Introduction
Psychotic disorders affect approximately 2.5–3% of the population and typically emerge in late adolescence or early adulthood (Chang et al., Reference Chang, Wong, Chen, Lam, Chan, Ng, Hung, Cheung, Sham, Chiu, Lam, Lee, Chiang, Chan, Lau, Lee, Leung, Leung, Lau and Bebbington2017b; McGrath et al., Reference McGrath, Al-Hamzawi, Alonso, Altwaijri, Andrade, Bromet, Bruffaerts, de Almeida, Chardoul, Chiu, Degenhardt, Demler, Ferry, Gureje, Haro, Karam, Karam, Khaled, Kovess-Masfety and Zaslavsky2023). The disorders are associated with profound psychosocial impairment and constitute one of the leading causes of disability worldwide (Vos et al., Reference Vos, Barber, Bell, Bertozzi-Villa, Biryukov, Bolliger, Charlson, Davis, Degenhardt, Dicker, Duan, Erskine, Feigin, Ferrari, Fitzmaurice, Fleming, Graetz, Guinovart, Haagsma and Murray2015). Patients with psychotic disorders often continue to encounter difficulties in employment and interpersonal relationships after symptom remission, making psychosocial functioning a critical outcome in psychosis research. Past studies have examined the determinants of psychosocial functioning and found that less severe positive and negative symptoms, shorter duration of untreated psychosis (DUP), less severe cognitive impairment, better premorbid adjustment, being female, higher educational levels, and past employment history significantly predicted better functional outcomes in psychosis (Santesteban-Echarri et al., Reference Santesteban-Echarri, Paino, Rice, González-Blanch, McGorry, Gleeson and Alvarez-Jimenez2017). Specialized early intervention (EI) programs for patients with early psychosis have been developed in the past two decades to improve clinical and functional outcomes (Nordentoft et al., Reference Nordentoft, Rasmussen, Melau, Hjorthøj and Thorup2014). Although EI services have demonstrated some efficacy in improving outcomes for early psychosis patients, a significant proportion of individuals continue to experience functional impairment despite achieving symptomatic remission, implying a significant unmet treatment need for achieving functional recovery (Chang et al., Reference Chang, Tang, Hui, Lam, Chan, Wong, Chiu and Chen2012a, Reference Chang, Chan, Jim, Lau, Hui, Chan, Lee and Chen2015, Reference Chang, Kwong, Chan, Jim, Lau, Hui, Chan, Lee and Chen2016b; Mustafa et al., Reference Mustafa, Malla, Joober, Abadi, Latimer, Schmitz, Jarvis, Margolese, Casacalenda, Abdel-Baki and Iyer2022). Of note, most studies have indicated short-term improvement in functional outcomes among individuals with early psychosis receiving EI services (Correll et al., Reference Correll, Galling, Pawar, Krivko, Bonetto, Ruggeri, Craig, Nordentoft, Srihari, Guloksuz, Hui, Chen, Valencia, Juarez, Robinson, Schooler, Brunette, Mueser, Rosenheck and Kane2018; Ho et al., Reference Ho, Chu, Lui, Lau, Lee, Hui, Chan, Chen and Chang2025). Nevertheless, studies with longer follow-up periods have yielded varying results, with the majority of studies failing to demonstrate improvement in functional outcomes, albeit some showed positive findings regarding employment levels (Chan et al., Reference Chan, Chan, Devlin, Bastiampillai, Mohan, Hui, Chang, Lee and Chen2019). This variability underscores the need for a deeper understanding of the longitudinal course of social-occupational functioning.
The majority of longitudinal studies examining functional outcomes utilized conventional approaches, which often rely on dichotomous outcomes, such as functional recovery, or group mean comparisons at the end of the follow-up (Correll et al., Reference Correll, Galling, Pawar, Krivko, Bonetto, Ruggeri, Craig, Nordentoft, Srihari, Guloksuz, Hui, Chen, Valencia, Juarez, Robinson, Schooler, Brunette, Mueser, Rosenheck and Kane2018; Santesteban-Echarri et al., Reference Santesteban-Echarri, Paino, Rice, González-Blanch, McGorry, Gleeson and Alvarez-Jimenez2017). Such methods tend to obscure individual differences and the dynamic course of functional change over time, failing to capture the complexity and ongoing fluctuations in functioning across individuals. To overcome these limitations, recent research has adopted person-centered analytic methods such as latent growth mixture modeling (LGMM), which allows for the identification of distinct subgroups of patients who share similar functional trajectories, providing a more nuanced perspective on recovery and disability in psychosis (Beunckens, Molenberghs, Verbeke, & Mallinckrodt, Reference Beunckens, Molenberghs, Verbeke and Mallinckrodt2008). However, existing studies on social and social-occupational functional trajectories have predominantly focused on short-term outcomes (3-year follow-up or less), with the majority of these studies identifying four distinct trajectories (Chang et al., Reference Chang, Chu, Kwong, Wong, Hui, Chan, Lee and Chen2018a; du Sert et al., Reference du Sert, Unrau, Dama, Palaniyappan, Shah, Joober, Raucher-Chéné, Malla and Lepage2025; Hall et al., Reference Hall, Holton, Ongur, Montrose and Keshavan2019; Hodgekins et al., Reference Hodgekins, Birchwood, Christopher, Marshall, Coker, Everard, Lester, Jones, Amos, Singh, Sharma, Freemantle and Fowler2015; Salagre et al., Reference Salagre, Grande, Solé, Mezquida, Cuesta, Díaz-Caneja, Amoretti, Lobo, González-Pinto, Moreno, Pina-Camacho, Corripio, Baeza, Bergé, Verdolini, Carvalho, Vieta and Bernardo2021; van der Ven et al., Reference van der Ven, Scodes, Basaraba, Pauselli, Mascayano, Nossel, Bello, Humensky, Susser, Wall and Dixon2020). Some predictors, such as male gender, lower educational attainment, lower socio-economic status, poorer premorbid adjustment, higher levels of negative symptom severity, and ethnic minority status, were found to significantly predict poor social-occupational functioning trajectories (Chang et al., Reference Chang, Chu, Kwong, Wong, Hui, Chan, Lee and Chen2018a; du Sert et al., Reference du Sert, Unrau, Dama, Palaniyappan, Shah, Joober, Raucher-Chéné, Malla and Lepage2025; Hall et al., Reference Hall, Holton, Ongur, Montrose and Keshavan2019; Hodgekins et al., Reference Hodgekins, Birchwood, Christopher, Marshall, Coker, Everard, Lester, Jones, Amos, Singh, Sharma, Freemantle and Fowler2015; Salagre et al., Reference Salagre, Grande, Solé, Mezquida, Cuesta, Díaz-Caneja, Amoretti, Lobo, González-Pinto, Moreno, Pina-Camacho, Corripio, Baeza, Bergé, Verdolini, Carvalho, Vieta and Bernardo2021; van der Ven et al., Reference van der Ven, Scodes, Basaraba, Pauselli, Mascayano, Nossel, Bello, Humensky, Susser, Wall and Dixon2020). One prior research investigated long-term functional trajectories over a 20-year period, focusing only on social functioning (Velthorst et al., Reference Velthorst, Fett, Reichenberg, Perlman, Van Os, Bromet and Kotov2017). It identified four distinct yet stable trajectories of social functioning in first-episode psychosis (FEP) patients (Velthorst et al., Reference Velthorst, Fett, Reichenberg, Perlman, Van Os, Bromet and Kotov2017). In addition, it found that poorer premorbid adjustment was associated with poor social functioning trajectory, which in turn was associated with worse real-life functional outcomes, namely unemployment and inability to live independently, at 20 years of follow-up (Velthorst et al., Reference Velthorst, Fett, Reichenberg, Perlman, Van Os, Bromet and Kotov2017).
Considering that functional impairment in psychosis often persists even after psychotic symptoms have improved and contributes to lifelong socioeconomic burden, it is crucial to determine whether social-occupational trajectory patterns established in the early years of illness can serve as predictors for long-term outcomes. Thus far, no study has explored prospective relationships between early-stage social-occupational trajectories and distal illness outcomes in patients with early psychosis. To address these gaps, the current study leverages a representative cohort with multiple timepoint assessments of young Chinese people with early psychosis from a specialized EI service in Hong Kong. Specifically, we aimed to (1) identify trajectory patterns of functioning, based on the ratings on Social and Occupational Functioning Scale (SOFAS) (Goldman, Skodol, & Lave, Reference Goldman, Skodol and Lave1992) over a 3-year period using LGMM analysis, (2) delineate baseline predictors that distinguish between these trajectory classes, and (3) examine how these trajectory classes relate to long-term functional outcomes at the 12-year follow-up.
Methods
Participants
This study was a 12-year follow-up of a single-blind randomized controlled trial (RCT) comparing a 1-year extension of specialized EI (3-year EI service) with a step-down standard care (SC) (2-year EI service) in patients who have completed a 2-year Early Assessment Service for Young People with Early Psychosis (EASY) program for FEP (Supplementary Figure S1) (Chang et al., Reference Chang, Chan, Jim, Lau, Hui, Chan, Lee and Chen2015, Reference Chang, Kwong, Chan, Jim, Lau, Hui, Chan, Lee and Chen2016b, Reference Chang, Kwong, Kei Lau, So, Man Wong, Kei Chan, Ting Jim, Ming Hui, Wa Chan, Ming Lee and Hai Chen2017a, Reference Chang, Kwong, Or, Lau, Chan, Jim, Hui, Chan, Lee and Chen2018b). In brief, the EASY program was a publicly funded, territory-wide service that provided early assessment and intervention for individuals aged 15–25 years with FEP in Hong Kong (Chung & Chen, Reference Chung, Chen, Chen, Lee, Chan and Wong2013; Tang et al., Reference Tang, Wong, Hui, Lam, Chiu, Chan, Chung, Tso, Chan, Yip, Hung and Chen2010). The service consisted of five clinical teams, each covering a geographically defined catchment area and comprising two psychiatrists, three case managers, and one social worker. The service adopted phase-specific case-management approach in which each FEP patient was assigned with a case manager who provided protocol-based psychosocial interventions (So, Reference So, Chen, Lee, Chan and Wong2013). The FEP patients were assertively followed up for a period of 2 years, after which they were managed by a step-down clinic in the third year of treatment without provision of case-management. Exclusion criteria for the study included intellectual disability, substance-induced psychosis, or psychotic disorder due to a general medical condition.
In the original RCT, 160 Chinese patients with FEP (DSM-IV criteria) were recruited from the EASY program between November 2010 and August 2011, randomly allocated to Extended EI (N = 82) or SC (N = 78), and underwent a 1-year clinical trial (Chang et al., Reference Chang, Chan, Jim, Lau, Hui, Chan, Lee and Chen2015). Participants were assessed at baseline, and at 1 year (end of RCT), and then re-interviewed at 1 and 2 years post-RCT. In the present study, participants were re-assessed 12 years after inclusion in the trial (Chu et al., Reference Chu, Fung, Lei, Wong, Chan, Lee, Hui, Chen and Chang2026a, Reference Chu, Yau, Fung, Fung, Lei, Wong, Chan, Lee, Hui, Chen and Chang2026b). This study examined social-occupational functioning trajectories in 148 participants who completed SOFAS at three or more timepoints, including baseline, and the 1-, 2-, and 3-year follow-up. The current study was approved by the local institutional review boards (UW 21-761) on 21 December 2021, and all participants provided written informed consent.
Assessments
Diagnosis was ascertained using Chinese-bilingual Structured Clinical Interview for DSM-IV (CB-SCID) at baseline, 1-, 3-, and 12-year follow-up (So et al., Reference So, Kam, Leung, Chung, Liu and Fong2003). Interview for the Retrospective Assessment of the Onset of Schizophrenia (IRAOS) was conducted at baseline to assess DUP, age, and mode of onset (Häfner et al., Reference Häfner, Riecher-Rössler, Hambrecht, Maurer, Meissner, Schmidtke, Fätkenheuer, Löffler and van der Heiden1992). Premorbid functioning and personality traits were evaluated by the Premorbid Adjustment Scale (PAS) and the Premorbid Schizoid-Schizotypal Traits (PSST) at RCT study entry, respectively (Cannon-Spoor, Potkin, & Wyatt, Reference Cannon-Spoor, Potkin and Wyatt1982; Foerster, Lewis, Owen, & Murray, Reference Foerster, Lewis, Owen and Murray1991). Psychopathology was assessed using Positive and Negative Syndrome Scale (PANSS) (Kay, Fiszbein, & Opler, Reference Kay, Fiszbein and Opler1987) and Calgary Depression Scale for Schizophrenia (CDSS) (Addington, Addington, & Maticka-Tyndale, Reference Addington, Addington and Maticka-Tyndale1993) at baseline, 1-, 2-, 3-, and 12-year follow-up. Consistent with prior research (Ho et al., Reference Ho, Chang, Kwong, Lau, Chan, Jim, Hui, Chan, Lee and Chen2018), positive, negative, and disorganization symptom dimensions were derived based on the Marder five-factor model on PANSS (Marder, Davis, & Chouinard, Reference Marder, Davis and Chouinard1997). Baseline insight was measured by PANSS item G12 (lack of judgment/insight) at study entry, with score ≤ 3 indicating good insight. Antipsychotic-induced extrapyramidal side-effects were assessed using Simpson–Angus Scale (SAS) (Simpson & Angus, Reference Simpson and Angus1970). Subjective quality of life was evaluated using the Chinese version 36-Item Short Form Survey (SF-36) (Lam, Gandek, Ren, & Chan, Reference Lam, Gandek, Ren and Chan1998; Ware & Sherbourne, Reference Ware and Sherbourne1992). Psychosocial functioning was measured by SOFAS (Goldman et al., Reference Goldman, Skodol and Lave1992) and Role Functioning Scale (RFS) (Goodman, Sewell, Cooley, & Leavitt, Reference Goodman, Sewell, Cooley and Leavitt1993) at baseline, 1-, 2-, 3-, and 12-year follow-up. Employment history was also assessed at baseline, 1, 2, 3, and 12 years after study entry. Sustained employment was defined as continuous employment for the last 12 months of 12-year follow-up.
Training was provided to research assistants about all the assessments used in the current study before subject recruitment. Videotaped interviews of 10 cases selected at baseline were independently rated by all research assistants for the purpose of interrater reliability evaluation. Intraclass correlation coefficients (ICCs) for PANSS positive symptom, negative symptom, and general psychopathology subscales, and CDSS total score were 0.95, 0.90, 0.91, and 0.87 respectively, indicating good interrater reliability. Satisfactory levels of concordance were also observed in SOFAS and RFS total scores, with 0.92 and 0.85, respectively.
Statistical analysis
LGMM was applied to identify latent trajectories of functioning over the initial 3-year follow-up in the current cohort. The LGMM is a data-driven and person-centered approach that identifies population subgroups with similar patterns of change over time (Ram & Grimm, Reference Ram and Grimm2009). In the current analysis, individual class membership was assigned on the basis of SOFAS scores measured at four timepoints over 3 years: at baseline, 1 year, 2 years, and 3 years of follow-up. To determine the optimal growth functions and number of trajectory classes, models with an increasing number of latent classes were fitted to the data and the best-fitting model was selected according to the following goodness-of-fit indices: Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), sample-size-adjusted BIC (SABIC), maximized log-likelihood (LogLik) and entropy. Lower values of AIC, BIC, and SABIC suggest a more parsimonious model, whereas higher values of LogLik and entropy indicate better model fit. To ensure sufficient representativeness of each trajectory class, the sample size of each trajectory class should have a minimum of 5% of the total sample.
Following the establishment of class membership, potential predictors for the identified trajectories were then examined. Univariate analyses were conducted to evaluate the associations between potential baseline predictors and class membership, using univariate multinomial logistic regression. The significant variables (p < 0.05) were included in a multivariable multinomial regression model. To examine whether early-stage functioning latent trajectories differentially predicted functional outcomes at 12-year follow-up, a series of analysis of covariance (ANCOVAs) were conducted with previously identified baseline predictive variables as covariates. The long-term functional outcomes (12-year follow-up) included SOFAS total score and RFS total and sub-scale scores. The recovery outcomes included functional recovery and personal recovery. The operational definition of functional recovery was adapted from our past studies (Chang et al., Reference Chang, Tang, Hui, Lam, Chan, Wong, Chiu and Chen2012a, Reference Chang, Chan, Jim, Lau, Hui, Chan, Lee and Chen2015, Reference Chang, Kwong, Chan, Jim, Lau, Hui, Chan, Lee and Chen2016b, Reference Chang, Kwong, Or, Lau, Chan, Jim, Hui, Chan, Lee and Chen2018b; Chu et al., Reference Chu, Fung, Lei, Wong, Chan, Lee, Hui, Chen and Chang2026a). Patients attained functional recovery if: (1) SOFAS score > 60, RFS ratings on independent living and immediate social network scores >5, and RFS ratings on work productivity and extended social network >4 at 12-year follow up; and (2) sustained employment (full-time or part-time work, or full-time study) and SOFAS score > 60 based on medical record review over the last 12 months of the 12-year follow up. The medical record review procedure was described in our previous study (Chu et al., Reference Chu, Fung, Lei, Wong, Chan, Lee, Hui, Chen and Chang2026a). Following previous research (Chu et al., Reference Chu, Fung, Lei, Wong, Chan, Lee, Hui, Chen and Chang2026a), patients were categorized as achieving personal recovery status if they attained a cutoff score ≥ 45 of the Questionnaire about the Process of Recovery (QPR-15) (Law, Neil, Dunn, & Morrison, Reference Law, Neil, Dunn and Morrison2014). Post-hoc comparison analyses with Tukey correction for multiple comparisons were performed to clarify differences between classes in various long-term outcomes. DUP was log-transformed due to its skewed distribution. All statistical analyses were performed using R version 4.5.0. The level of significance was at p < 0.05.
Results
Characteristics of participants
Out of the initial 160 patients, a total of 148 participants were included in the trajectory analysis and 102 participants completed the 12-year follow-up assessment study. Among the study sample, 52.7% of the included participants were male, and the mean age of onset was 20.3 years (SD = 3.17) and median DUP of the sample was 90 days. Attrition for trajectory analysis showed significant better premorbid adjustment and baseline role functioning in the included participants as compared to the excluded participants (Supplementary Table S1). Attrition for outcome analysis demonstrated that patients who completed 12-year follow-up had fewer PSSTs than dropouts (Supplementary Table S2).
Latent classes of psychosocial functioning trajectories
Table 1 presents fit statistics for the tested LGMM models. Analyses with goodness-of-fit statistics suggested that a four-class solution was the most parsimonious with low SABIC, high Loglik and entropy, and had a minimum of 5% in each class. The four functioning trajectories identified in the four-class solution are illustrated in Figure 1. The first class was referred to as the persistently-good (12.8% [N = 19]) trajectory which was characterized by persistent good functioning over 3 years. The second class was termed the gradually-improved trajectory (10.8% [N = 16]) in which patients with moderate functioning showed steady improvement across 3-year follow-up. The third class was the suboptimal-stable trajectory (48.6% [N = 72]) in which patients displayed stable, moderate levels of functioning over 3 years. The fourth class was referred to as the persistently-poor trajectory (27.7% [N = 41]), which was characterized by slightly improving but low functioning levels throughout 3-year follow-up.
Goodness-of-fit statistics of LGMM models with 1–5 Class solutions of social-occupational functioning trajectories
Table 1. Long description
The table contains 7 columns and 6 rows. The columns are labeled: Number of classes, Loglik, A I C, B I C, S A B I C, Entropy, and Class size in percent.
* Row 1 (1 Class): Loglik -2,126; A I C 4,268; B I C 4,292; S A B I C 4,266; Entropy 1.0000; Class size 100.
* Row 2 (2 Classes): Loglik -2,118; A I C 4,258; B I C 4,291; S A B I C 4,257; Entropy 0.7546; Class sizes 78.38 and 21.62.
* Row 3 (3 Classes): Loglik -2,116; A I C 4,260; B I C 4,302; S A B I C 4,258; Entropy 0.7083; Class sizes 54.05, 26.35, and 19.59.
* Row 4 (4 Classes): Loglik -2,112; A I C 4,259; B I C 4,310; S A B I C 4,256; Entropy 0.7505; Class sizes 12.84, 27.70, 48.65, and 10.81.
* Row 5 (5 Classes): Loglik -2,109; A I C 4,258; B I C 4,318; S A B I C 4,254; Entropy 0.8367; Class sizes 16.89, 4.73, 43.24, 11.49, and 23.65.
Footnotes indicate that higher Loglik and entropy values, and lower A I C, B I C, and S A B I C values, indicate better model fit.
AIC, Akaike information criterion; BIC, Bayesian information criterion; LGMM, latent growth mixing modeling; Loglik, maximized log-likelihood values; SABIC, sample-size adjusted Bayesian information criterion.
a Higher Loglik and entropy indicates better model fit.
b Lower values of AIC, BIC, and SABIC indicate better model fit.
Social-occupational trajectories over 3 years derived from growth mixture modeling.
Figure 1. Long description
The x-axis represents Timepoint with four intervals: Baseline, 1 year, 2 years, and 3 years. The y-axis represents S O F A S score ranging from 20 to 90. Individual participant data are shown as thin background lines, while bold lines represent the mean for four classes.
* Persistently-Good (dark blue): Starts at 75.4, peaks at 82.8 at 1 year, and remains high, ending at 81.6.
* Gradually-Improved (light blue): Shows a steady linear increase from 62.6 at Baseline to 71.1 at 1 year, 75.6 at 2 years, and 79.1 at 3 years.
* Suboptimal-Stable (green): Maintains a relatively flat trajectory, starting at 62.4, dipping slightly to 60.3 at 2 years, and ending at 62.0.
* Persistently-Poor (dark red): Starts at the lowest point of 41.3 and shows a slight upward trend to 47.3 at 1 year, 48.9 at 2 years, and 51.2 at 3 years.
Predictors of social-occupational functional trajectories
We first identified baseline characteristics associated with the trajectories using univariate multinomial logistic regression analysis (Table 2). Group differences were observed in PAS social and academic scores, log DUP, PANSS positive symptom, negative symptom, and disorganization scores, CDSS total score, insight, and SAS score. As reflected in Table 3, we then performed multivariate logistic regression analysis utilizing persistently-good class as reference and found that patients in persistently-good class had lower PANSS negative symptom score (p = 0.008) at baseline as compared to patients in persistently-poor class. In addition, it is suggestive that patients in the persistently-good class had shorter DUP (p = 0.053) and lower SAS scores (p = 0.073) compared to those in the persistently-poor class, and lower SAS scores (p = 0.051) compared to those in the suboptimal-stable class, although these associations did not reach statistical significance.
Baseline comparison between four social-occupational trajectories in demographics, premorbid adjustment, illness characteristics, symptom severity, functioning, and treatment profiles
Table 2. Long description
The table compares four groups: Persistently-good (N = 19), Gradually-improved (N = 16), Suboptimal-stable (N = 72), and Persistently-poor (N = 41).
* Demographics: No significant differences were found in sex, marital status, age at entry (mean 21.62 to 23.17), or education level.
* Premorbid personality and adjustment: Significant differences were found in P A S social domain (P = 0.038) and P A S academic domain (P = 0.018), with the Persistently-good group showing better adjustment (lower scores) than the Persistently-poor group.
* Illness characteristics: Log D U P (duration of untreated psychosis) was significantly different (P = 0.013), being lowest in the Persistently-good group (1.62) and highest in the Persistently-poor group (2.15). Diagnosis and psychiatric admission history showed no significant differences.
* Symptom severity and functioning: Highly significant differences (P < 0.001) were observed in P A N S S positive, negative, and disorganization scores. The Persistently-poor group had the highest symptom severity (Negative score = 16.32) compared to the Persistently-good group (Negative score = 8.42). C D S S total score (P = 0.006) and Good insight (P = 0.031) also showed significant variation.
* Treatment profiles: Significant differences were found in S A S scores (P = 0.009), with the Persistently-poor group scoring highest (3.49) and the Persistently-good group lowest (0.95). Treatment condition, use of S G A, and D A I scores were not significantly different.
Note: Bold values indicate that the results are statistically significant (p < 0.05). CDSS, Calgary Depression Scale for Schizophrenia; DAI, Drug Attitude Inventory; DUP, duration of untreated psychosis; MCS, mental component summary; PANSS, Positive and Negative Syndrome Scale; PAS, Premorbid Adjustment Scale; PCS, physical component summary; PSST, Premorbid Schizoid and Schizotypal Traits; RFS, Role Functioning Scale; SAS, Simpson-Angus Scale; SD, standard deviation; SGA, second-generation antipsychotics; SF-36, 36-Item Short Form Survey.
a Good insight is defined by PANSS G12 score ≤ 3.
Predictors of membership to functioning trajectories using persistently-good class as reference (N = 19) in multinomial logistic regression
Table 3. Long description
The table presents multinomial logistic regression results for nine variables across three trajectory groups. Each group includes Z and P values.
* Gradually-improved (N = 16):
- P A S social domain score: Z 0.89, P 0.371.
- P A S academic domain score: Z 0.34, P 0.737.
- Log D U P: Z minus 0.81, P 0.416.
- P A N S S positive symptom score: Z 0.52, P 0.606.
- P A N S S negative symptom score: Z minus 0.03, P 0.979.
- P A N S S disorganization score: Z 1.09, P 0.277.
- C D S S total score: Z 0.56, P 0.579.
- Good insight: Z minus 0.57, P 0.569.
- S A S score: Z 0.83, P 0.405.
* Suboptimal-stable (N = 72):
- P A S social domain score: Z 0.55, P 0.579.
- P A S academic domain score: Z minus 0.81, P 0.416.
- Log D U P: Z 1.24, P 0.216.
- P A N S S positive symptom score: Z 1.16, P 0.244.
- P A N S S negative symptom score: Z 1.44, P 0.149.
- P A N S S disorganization score: Z 0.64, P 0.524.
- C D S S total score: Z 0.12, P 0.906.
- Good insight: Z minus 0.91, P 0.362.
- S A S score: Z 1.95, P 0.051.
* Persistently-poor (N = 41):
- P A S social domain score: Z minus 0.33, P 0.744.
- P A S academic domain score: Z 1.35, P 0.178.
- Log D U P: Z 1.93, P 0.053.
- P A N S S positive symptom score: Z 1.35, P 0.176.
- P A N S S negative symptom score: Z 2.65, P 0.008 (statistically significant).
- P A N S S disorganization score: Z 0.85, P 0.397.
- C D S S total score: Z minus 0.02, P 0.983.
- Good insight: Z 0.02, P 0.987.
- S A S score: Z 1.79, P 0.073.
Note: Bold values indicate that the results are statistically significant (p < 0.05). CDSS, Calgary Depression Scale for Schizophrenia; DUP, duration of untreated psychosis; PANSS, Positive and Negative Syndrome Scale; PAS, Premorbid Adjustment Scale; SAS, Simpson-Angus Scale.
a Variables demonstrating statistical significance in univariate analyses were included in this multivariable logistic regression model.
b Good insight is defined by PANSS G12 score ≤ 3.
Association between trajectory classes and 12-year outcomes
Table 4 shows the relationship between trajectory classes and functional and recovery outcomes at 12-year follow-up. Patients with persistently-good trajectory had better outcome than those with persistently-poor class in the majority of functional measures, including RFS total score (p < 0.001), RFS work (p = 0.004) and extended social (p = 0.032) domain scores. Patients with persistently-good trajectory had a better SOFAS score at 12-year follow-up than patients with all other trajectories (p < 0.02). Better SOFAS (p = 0.021) and RFS (p = 0.021) total scores were also observed in patients with suboptimal-stable trajectory relative to those with persistently-poor class. Additionally, rates of attaining functional recovery at 12-year follow-up was significantly higher in patients with persistently-good trajectory (70%) as compared to patients with gradual-improving (14.3%) and persistently-poor trajectory (6.9%).
Association between social-occupational functioning trajectories and 12-year outcomes
Table 4. Long description
The table consists of seven columns: Variables, Persistently-good, Gradually-improved, Suboptimal-stable, Persistently-poor, P value (interclass X squared), and Significant differences between classes.
Key data points include:
* S O F A S scores: Persistently-good (73.80), Gradually-improved (58.57), Suboptimal-stable (60.79), and Persistently-poor (52.86). P < 0.001.
* R F S Total scores: Persistently-good (25.20), Gradually-improved (23.57), Suboptimal-stable (22.11), and Persistently-poor (19.38). P < 0.001.
* Functional recovery rates: Persistently-good (70.0 percent), Gradually-improved (14.3 percent), Suboptimal-stable (28.6 percent), and Persistently-poor (6.9 percent). P < 0.001.
* Personal recovery rates: Persistently-good (70.0 percent), Gradually-improved (71.4 percent), Suboptimal-stable (49.1 percent), and Persistently-poor (37.9 percent). P = 0.205.
Significant differences are noted between the Persistently-good group and the Persistently-poor group across most functional metrics. Initialisms used include S O F A S (Social and Occupational Functioning Assessment Scale) and R F S (Role Functioning Scale).
Note: Bold values indicate that the results are statistically significant (p < 0.05). GI, Gradually-improved; PANSS, Positive and Negative Syndrome Scale; PG, Persistently-good; PP, Persistently-poor; RFS, Role Functioning Scale; SOFAS, Social and Occupational Functioning Assessment Scale; SS, Suboptimal-stable.
Discussion
We identified four distinct trajectories of social-occupational functioning over 3 years in young people with early psychosis. Specifically, 12.8% of our participants maintained a high level of social-occupational functioning throughout the entire follow-up period (classified as persistently-good). In addition, 10.8% showed gradual improvement over time (classified as gradually-improved) whereas 48.6% exhibited stable, moderate levels of functioning across the 3-year follow-up (classified as suboptimal-stable). The remaining 27.7% of our patients displayed severe, persistent functional deficits throughout the 3-year follow-up. Our identification of a four-class solution as the best model fit is aligned with the majority of prior studies on early functional trajectories in early psychosis (Chang et al., Reference Chang, Chu, Kwong, Wong, Hui, Chan, Lee and Chen2018a; Hall et al., Reference Hall, Holton, Ongur, Montrose and Keshavan2019; Salagre et al., Reference Salagre, Grande, Solé, Mezquida, Cuesta, Díaz-Caneja, Amoretti, Lobo, González-Pinto, Moreno, Pina-Camacho, Corripio, Baeza, Bergé, Verdolini, Carvalho, Vieta and Bernardo2021; van der Ven et al., Reference van der Ven, Scodes, Basaraba, Pauselli, Mascayano, Nossel, Bello, Humensky, Susser, Wall and Dixon2020). Nonetheless, discrepant findings in trajectory patterns were observed. Some studies, including the current study, have shown a relatively parallel pattern, while others have indicated crossover patterns over time (Chang et al., Reference Chang, Chu, Kwong, Wong, Hui, Chan, Lee and Chen2018a; Hall et al., Reference Hall, Holton, Ongur, Montrose and Keshavan2019; Salagre et al., Reference Salagre, Grande, Solé, Mezquida, Cuesta, Díaz-Caneja, Amoretti, Lobo, González-Pinto, Moreno, Pina-Camacho, Corripio, Baeza, Bergé, Verdolini, Carvalho, Vieta and Bernardo2021). Notably, we found that 27.7% of our patients were categorized into the worst trajectory pattern (persistently-poor), which was at the lower-end among previous studies (Chang et al., Reference Chang, Chu, Kwong, Wong, Hui, Chan, Lee and Chen2018a; du Sert et al., Reference du Sert, Unrau, Dama, Palaniyappan, Shah, Joober, Raucher-Chéné, Malla and Lepage2025; Salagre et al., Reference Salagre, Grande, Solé, Mezquida, Cuesta, Díaz-Caneja, Amoretti, Lobo, González-Pinto, Moreno, Pina-Camacho, Corripio, Baeza, Bergé, Verdolini, Carvalho, Vieta and Bernardo2021; Velthorst et al., Reference Velthorst, Fett, Reichenberg, Perlman, Van Os, Bromet and Kotov2017). This may be attributed to the use of different functional assessments to derive trajectories and variations in the follow-up duration. Although the number of patients categorized into the worst trajectory pattern was relatively low as compared to other studies, more than one-fourth of our patients still experienced severe functional impairment, reaffirming the importance of social-occupational functioning as a critical treatment target in the early course of psychotic disorders.
Our results highlighted the importance of negative symptoms in predicting social-occupational functioning by demonstrating that greater negative symptom severity at baseline predicted a persistently-poor social-occupational functioning trajectory, which concorded with the majority of studies examining trajectories (du Sert et al., Reference du Sert, Unrau, Dama, Palaniyappan, Shah, Joober, Raucher-Chéné, Malla and Lepage2025; Hodgekins et al., Reference Hodgekins, Birchwood, Christopher, Marshall, Coker, Everard, Lester, Jones, Amos, Singh, Sharma, Freemantle and Fowler2015; Marder & Umbricht, Reference Marder and Umbricht2023; Salagre et al., Reference Salagre, Grande, Solé, Mezquida, Cuesta, Díaz-Caneja, Amoretti, Lobo, González-Pinto, Moreno, Pina-Camacho, Corripio, Baeza, Bergé, Verdolini, Carvalho, Vieta and Bernardo2021). Findings that negative symptoms have been consistently found to be a key determinant of poorer functional outcomes in early psychosis (Santesteban-Echarri et al., Reference Santesteban-Echarri, Paino, Rice, González-Blanch, McGorry, Gleeson and Alvarez-Jimenez2017) were also consistent with research on FEP patients without EI (Chang et al., Reference Chang, Hui, Chan, Lee and Chen2016a; Faerden et al., Reference Faerden, Friis, Agartz, Barrett, Nesvåg, Finset and Melle2009, Reference Faerden, Finset, Friis, Agartz, Barrett, Nesvåg, Andreassen, Marder and Melle2010; Milev, Ho, Stephan Arndt, & Andreasen, Reference Milev, Ho, Stephan Arndt and Andreasen2005) and chronic schizophrenia (Ventura et al., Reference Ventura, Hellemann, Thames, Koellner and Nuechterlein2009). In fact, negative symptoms are prevalent in the early course of the illness, with roughly one-third of FEP patients having negative symptoms during assessment (Mittal, Chang, & Walker, Reference Mittal, Chang and Walker2026). Previous research has indicated that negative symptoms predicted functional outcomes, above and beyond positive symptoms, other symptom dimensions, and cognitive impairment in early psychosis (Mittal et al., Reference Mittal, Chang and Walker2026). Persistent negative symptoms in the initial course of the illness further predicted poorer long-term functional outcomes (Chang et al., Reference Chang, Ho, Tang, Wong, Hui, Chan, Lee, Suen and Chen2019; Mittal et al., Reference Mittal, Chang and Walker2026). Unfortunately, negative symptoms remain minimally responsive to currently available antipsychotics (Correll & Schooler, Reference Correll and Schooler2020; Moccia et al., Reference Moccia, Bardi, Anesini, Barbonetti, Kotzalidis, Rossi, Caso, Grisoni, Mandracchia, Margoni, Callovini, Janiri, Mazza, Simonetti, Montanari, Autullo, Camardese, Pepe, Di Nicola and Sani2025). To date, there is still no FDA-approved medication specifically designed for targeting negative symptoms (Correll & Schooler, Reference Correll and Schooler2020). On the other hand, some studies have suggested that psychosocial interventions, such as cognitive behavioral therapy and cognitive remediation therapy, can have demonstrated positive effects in improving negative symptoms (Cella et al., Reference Cella, Roberts, Pillny, Riehle, O’donoghue, Lyne, Tomlin, Valmaggia and Preti2023; Correll & Schooler, Reference Correll and Schooler2020; Melville, Preisig, Zheng, & Kurtz, Reference Melville, Preisig, Zheng and Kurtz2025). Cognitive behavioral therapy has shown a small to moderate effect on negative symptoms, whereas longer cognitive remediation therapy with bridging activities that relate to daily living may lead to greater improvements in negative symptoms (Cella et al., Reference Cella, Roberts, Pillny, Riehle, O’donoghue, Lyne, Tomlin, Valmaggia and Preti2023; Melville et al., Reference Melville, Preisig, Zheng and Kurtz2025). More research should be conducted to verify the effectiveness of psychosocial interventions on improving negative symptoms, which may lead to the development of effective, targeted interventions for negative symptoms in the early stage of illness.
Our findings that longer DUP and more severe antipsychotic-induced extrapyramidal side-effects at baseline were associated with poorer social-occupational functioning trajectory were borderline significant. Regarding DUP, a meta-analytical review has revealed that a longer DUP was significantly correlated with poorer social functioning, albeit not with employment (Penttilä et al., Reference Penttilä, Jaä¨skel̈ainen, Hirvonen, Isohanni and Miettunen2014). Another meta-analysis observed an association between prolonged DUP and lower likelihood of attaining functional recovery; however, subgroup analysis did not reveal an association between DUP and sub-domains of functioning, specifically relationships and vocational functioning (Santesteban-Echarri et al., Reference Santesteban-Echarri, Paino, Rice, González-Blanch, McGorry, Gleeson and Alvarez-Jimenez2017). The majority of studies of FEP without an EI component and of early psychosis have reported significant associations, although a few have not (Chang et al., Reference Chang, Tang, Hui, Lam, Wong, Chan, Chiu, Chung, Law, Tso, Chan, Hung and Chen2012b; de Pablo et al., Reference de Pablo, Guinart, Armendariz, Aymerich, Catalan, Alameda, Rogdaki, Baringo, Soler-Vidal, Oliver, Rubio, Arango, Kane, Fusar-Poli and Correll2024; Larsen, Moe, Vibe-Hansen, & Johannessen, Reference Larsen, Moe, Vibe-Hansen and Johannessen2000; Robinson et al., Reference Robinson, Woerner, McMeniman, Alan Mendelowitz and Bilder2004). Of note, since our study utilized SOFAS to examine global functioning, which incorporated both role and social functioning, we may not be able to directly compare our current findings with results from prior research. Inconsistent findings were also reported in studies using a trajectory analysis approach, with some reporting a significant association between prolonged DUP and poor functioning trajectory (Chaudhry et al., Reference Chaudhry, Weiss, Mojtabai, Farrow, Shrestha and Long2025; du Sert et al., Reference du Sert, Unrau, Dama, Palaniyappan, Shah, Joober, Raucher-Chéné, Malla and Lepage2025; Salagre et al., Reference Salagre, Grande, Solé, Mezquida, Cuesta, Díaz-Caneja, Amoretti, Lobo, González-Pinto, Moreno, Pina-Camacho, Corripio, Baeza, Bergé, Verdolini, Carvalho, Vieta and Bernardo2021), while others failed to find any relationship (Chang et al., Reference Chang, Chu, Kwong, Wong, Hui, Chan, Lee and Chen2018a; Hall et al., Reference Hall, Holton, Ongur, Montrose and Keshavan2019; Hodgekins et al., Reference Hodgekins, Birchwood, Christopher, Marshall, Coker, Everard, Lester, Jones, Amos, Singh, Sharma, Freemantle and Fowler2015). As for the extrapyramidal side-effects, very few studies have examined their relationship with functioning. One study with cross-sectional design suggested that extrapyramidal side-effects were associated with social functioning (Gao et al., Reference Gao, Huang, Huang, Zhou, Shi, Yu and Pu2022). Another study with longitudinal design showed that patients with schizophrenia who had more severe extrapyramidal side-effects tended to have less improvement in their social functioning over time (Park et al., Reference Park, Jung, Hwang, Jung, Yoon, Kim, Ahn and Kim2016). Notably, our findings should be interpreted with caution, as the results were only borderline significant. Thus far, no study has examined the effect of extrapyramidal side-effects on social-occupational functioning with the use of trajectory analysis. More studies are needed to verify our findings. Moreover, we did not observe any positive effect of extended EI in our trajectory analysis. This is, nonetheless, in line with our previous post-RCT follow-up study which revealed that the beneficial effect of extended EI on functional outcomes could not be sustained 2 years after discontinuation of specialized care (Chang et al., Reference Chang, Chan, Jim, Lau, Hui, Chan, Lee and Chen2015, Reference Chang, Kwong, Kei Lau, So, Man Wong, Kei Chan, Ting Jim, Ming Hui, Wa Chan, Ming Lee and Hai Chen2017a).
To the best of our knowledge, this is the first study examining the prospective relationships between early-stage functioning trajectories in an early psychosis sample and functional outcomes at long-term follow-up. Our results indicated that, among the four trajectory classes, patients in the persistently-poor trajectory exhibited worse functional outcomes encompassing poorer global functional levels, and discrete functional subdomains, as well as a lower likelihood of achieving functional recovery at 12-year follow-up, whereas patients with persistently-good trajectory class had the best functional outcomes in the majority of domains. Thus, our findings indicate that early-course social-occupational functioning trajectory patterns represent a valuable prognostic indicator of long-term functional outcomes and underscore the importance of social-occupational functioning as a key treatment target in EI for psychosis.
Several methodological limitations should be warranted when interpreting our findings. First, the modest sample size of our cohort, especially in the persistently-good and gradually-improved classes, may compromise the statistical power to detect subtle yet potentially significant baseline predictors of trajectory membership. Attrition at 12-year follow-up may further compromise the statistical power to detect potential outcome associations. Second, the generalizability of our results to older early psychosis patients may be limited since the EASY program exclusively treated only patients aged 15–25 years at the time the study was conducted. Third, variables that may be relevant to social-occupational functioning, namely cognitive impairment, self-stigma, and self-efficacy, were not evaluated in the current study (Chu et al., Reference Chu, Ng, Chu, Lui, Lau, Chan, Lee, Hui, Chen, Lui and Chang2023; Santesteban-Echarri et al., Reference Santesteban-Echarri, Paino, Rice, González-Blanch, McGorry, Gleeson and Alvarez-Jimenez2017; Wright et al., Reference Wright, Browne, Cather, Pratt, Bartels and Mueser2021). Fourth, the participants included in the trajectory analysis had lower PAS and RFS scores compared with those who were excluded, suggesting that patients with poorer premorbid adjustment and role functioning and potentially worse treatment outcomes may be under-represented in this investigation. This attrition bias may contribute to an overestimation of the proportion of patients allocated to persistently-good and gradually-improved trajectories. Fifth, only 102 of the initial 160 patients completed the 12-year follow-up. Attrition analysis suggested that patients included in the 12-year outcome analysis had fewer PSSTs, indicating that patients with poorer premorbid personality and potentially worse functional outcomes may be inadequately represented in our study, which may in turn lead to an underestimation of poor outcomes in the persistently-poor trajectories (Chu et al., Reference Chu, Fung, Lei, Wong, Chan, Lee, Hui, Chen and Chang2026a). Sixth, this study was not pre-registered on the Open Science Framework. This may limit the transparency and reproducibility of the study. Seventh, although the LGMM provides a novel, data-driven approach to identify subgroups, its results may pose interpretability challenges. The model may extract classes that are mathematically distinct, but not theoretically meaningful or clinically useful.
In conclusion, less than one-fourth of the participants showed persistently-good functioning or improvement in early-course functioning, with the majority of patients demonstrating suboptimal or poor functional levels. Patients in the persistently-poor functional trajectory class had greater severity of negative symptoms at baseline as compared to those in the persistently-good class. Notably, more than one-fourth of the participants exhibited a persistently poor social-occupational functioning trajectory, which, among the four identified trajectories, predicted the worst functional outcomes at 12-year follow-up. These findings highlight the importance of tracking functional changes during the initial years of illness. In hope of improving long-term functional outcomes, more intensive interventions targeting functional improvement, such as individual placement and support, are needed.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291726105741.
Data availability statement
The data that support the findings of the study are available from the corresponding author upon reasonable request.
Acknowledgements
The authors thank all the coordinating clinicians and staff from the psychiatric units involved in the study. We are also grateful to the individuals who participated in the study.
Author contribution
Conceptualization: W.C.C., E.Y.H.C.; Data collection: V.S.C.F., J.H.C.L., G.H.S.W.; Designed the study: W.C.C., E.Y.H.C.; Finalized: W.C.C., R.S.T.C.; Project implementation: E.H.M.L., C.L.M.H., S.K.W.C.; Statistical analysis: C.C.Z.F., N.W.T.C., R.S.T.C.; Supervision: E.H.M.L., C.L.M.H., S.K.W.C.; Wrote the initial draft: C.C.Z.F., N.W.T.C., R.S.T.C. All authors critically reviewed the report and approved the final submitted version. All authors had final responsibility for the decision to submit for publication.
Funding statement
This study was supported by a grant from the Food and Health Bureau (SMH-29) and Seed Fund for Basic Research (202111159039) of the University of Hong Kong. The funding bodies had no involvement in any aspect of the study or manuscript preparation.
Competing interests
The authors declared no competing interests. The authors alone are responsible for the content and writing of the paper.
Open access
