing efforts have focused on delineating prodromal phases and developing reliable diagnostic instruments [6–8]. The most well-established example is the clinical high-risk for psychosis (CHR-P). CHR-P is defined by psychotic symptoms that do not meet the Diagnostic and Statistical Manual of Mental Disorders (DSM) or the International Classification of Diseases (ICD) diagnostic thresholds in duration or severity [9]. The most widely used assessment tools for CHR-P include the Comprehensive Assessment of At-Risk Mental States (CAARMS) and the Structured Interview for Psychosis-Risk Syndromes (SIPS) [10, 11]. The criteria for CHR-P are primarily based on the presence of attenuated positive symptoms, such as unusual thought content or delusional thinking, persecutory ideas, grandiosity, hallucinations, and disorganized communication [12, 13].

Similarly, the clinical high-risk for bipolar disorder (CHR-BD), also referred to as the bipolar prodrome, has been proposed. It is characterized by episodic mood fluctuations that are typically subthreshold in intensity and potentially prodromal, with or without additional risk factors or symptomatology [14]. Subthreshold manic symptoms, mood swings, and depressive episodes in individuals with a family history of bipolar disorder are considered strong indicators [15–17]. The Bipolar Prodrome Symptom Interview and Scale (BPSS) was developed to characterize CHR-BD. This tool includes mania, depression, and general symptom indices [18].

CHR-P is defined by attenuated positive psychotic symptoms and is assessed using the SIPS or CAARMS, whereas CHR-BD is characterized by episodic subthreshold mood symptoms and is evaluated using the BPSS. Nevertheless, both constructs share a common framework: they identify individuals with subthreshold psychopathology who experience functional distress but have not yet met criteria for a full psychiatric diagnosis.

Despite growing research, the clinical significance and diagnostic boundaries of these risk states remain a matter of debate. Identifying CHR-P using the SIPS implies that positive symptoms alone define psychosis risk. However, positive symptoms such as hallucinations and delusions are not specific to schizophrenia; rather, they can also be experienced in other conditions, including bipolar disorder [19, 20]. Likewise, mood instability is a common feature among patients with schizophrenia or personality disorders [21, 22]. Positive psychotic symptoms and mood instability are increasingly recognized as transdiagnostic phenomena [3, 21–23].

Such symptom overlaps challenge the assumption that CHR-P and CHR-BD represent discrete and disease-specific risk states. Furthermore, longitudinal studies have shown that transition rates from CHR states to full-threshold disorders are lower than previously expected, and that clinical trajectories are often heterogeneous rather than disorder-specific [24]. CHR-P and CHR-BD may capture meaningful early features of psychosis and bipolarity, yet the boundaries between the two high-risk states may not be entirely distinct in clinical reality.

The present study aimed to investigate the psychopathological features of help-seeking individuals who had not met full diagnostic thresholds but exhibited subthreshold psychotic and episodic mood symptoms, as classified using the SIPS and BPSS. Through comprehensive assessments across multiple symptom domains, we aimed to characterize both overlapping and distinct features within this high-risk population.

We hypothesized that CHR-P and CHR-BD are not entirely distinct risk states and may co-occur, and that symptom structure—the patterns of relationships among symptoms—would differ according to overlap status.

Methods

Participants

All participants were recruited from an ongoing prospective observational cohort study at the psychiatry clinic of Seoul National University Hospital, which began in July 2020. Help-seeking individuals without formal diagnoses of major psychiatric illnesses are enrolled to track longitudinal changes in symptom profiles and diagnostic transitions. Baseline evaluations were conducted at registration, and follow-up assessments were scheduled at 6, 12, and 24 months. Only baseline data were used in this study. Subjects with an intellectual quotient (IQ) below 70 and those with psychiatric symptoms due to physical or neurological conditions were excluded. This study was approved by the Institutional Review Board of Seoul National University Hospital (IRB No. 2006-088-1132). All participants provided written informed consent. This study was conducted in accordance with the Declaration of Helsinki.

Diagnosis

Participants with a fully syndromal psychiatric diagnosis, such as schizophrenia, bipolar disorder, or major depressive disorder, were excluded based on DSM-5 criteria. Afterward, all subjects were assessed with SIPS and BPSS to identify clinical high-risk for psychosis (CHR-P) and bipolar disorder (CHR-BD), respectively.

The SIPS is a clinician-administered, semi-structured interview used to assess CHR-P [11]. It uses the Scale of Psychosis-Risk Symptoms (SOPS), which rates each symptom on a 7-point scale from 0 (absent) to 6 (severe/psychotic). Scores of 3 to 5 on positive symptom items indicate subthreshold or high risk, and a score of 6 reflects psychotic-level intensity. In this study, CHR-P was defined by a score of 3 to 5 on at least one positive symptom item, with evidence of symptom progression or worsening over the past year. We used the Korean version of the SIPS, which has established reliability and validity [25].

The BPSS is a semi-structured interview developed to assess attenuated or subsyndromal manic symptoms. It includes 10 core manic symptoms with severity measured on a 0 (absent) to 6 (severe) scale [18]. The risk state is defined by a score of 3 or greater in mood elevation or irritability, accompanied by at least one additional attenuated manic symptom rated 3 or greater, presence of subjective distress or functional impairment, and absence of a history of (hypo)manic episode. In cases where irritability was the only core symptom present, distractibility or increased energy/goal-directed activity alone were considered insufficient, and an additional symptom criterion was required. In this study, we used the Korean version of the Bipolar Prodrome Symptom Scale–Abbreviated Prospective (BPSS-AP) [26] and operationalized this risk state as clinical high-risk for bipolar disorder (CHR-BD) [14]. The BPSS was selected because it was originally modeled after SIPS and uses a closely parallel 0–6 severity rating scale, allowing a more consistent assessment framework [18].

Based on SIPS and BPSS classifications, participants were categorized into three subgroups: CHR-P-only, CHR-BD-only, and CHR-P/BD-overlap.

Measures

The goal of this study was to investigate psychopathological characteristics of the high-risk population, focusing on their subjective experiences using a dimensional approach. Psychometric tools were employed to evaluate multiple symptom domains, enabling the detection of even subtle deviations in the general population. Delusional thoughts and perceptual abnormalities were evaluated as they represent core positive symptoms relevant to psychosis risk, and mood symptoms were investigated as core affective dimensions relevant to bipolarity. Obsessive-compulsive (OC) symptoms were also assessed as accumulating evidence suggests that OC symptomatology is frequently observed across schizophrenia-spectrum and bipolar disorders, and may represent a transdiagnostic feature [27, 28].

Delusional thoughts were assessed using the Korean version of the 21-Item Peters et al. Delusions Inventory (PDI-21), which measures delusional thoughts in the general population [29, 30]. It evaluates the presence of delusional symptoms across three domains—distress, preoccupation, and conviction. The total score, as the sum of three domains, was used in the analysis.

OC symptoms were evaluated using the Korean version of Obsessive-Compulsive Inventory-Revised (OCI-R), an 18-item self-report measure designed to assess symptoms of obsessive-compulsive disorder (OCD) [31, 32].

We used the Korean version of Launay–Slade Hallucination Scale-Revised (LSHS) to assess perceptual abnormalities [33–35]. LSHS is a 12-item self-report questionnaire evaluating hallucinatory predisposition. Three subscales include vivid/intrusive mental events, hallucinations with a religious theme, and hallucinatory experiences.

The Korean version of Bipolar Depression Rating Scale (BDRS) was used to measure the mood symptoms [36, 37]. The BDRS consists of 20 items evaluating depressive and mixed symptoms of bipolar depression.

Additionally, the Clinical Global Impressions Scale – Severity (CGI-S) [38] and the Young Mania Rating Scale (YMRS) [39, 40] were administered to assess global severity and manic symptoms.

Statistical analysis

Comparative analyses were performed across the three high-risk subgroups. Demographic and clinical variables of the subjects were analyzed using the Kruskal-Wallis test followed by Dunn’s post-hoc test for non-normally distributed continuous and ordinal scale variables. Categorical variables were compared using the χ² test or Fisher’s exact test, as appropriate.

Exploratory factor analysis was conducted separately for the PDI-21, OCI-R, LSHS, and BDRS. We used the principal component method with varimax rotation to identify underlying symptom dimensions and to reduce item-level data into interpretable factors. For each dataset, Spearman correlation matrices were computed to account for ordinal data. The number of factors was determined based on scree plot inspection (elbow method), with the Kaiser criterion considered as a supplementary reference [41]. Factor loadings were calculated for each item, and individual-level factor scores were derived by weighting item responses according to their respective factor loadings. For interpretability, each factor was labeled using items with the highest loadings. These factors were included in the correlation plot to investigate the relationships between variables across the datasets.

We constructed a network model using the qgraph R package to analyze data from the PDI-21, OCI-R, LSHS, BDRS, and YMRS scales [42]. Each symptom factor was represented as a circular node, and associations between factors were depicted as edges, with edge thickness reflecting their strength.

The network was estimated as a Gaussian graphical model representing partial correlations among variables, using the graphical least absolute shrinkage and selection operator (glasso) with Extended Bayesian Information Criterion (EBIC) model selection. Nodes were colour-coded according to their community structure, which was identified using the Louvain algorithm for community detection with a resolution parameter of 0.4 [43].

To characterize network topology, centrality indices—including strength, betweenness, closeness, and eigenvector centrality—were calculated using the qgraph and igraph R packages [42, 44]. The accuracy of edge weights and the stability of centrality estimates were assessed using the case-dropping bootstrap procedure, yielding correlation stability (CS) coefficients, as implemented in the bootnet R package [45].

All analyses were performed using R software (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria).

Results

Baseline characteristics and the identification of high-risk groups

A total of 231 subjects were included in the analysis, of whom 119 (51.5%) were female. The average age was 23.3 years (SD = 6.0; range: 13–47), and 80 (34.6%) had a family history of psychiatric illness (Table 1).

Table 1.

Comparison of sociodemographic and clinical characteristics between groups

Variables	CHR-P-only (N = 52)	CHR-BD-only (N = 73)	CHR-P/BD-overlap (N = 106)	Total (N = 231)	p-value	Post-hoc
Age, mean (SD)	25.2 (7.5)	23.2 (5.7)	22.3 (5.3)	23.3 (6.0)	0.108
Sex, n (%)					0.919
Male	26 (50.0%)	34 (46.6%)	52 (49.1%)	112 (48.5%)
Female	26 (50.0%)	39 (53.4%)	54 (50.9%)	119 (51.5%)
Familial history, n (%)					0.338
Yes	16 (30.8%)	23 (31.5%)	41 (38.7%)	80 (34.6%)
No	36 (69.2%)	50 (68.5%)	65 (61.3%)	151 (65.4%)
Years of education, mean (SD)	13.5 (2.1)	13.3 (2.3)	13.0 (2.3)	13.2 (2.3)	0.501
CGI-S, mean (SD)	3.7 (1.0)	3.5 (1.1)	3.6 (1.0)	3.6 (1.0)	0.511
PDI-21 total, mean (SD)	66.3 (50.7)	72.2 (58.0)	86.3 (58.8)	77.3 (57.2)	0.081
Grandiose/bizarre beliefs	19.5 (20.7)	17.5 (18.9)	22.7 (21.3)	20.3 (20.5)	0.179
Paranoid features	25 (19.8)	25.7 (20.9)	32.7 (23)	28.8 (21.9)	0.060
Passivity experiences	13.4 (11.9)	16.8 (14.7)	20.5 (15.6)	17.7 (14.8)	0.015*	P vs. P/BD
OCI-R total, mean (SD)	18.4 (13.3)	20.5 (13.7)	24.4 (14.3)	21.8 (14.1)	0.025*	P vs. P/BD
Compulsions	7.1 (6.0)	7.8 (5.9)	9.5 (6.5)	8.4 (6.3)	0.037*	P vs. P/BD
Obsessing	6.9 (4.4)	7.5 (4.7)	8.8 (4.7)	7.9 (4.7)	0.036*	P vs. P/BD
Hoarding	5.3 (4.3)	6.7 (4.5)	7.6 (4.7)	6.8 (4.6)	0.013*	P vs. P/BD
LSHS total, mean (SD)	14.9 (9.9)	13.7 (11.1)	18.3 (10.7)	16.1 (10.8)	0.009*	BD vs. P/BD
Vivid mental events	7.3 (5.1)	7.6 (5.4)	9.6 (5.4)	8.4 (5.4)	0.014*	P vs. P/BD, BD vs. P/BD
Religious themes	2.1 (2.7)	1.4 (2.8)	2.4 (2.8)	2.0 (2.8)	0.017*	BD vs. P/BD
Auditory/visual hallucinations	5.7 (4.6)	4.9 (5.0)	7.0 (4.9)	6.1 (4.9)	0.008*	BD vs. P/BD
BDRS total, mean (SD)	18.4 (8.7)	19 (7.9)	21.3 (7.2)	20.0 (7.9)	0.042*	P vs. P/BD
Depression	7.7 (4.6)	7.9 (4.4)	9.0 (4.2)	8.4 (4.4)	0.115
Mixed features	3.0 (2.2)	3.5 (1.7)	3.9 (1.6)	3.6 (1.8)	0.001*	P vs. BD, P vs. P/BD
YMRS total, mean (SD)	8.8 (5.6)	8.2 (5.4)	9.9 (4.6)	9.1 (5.1)	0.062

* p < 0.05

Abbreviations: CHR-P-only (P), clinical high-risk for psychosis; CHR-BD-only (BD), clinical high-risk for bipolar disorder; CHR-P/BD-overlap (P/BD), clinical high-risk for both psychosis and bipolar disorder; CGI-S, Clinical Global Impression-Severity; PDI-21, 21-Item Peters et al. Delusions Inventory; OCI-R, Obsessive-Compulsive Inventory-Revised; LSHS, Launay–Slade Hallucination Scale; BDRS, Bipolar Depression Rating Scale; YMRS, Young Mania Rating Scale

Fig. 1 illustrates the distribution of participants evaluated with the SIPS and BPSS to identify CHR-P and CHR-BD, respectively. 158 subjects met the criteria for CHR-P, of whom 106 (67.1%) were concurrently identified as CHR-BD. Conversely, 179 participants met the criteria for CHR-BD, of whom 106 (59.2%) also met the CHR-P criteria. Consequently, 52 individuals were classified as the CHR-P-only group, 73 as the CHR-BD-only group, and 106 individuals as the CHR-P/BD-overlap group

Fig. 1.

illustrates the distribution of participants evaluated with the SIPS and BPSS to identify CHR-P and CHR-BD, respectively. 158 subjects met the criteria for CHR-P, of whom 106 (67.1%) were concurrently identified as CHR-BD. Conversely, 179 participants met the criteria for CHR-BD, of whom 106 (59.2%) also met the CHR-P criteria. Consequently, 52 individuals were classified as the CHR-P-only group, 73 as the CHR-BD-only group, and 106 individuals as the CHR-P/BD-overlap group

Comparing sociodemographic and clinical characteristics between groups

Table 1 represents group comparisons of sociodemographic characteristics and clinical variables, including total and factor scores from each psychopathological scale. There were no significant differences in age, sex, familial history, and years of education between groups. Overall symptom severity, measured by the CGI-S, did not differ significantly among the groups (p = 0.511).

Multiple domains of psychopathology were assessed using the PDI-21, OCI-R, LSHS, BDRS, and YMRS. We compared total scores as well as factor scores derived from principal-component exploratory factor analysis. Item-level results are presented in Supplementary Table S1.

In the PDI-21, three factors were identified: [1] grandiose/bizarre beliefs [2], paranoid features, and [3] passivity experiences. The OCI-R was reduced to three factors: [1] compulsions (washing, ordering, checking, and neutralizing) [2], obsessing, and [3] hoarding. The LSHS was also reduced to three factors: [1] vivid mental events [2], religious themes, and [3] auditory/visual hallucinations. Finally, the BDRS was reduced to two factors: [1] depression and [2] mixed features. Factor loadings for each scale are presented in the Supplementary Table S2. Model fit was examined using commonly reported indices. The PDI-21 showed chi-square (χ²) = 486.733 (p < 0.001), comparative fit index (CFI) = 0.782, Tucker–Lewis index (TLI) = 0.753, and standardized root mean square residual (SRMR) = 0.078; the OCI showed χ² = 437.303 (p < 0.001), CFI = 0.823, TLI = 0.795, and SRMR = 0.073; the LSHS showed χ² = 183.617 (p < 0.001), CFI = 0.879, TLI = 0.849, and SRMR = 0.066; and the BDRS showed χ² = 511.095 (p < 0.001), RMSEA = 0.094, CFI = 0.706, TLI = 0.670, and SRMR = 0.093.

Total PDI-21 scores did not differ significantly among the three groups (p = 0.081). Factor scores of the PDI-21 revealed significant differences in the passivity experiences factor (p = 0.015), with the CHR-P/BD-overlap group showing the highest scores. Among individual items, the CHR-P/BD-overlap group reported higher scores on feeling looked at oddly because of one’s appearance (p = 0.001), as well as on items related to the end of the world belief (p = 0.031) and alien thoughts (p = 0.031).

OC symptoms were assessed using the OCI-R, and the total score showed a significant difference among the groups (p = 0.025). Post-hoc analysis revealed that the CHR-P/BD-overlap group exhibited higher levels of OC symptoms compared to the CHR-P-only group. All factor scores—compulsions, obsessing, and hoarding—showed significant differences (p = 0.037, 0.036, and 0.013, respectively), with the CHR-P/BD-overlap group scoring the highest. The CHR-P/BD-overlap group also reported significantly higher scores on individual items — saving many things, upset if others change arrangements, avoiding throwing things away, and nasty thoughts (p = 0.021, 0.004, 0.021, and 0.046, respectively).

Perceptual abnormalities were assessed using the LSHS, which showed significant differences among the three groups (p = 0.009). Factor scores also revealed significant differences in vivid mental events, religious themes, and auditory/visual hallucinations, with the CHR-P/BD-overlap group scoring highest across all factors (p = 0.014, 0.017, and 0.008, respectively).

The total scores of the BDRS showed a significant difference (p = 0.042), with the CHR-P/BD-overlap group also scoring the highest. Factor scores revealed a significant difference in mixed features, with the CHR-P-only group scoring the lowest (p = 0.001). CHR-P-only group scored lowest in depressed mood and lability, while scoring highest in social impairment (i.e., reduced social interaction) (p = 0.001, < 0.001, and 0.005, respectively). CHR-P/BD-overlap group scored highest in anxiety, anhedonia, and suicidal ideation (p = 0.009, 0.027, and 0.025, respectively), while the CHR-BD-only group scored lowest in psychotic symptoms (p = 0.046).

YMRS did not differ significantly among the three groups (p = 0.062). The CHR-P-only group showed the lowest score in sleep deprivation (p = 0.019), and the CHR-BD-only group scored the lowest in the content of thoughts item, which assesses the presence of hyper-religiosity, grandiosity, delusions, or other unusual or grandiose ideas (p = 0.012).

Correlation between psychiatric symptoms from different domains

We explored correlations among different symptom domains, as shown in Fig. 2. The CHR-P-only group showed lower inter-domain symptom correlations, whereas the CHR-BD-only and CHR-P/BD-overlap groups showed more widespread inter-domain correlations. For example, while OC symptoms were positively correlated to perceptual abnormalities and delusions in CHR-BD-only and CHR-P/BD-overlap, such relationships were not consistently observed in the CHR-P-only. This pattern becomes more apparent when comparing the total number of significant correlations across groups: CHR-P-only had 37, CHR-BD-only had 58 and the CHR-P/BD-overlap group had 54.

Fig. 2.

Correlation of psychopathological characteristics. The Spearman correlation coefficients between psychopathological characteristics are depicted for (A) CHR-P-only, (B) CHR-BD-only, and (C) CHR-P/BD-overlap. CHR-P-only group exhibited a unique pattern with a lower degree of correlation between symptoms across different domains. Correlation values with statistical significance were presented. Red indicates positive correlation, and blue indicates negative correlation. Darker colours indicate higher correlations. Abbreviations: PDI-Gr/Bz, grandiose/bizarre beliefs; PDI-Para, Paranoid features; PDI-Pass, passivity experiences; OCI-Comp, compulsive behavior (including washing, ordering, checking, and neutralizing); OCI-Obs, obsessing; OCI-Hrd, hoarding; LSHS-Vvd, vivid mental events; LSHS-Rel, religious themes; LSHS-Hal, auditory/visual hallucinations; BDRS-Dep, depression; BDRS-Mix, mixed features

Network structures of symptom factors

We constructed network models for the CHR-P-only, CHR-BD-only, and CHR-P/BD-overlap groups using the PDI-21, LSHS, OCI-R, BDRS, and YMRS scales. Figure 3 presents these models and their centrality indices.

Fig. 3.

Network structures of psychiatric symptom factors. Network models of psychiatric symptom factors are shown for (A) CHR-P-only, (B) CHR-BD-only, and (C) CHR-P/BD-overlap. In the CHR-P-only group, symptom factors segregated into four distinct communities corresponding to mood symptoms, perceptual abnormalities, delusional ideation, and OC symptoms, with relatively weak interconnections across domains, indicating a fragmented network structure. In contrast, the CHR-BD-only group exhibited a more cohesive organization, with symptom factors integrated into two densely interconnected communities. The CHR-P/BD-overlap group demonstrated an intermediate pattern, comprising three communities: one dominated by mood-related symptoms (BDRS and YMRS), a second clustering obsessive–compulsive symptoms and delusional ideation (OCI-R and PDI factors), and a third consisting primarily of perceptual abnormalities (LSHS factors). Centrality measures (strength, betweenness, closeness, and eigenvector centrality) are displayed to the right of each network, illustrating the relative importance of individual symptom factors within each group. Abbreviations: PDI-Gr/Bz, grandiose/bizarre beliefs; PDI-Para, Paranoid features; PDI-Pass, passivity experiences; OCI-Comp, compulsive behavior (including washing, ordering, checking, and neutralizing); OCI-Obs, obsessing; OCI-Hrd, hoarding; LSHS-Vvd, vivid mental events; LSHS-Rel, religious themes; LSHS-Hal, auditory/visual hallucinations; BDRS-Dep, depression; BDRS-Mix, mixed features

In the CHR-P-only group (Fig. 3A), symptoms segregated into four communities: mood symptoms (YMRS and BDRS factors), perceptual abnormalities (LSHS factors), delusional ideation (PDI-21 factors), and OC symptoms (OCI-R factors). These domains showed relatively weak interconnections, indicating a fragmented network structure.

In contrast, the CHR-BD-only group (Fig. 3B) exhibited a more cohesive network, with two densely interconnected communities. For the CHR-P/BD overlap group (Fig. 3C), the network demonstrated partial integration, comprising three communities: one dominated by mood symptoms (YMRS and BDRS factors), a second consisting of perceptual abnormalities (LSHS factors), and a third encompassing delusional ideation and OC features (PDI-21 and OCI-R factors).

Centrality analyses further highlighted distinct organizing principles across groups. In the CHR-P-only group, the PDI paranoia factor showed the highest centrality, underscoring its dominant role in structuring the symptom network. In contrast, mood-related symptoms (BDRS depression and mixed factors, and YMRS) exhibited low centrality, suggesting that affective symptoms were relatively peripheral in this group.

In the CHR-BD-only group, mood-related factors—particularly the BDRS depression and mixed features—emerged as the most central nodes, indicating that affective symptoms played a key organizing role. The OCI-R obsessing factor also showed elevated centrality, suggesting that intrusive obsessive thoughts contributed meaningfully to network connectivity.

In the CHR-P/BD-overlap group, both BDRS depression and mixed features factors demonstrated high centrality across metrics, reflecting the central influence of mood symptoms. Additionally, the LSHS religious themes factor functioned as a prominent hub, bridging perceptual, affective, and cognitive symptom domains.

The network accuracies and stabilities are presented in Supplementary Fig. S1. Edge weight estimates were relatively accurate, and strength centrality showed moderate to strong stability across groups.

Additional network analysis was conducted in the total high-risk population (Supplementary Fig. S2). The overall high-risk network showed a less differentiated structure with two communities identified. The network reflected a mixed pattern of the subgroup structures. Key symptoms included mood symptoms, paranoid features, and obsessive thoughts.

Discussion

Consistent with our hypothesis, a substantial overlap between CHR-P and CHR-BD was observed. Among individuals who met SIPS criteria and were thereby identified as CHR-P, 106 out of 158 (67.1%) were also BPSS-positive and thus met criteria for CHR-BD. Conversely, 106 out of 179 individuals (59.2%) meeting BPSS criteria were SIPS-positive. Overall, 106 out of 231 participants (45.9%) met both criteria, indicating that CHR-P and CHR-BD frequently co-exist.

Previous studies have reported that the symptom profiles of CHR-P and CHR-BD are highly overlapping and difficult to distinguish [13, 46, 47]. These findings suggest that psychosis-risk and bipolar-risk states may stem from shared early vulnerability rather than representing entirely distinct prodromal pathways. Clinically, this overlap underscores the need for careful evaluation of mood-related features even when psychotic symptoms predominate. Indeed, bipolar disorder with psychotic features is frequently misdiagnosed as schizophrenia or other psychotic disorders, and misdiagnosis is more frequent in the early stage of illness [48–50]. Therefore, the presence of positive symptoms should not automatically be attributed to an emerging psychotic disorder; clinicians should also consider potential underlying bipolarity, as mood-related fluctuations may indicate a bipolar-spectrum trajectory rather than pure psychosis.

Subsequently, we categorized the high-risk population into three groups according to the overlap status of the two risk constructs: CHR-P-only, CHR-BD-only, and CHR-P/BD-overlap. In support of our second hypothesis, symptom structures differed across these subgroups.

The CHR-P/BD-overlap group represents individuals with both psychosis- and bipolar-related high-risk features. These individuals exhibited a broad spectrum of highly correlated psychiatric symptoms, with greater symptom severity. Network analysis further supported this interconnected symptom structure. Centrality analysis indicated that mood symptoms constituted the key node within the network. This implies that treating mood symptoms could also reduce psychotic symptoms concurrently. As previously reported, mood-congruent psychotic symptoms tend to fluctuate with the mood state [51], and OC symptoms in bipolar disorder are mood-state dependent as well [52].

In contrast, the CHR-P-only group showed a fragmented symptom organization consisting of four independent communities with weak intercorrelations across domains. This suggests that symptoms in CHR-P may operate as relatively discrete treatment targets. The key symptom of this group was paranoia, which is consistent with evidence that positive symptoms have the highest predictive value for transition to psychosis [53], and that attenuated psychotic symptoms constitute the core diagnostic feature of the SIPS [11]. In a recent study, paranoia was reported by 91% of CHR-P individuals compared with 59% of first-episode psychosis patients, suggesting that paranoia may represent one of the most prominent early markers of psychosis risk [54].

Most symptom item scores were lower in the CHR-P-only group; however, the ‘social isolation’ item showed the highest score across the three groups. This aligns with previous studies that have identified social withdrawal as one of the most prevalent and earliest features observed in CHR-P [55, 56]. Social isolation may reflect not only depressive tendencies but also a negative-symptom dimension, which has been recognized as an important determinant of functional outcomes in both schizophrenia [57, 58] and CHR-P [59]. While positive symptoms remain the key indicator of psychosis risk, systematic assessment of negative symptoms is equally important for early identification.

In contrast to the CHR-P-only group, the CHR-BD-only group exhibited a more integrated symptom network structure. Mood symptoms and obsessive thoughts showed high and relatively evenly distributed centrality, suggesting stronger interconnections between symptom domains. OC symptoms are known to be more common in younger bipolar disorder patients, suggesting that they may reflect an early manifestation of bipolar vulnerability [60, 61]. The co-occurrence of OC and mood symptoms may therefore indicate a bipolar high-risk state.

Schizophrenia and bipolar disorder themselves share substantial overlap in genetic, phenomenological, and therapeutic aspects [12, 62, 63]. Longitudinal studies have documented diagnostic shifts between schizophrenia-spectrum and bipolar-spectrum disorders over time [64, 65]. In this study, we demonstrated that the boundary is even less distinct in the subthreshold high-risk states. Moreover, individuals in the early stage of illness could be pluripotent — that is, they occupy a non-specific period of vulnerability that may evolve into a wide range of psychiatric disorders [66, 67]. Clinicians should therefore evaluate a broader range of diagnostic possibilities rather than anchoring on a single diagnosis.

In psychiatric disorders, no single symptom is pathognomonic for a specific diagnosis [68]. We observed that CHR-P individuals could exhibit attenuated mood instability, whereas CHR-BD individuals could show attenuated psychotic features, indicating overlap in core symptoms across these states. Group differences were characterized more by patterns of symptom co-existence than by isolated symptoms. A broader focus across multiple symptom domains is therefore important in early-stage psychopathology, as clinical approaches may differ depending on the configuration of co-occurring symptoms.

This study has several limitations. The relatively small sample size may have affected statistical power and generalizability. Recruitment from a single institution also introduces potential regional and population-specific biases, which we aim to address through future multi-center studies. In addition, other clinically relevant features, including anxiety, cognitive domains, a history of ADHD, and social functioning, were not comprehensively assessed. Anxiety, in particular, represents an important transdiagnostic dimension, and the use of a dedicated anxiety scale rather than indirect assessment through BDRS items may have provided more detailed characterization. Finally, the cross-sectional design limits conclusions regarding symptom trajectories. Nevertheless, this sample is part of an ongoing prospective study, and forthcoming longitudinal data will enable evaluation of outcomes such as transition to psychosis, symptom persistence, and functional outcomes.

Conclusion

This study examined the characteristics of high-risk individuals with subsyndromal psychiatric symptoms. Consistent with our hypothesis, we observed a substantial overlap between CHR-P and CHR-BD, suggesting that high-risk states extend beyond the prodrome of any single disorder. CHR-P-only was characterized by relatively fragmented symptom networks centered around paranoia, whereas CHR-BD-only showed more tightly integrated structures with mood symptoms and obsessive thoughts as key nodes. The CHR-P/BD-overlap group demonstrated partially integrated network structures, with mood symptoms and perceptual abnormalities with religious themes emerging as central features. To better capture early psychopathology, dimensional and spectrum-based approaches are required. Our cohort is currently being followed longitudinally, and forthcoming data will allow us to examine how these symptom constellations evolve and whether risk-construct overlap predicts transition outcomes.

Electronic Supplementary Material

Below is the link to the electronic supplementary material.

Author contributions

HJ, JYK, JHJ, and SHK conceptualised and designed the study. JJ, HY, JA, JB, JIK, and NK developed the assessment protocol and contributed to data collection. HJ and JHJ conducted the statistical analyses, with overall guidance provided by JYK and SHK. HJ wrote the first draft of the manuscript, and JHJ and SHK contributed substantially through critical revision. All authors contributed to the interpretation of the results, critically reviewed the manuscript, and approved the final version.

Funding

This work was supported by a grant from the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (no. HL19C0026).

Data availability

The datasets generated and analysed during the current study are not publicly available due to data protection regulations. Access to data is limited to the researchers who have obtained permission for data processing. Further inquiries can be made to the corresponding author.

Declarations

Ethical approval and consent to participate

This study was approved by the Institutional Review Board of Seoul National University Hospital (IRB No. 2006-088-1132). Written informed consent was obtained from all participants. The study was conducted in accordance with the ethical principles of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.