Three-year outcomes and predictors for full recovery in patients with early-stage psychosis

The study addresses how recovery, remission, and relapse unfold over three years in early-stage psychosis, with a focus on first-episode psychosis (FEP). Early-stage psychosis was defined as within ≤2 years of antipsychotic treatment. Given variability and gaps in prior research, the authors aimed to quantify remission, recovery, dropout, relapse, medication adherence, and LAIA use trajectories in a Korean national cohort. They also sought to test predictors for full recovery at year 3, including childhood trauma and rumination, in more homogeneous diagnostic groups (schizophrenia spectrum disorders and PNOS). Understanding these trajectories and predictors is important to guide mental health services and targeted interventions to improve outcomes after FEP.

Prior work in FEP shows wide ranges of symptomatic remission: a systematic review reported a mean of 36% (17–78%) over 6 months to 7 years, and a later meta-analysis reported 58% over a mean 5.5 years. Recovery estimates vary: 17% in schizophrenia over 10 years and 38% in FEP over 7.2 years. Predictors of symptomatic remission and recovery include better premorbid functioning, milder baseline symptoms, early treatment response, and shorter duration of untreated psychosis (DUP). Other recovery-associated factors: female sex, higher education, better neurocognition, premorbid IQ, better occupational status, being married, good medication adherence, favorable personality, and fewer baseline negative symptoms. Relapse after FEP is common, with positive symptom relapse prevalence around 28% at 1 year, 43% at 1.5–2 years, 54% at 3 years, and 80% at 5 years. Risk factors for relapse include medication non-adherence (4-fold), persistent substance use disorder (3-fold), critical caregiver comments (2.3-fold), and poorer premorbid adjustment (2.2-fold). Oral antipsychotic discontinuation rates in FES are high (42% at 12 months), suggesting LAIA can help, yet LAIA use post-first hospitalization for schizophrenia remains low (8–10%).

Design and setting: The Korea Early Psychosis Study (KEPS) is a prospective naturalistic observational cohort conducted across 11 hospitals (December 2014–December 2021). For this analysis, data from January 2015 to July 2020 were used. Participants: 657 patients with early-stage psychosis were enrolled; early-stage was defined as adequate antipsychotic treatment for >4 weeks and ≤2 years. Analyses were restricted to schizophrenia spectrum disorders (schizophrenia, schizophreniform disorder) and psychotic disorder not otherwise specified (PNOS), yielding 534 participants with variable follow-up durations. Diagnoses and ethics: Diagnoses followed DSM-IV criteria and the Korean MINI. Two experienced psychiatrists at each site reached consensus. Written informed consent was obtained; ethics approval: CUH 2014-11-002. Assessments and timepoints: Baseline sociodemographics and clinical variables included: family history of psychosis, DUP (time from psychosis onset to antipsychotic initiation or hospitalization), duration of illness, PANSS, CGI, SOFAS, CDSS, C-SSRS, and comorbid mental disorders. Clinical psychopathology was evaluated at baseline, 2, 6, 9, 12, 18, 24, 30, and 36 months; CDSS and C-SSRS at baseline, 6, 12, 24, and 36 months. Inter-rater reliability was supported by standardized training and workshops. Self-rated measures at baseline: BFI-10, BCSS, BRS, Brooding Scale, Diet History Questionnaire (categorized poor 20–49, usual 50–79, good 80–100), ETISR-SF, FACES-III, Family Intimacy (FI; 5-point Likert), Physical Activity Rating (PAR; frequency and intensity), and K-SWN. Operational definitions: Symptomatic remission: PANSS P1, P2, P3, N1, N4, N6, G5, G9 ≤3. Full recovery required both (a) symptomatic recovery (≤2 on the same PANSS items) and (b) functional recovery (adequate social interaction and occupational/role functioning). Required durations: remission 2 months at 6-month follow-up and 6 months at 12/24/36 months; for full recovery, 2 months at 6-month, 6 months at 12-month, and 12 months at 24- and 36-month follow-ups. Adherence categories used thresholds of Good (≥80), Fair (50–79), Poor (≤49). LAIA prescriptions were tracked, with agents and chlorpromazine-equivalent doses recorded. Sample sizes at follow-up (eligible/assessed varied): approximately n=382 at 6 months, 332 at 12 months, 255 at 24 months, 157 at 36 months for key outcomes (e.g., adherence in Table 3). Statistical analysis: Stepwise logistic regression on year-3 completers (n=157) identified predictors of full recovery. Subgroup analyses were conducted for schizophrenia, schizophreniform disorder, and PNOS. Between-group comparisons of follow-up vs. dropout were examined (Table S2).

- Symptomatic remission rates at 6, 12, 24, 36 months: 76.10%, 69.20%, 79.50%, 79.10%. - Full recovery rates at 6, 12, 24, 36 months: 22.80%, 26.40%, 28.60%, 39.60%. - Cumulative dropout rates at 6, 12, 24, 36 months: 25.4%, 29.5%, 38.6%, 51.1%. - Relapse rates at 6, 12, 24, 36 months: 3.7%, 8.9%, 19.0%, 38.9%. - Good adherence (≥80) at 6, 12, 24, 36 months: 87.8% (318/362), 88.0% (271/308), 91.9% (217/236), 93.9% (138/147). PNOS subgroup had lower adherence across follow-ups. - LAIA prescription: 9.7% at baseline; 18.3% (6 m), 21.7% (12 m), 22.0% (24 m), 25.5% (36 m). Aripiprazole once-monthly and paliperidone palmitate once-monthly were the primary agents. - Mental health welfare center use among eligible patients remained low (users: 7.4% at baseline; 12.4% at 6 m; 11.0% at 12 m; 10.4% at 24 m; 12.1% at 36 m), despite a sizable proportion being candidates (44–60%). - Predictors of full recovery at year 3 (n=157): shorter DUP (log[DUP+1] OR 0.534, 95% CI 0.364–0.758; p=0.001), higher Family Intimacy (FI) (OR 2.262, 95% CI 1.361–3.967; p=0.003), higher PAR intensity (OR 1.231, 95% CI 1.013–1.512; p=0.040). In schizophrenia subgroup, additional negative association with PANSS-positive (OR 0.929, 95% CI 0.865–0.991; p=0.033).

Observed remission (≈69–80%) and full recovery (≈23–40%) over 3 years are similar to or better than prior reports, despite stringent recovery criteria (PANSS core items ≤2 plus functional criteria). Early advantages in remission/recovery for schizophreniform disorder and PNOS at 6–12 months diminished or reversed by 24–36 months for PNOS, potentially reflecting higher dropout leaving more severe cases. Dropout rates align with prior disengagement estimates (~20–40%), underscoring the need for robust, hospital-based case management to maintain engagement. Relapse rates were relatively low compared to some literature; possible explanations include high measured adherence and moderate LAIA use, though adherence estimates excluded dropouts without ratings. LAIA utilization (18–26%) exceeded regional averages, suggesting potential contribution to stability and further room for improvement. Low uptake of community mental health welfare centers despite many eligible patients suggests ineffective referral pathways or acceptability barriers among clinicians, patients, or service limitations. Key predictors of full recovery—shorter DUP, stronger family intimacy, and greater physical activity intensity—highlight the importance of early detection/treatment, family-focused interventions, and exercise programs. Frequency of physical activity was not associated with recovery, suggesting intensity may be more relevant. In schizophrenia, lower positive symptom burden additionally predicted recovery.

Over three years in early-stage psychosis, remission and recovery rates were favorable relative to previous findings, yet a persistent gap between symptomatic remission and functional recovery remains. Targeted psychosocial interventions are needed to bridge this gap. Specifically, strategies to reduce DUP via early intervention and streamlined referral, enhance family intimacy/support, and promote structured, adequately intense physical activity may improve recovery outcomes. Broader use of LAIA and improved linkage to community mental health resources could further reduce relapse and disengagement.

- Outcome calculations excluded dropouts for remission, recovery, relapse (dropouts without relapse not counted), and adherence, potentially biasing estimates; medical record review is planned to address this. - Key prognostic variables such as premorbid functioning and cognition were not collected, which may limit predictive modeling. - The broad early-stage definition may have introduced heterogeneity, warranting cautious interpretation. - Adherence was not corroborated with family informants, possibly affecting validity. - Childhood trauma was assessed via self-report (ETISR-SF); interview-based assessments could improve reliability.